Lab of the Future: Digitization, Automation and Lab Sustainability

SPONSORED BY Lab of the Future: Digitization, Automation and Lab Sustainability Data Reporting: Connecting the Islands of Automation Towards the Lab of the Future How To Future-Proof Your LIMS: Handling Software Updates Credit: iStock Technologies Helping To Drive Digital Transformation and Laboratory Efficiencies 4 How To Reduce Data Integrity Risk 8 Towards the Lab of the Future 13 New Advances in the Bioprocess Pipeline 16 Infographic: The High-Throughput Technologies Speeding Up Biomedical Science 19 Data Reporting: Connecting the Islands of Automation 21 The Future of Digital Health With Professor Michael Snyder 25 Lab Sustainability: A Holistic Approach 27 Failing Faster, Succeeding Sooner: Technologies That Are Accelerating Drug Discovery 30 How To Future-Proof Your LIMS: Handling Software Updates 35 Contents 2 TECHNOLOGYNETWORKS.COM SYNTHIA™ Retrosynthesis Software is a unique web-based platform that revolutionizes the way organic chemists design pathways to complex targets. This decision-support tool is the strategic partner of synthetic organic chemists. Pioneer Your Pathway Go Beyond the literature www.SynthiaOnline.com MK_AD12813EN Vers. 01 09/2023 © 2023 Merck KGaA, Darmstadt, Germany and/or its affi liates. All Rights Reserved. Merck, the vibrant M, Sigma-Aldrich and SYNTHIA are trademarks of Merck KGaA, Darmstadt, Germany or its affi liates. All other trademarks are the property of their respective owners. Detailed information on trademarks is available via publicly accessible resources. The Life Science business of Merck operates as MilliporeSigma in the U.S. and Canada.in the U.S. and Canada. 4 TECHNOLOGYNETWORKS.COM Technologies Helping To Drive Digital Transformation and Laboratory Efficiencies Lee Baker The digital transformation of laboratories is a rapidly evolving field, with advances in automation and digitization fundamentally changing the way labs are designed, built and run from the ground up. Several technologies are helping to drive this transformation, ranging from artificial intelligence (AI) to laboratory information management systems (LIMS). Overall, the integration of these technologies is enabling labs to operate more efficiently, with increased speed, accuracy and flexibility. Labs will be able to leverage these advances to accelerate the pace of scientific discovery and drive innovation in their respective fields. In this article, we highlight some of the key technologies fueling digital transformation and explore the impacts they are having on laboratory efficiencies. Digitizing the laboratory Modern lab managers are acutely aware of the benefits that digitization can bring to their operations. Firstly, digitization improves data management by providing a centralized platform for storing, accessing and analyzing data. This allows for efficient data sharing and collaboration among researchers, facilitating faster decision-making and accelerating scientific discovery. Additionally, digitization enhances laboratory efficiencies by automating manual tasks and streamlining workflows. It reduces human errors, increases productivity and frees up researchers’ time to focus on more complex and critical tasks. Moreover, digitization enables real-time monitoring and analysis, providing valuable insights and enabling proactive decision-making. This leads to improved quality control, optimized resource utilization and reduced costs. Furthermore, digitization enhances compliance and traceability by maintaining comprehensive records and audit trails. It supports reproducibility and data integrity. Overall, digitization offers laboratories the potential to improve their research capabilities, enhance collaboration, increase efficiency and accelerate scientific progress. Advances in digital technologies such as LIMS and electronic lab notebooks (ELNs) are already helping to drive the digitization of laboratories and improve the accuracy and reproducibility of scientific research. Improving data quality with LIMS LIMS are comprehensive software systems designed to manage laboratory workflows, data and samples, offering significant advantages to laboratories. One key benefit of LIMS is their ability to streamline operations. LIMS provide a centralized platform for managing all aspects of laboratory activities, from sample collection and tracking to data analysis and reporting. By automating routine tasks such as sample registration, labeling and tracking, LIMS eliminate the need for manual data entry and reduce Lab of the Future Credit: iStock 5 TECHNOLOGYNETWORKS.COM Lab of the Future Credit: iStock the chances of human errors. This automation leads to increased productivity as laboratory staff can focus on more critical tasks and scientific analysis. Furthermore, LIMS play a crucial role in improving data quality within laboratories. By standardizing data collection methods and enforcing data integrity rules, LIMS ensure consistent and accurate data entry. Real-time access to data allows researchers to retrieve information quickly, facilitating decision-making and enabling timely responses to experimental outcomes. Moreover, LIMS offer features like data validation and data auditing, enabling better quality control and compliance with regulatory requirements. The versatility of LIMS extends beyond data management. They support workflow automation, ensuring that laboratory processes follow predefined protocols and reducing the potential for human error. With customizable workflows, laboratories can define specific steps and procedures, resulting in standardized and efficient operations. LIMS can also aid in compliance with regulatory guidelines and industry standards by providing traceability and audit trails for samples, tests and processes. Additionally, LIMS generate comprehensive reports, which are essential for documentation, analysis and sharing results. Researchers can easily retrieve and compile data for analysis, track experimental progress and generate reports that meet specific requirements. This functionality saves time and effort, streamlining the process of reporting findings to stakeholders or regulatory bodies. “Labs produce and gather huge volumes of data,” says Dr. Becky Upton President of the Pistoia Alliance. “These datasets have the potential to give researchers much deeper insight into their research questions and to make new connections, but in order to gain insights and answers from this data, we need to make sure it is stored according to best practice FAIR principles (findable, accessible, interoperable and reusable) and make it more easily accessible through semantic enrichment,” she adds. LIMS – and especially cloud-based LIMS – can help labs achieve this by providing a centralized repository for data and by making it easy to search and analyze data. Additionally, LIMS can help labs comply with regulations by tracking and recording data in a secure and auditable way. Dr. Andrew Buchanan, principal scientist at AstraZeneca, underscores that the keys in driving digital transformation are “the ability to barcode and track individual molecules from sequence, format, batch expression/QC and the functional data that is generated with the asset. [These] are essential for how we deliver the work flows we operate. Without this, the error rate is high.” Moving to the cloud Cloud computing refers to the delivery of computing services over the internet, enabling remote access to data, software and resources. In the laboratory, cloud-based LIMS can revolutionize data management and analysis. By leveraging cloud computing, labs can store and analyze large datasets, collaborate with colleagues and access software and resources on-demand. Researchers can conveniently access data and software from anywhere with an internet connection, promoting remote collaboration and enhancing flexibility. Furthermore, cloud computing offers cost-saving advantages by reducing the need for traditional IT infrastructure, such as servers and storage devices. Instead of investing in expensive hardware and software, labs can utilize cloud computing services and pay for computing resources as needed. This cost-effective approach allows labs to allocate resources efficiently while leveraging the power and scalability of cloud-based solutions. For example, Carnegie Mellon University, in conjunction with an industrial partner, recently built the world’s first university cloud lab. In this endeavor, Carnegie Mellon’s cloud lab is spearheading groundbreaking efforts to harness the full potential of cloud computing. With a specific focus on data management, the lab is developing cutting-edge methodologies and tools that enable seamless storage, retrieval and analysis of vast datasets. This approach not only streamlines research processes but also facilitates collaboration by empowering researchers to access and collaborate on data from any location with internet connectivity. By leveraging cloud computing infrastructure, the lab is optimizing resource allocation, enhancing computational capabilities, and driving efficiencies across various scientific and technological endeavors. Carnegie Mellon’s researchers plan to use the digital space to pave the way for advancements in fields ranging from AI and data analytics to high-performance computing and Internet of Things (IoT) applications. Optimizing experimental design with AI Over the years, AI has undergone significant evolution, transforming the way scientists approach research and experimentation. With its ability to learn from data, recognize patterns and make informed decisions, AI has become a valuable tool in scientific endeavors. It has the potential to revolutionize various aspects of scientific research and has already found applications in drug discovery and diagnostics. One area where AI excels is in optimizing experimental design. By simulating different scenarios and predicting outcomes, AI can assist researchers in identifying the most 6 TECHNOLOGYNETWORKS.COM Lab of the Future promising avenues of research. This helps in reducing the time and cost associated with experimental design, enabling scientists to focus their efforts on areas with higher potential for success. For example, researchers at the Broad Institute of MIT and Harvard are using AI to design experiments to solve a gene therapy problem. The Broad Institute’s AI screening technology, Fit4Function, was 90% successful at finding viral vectors that don’t cause disease but can deliver potentially life-changing gene therapies to specific cells in the body. In drug discovery, AI is being utilized to accelerate the identification and development of new therapeutic compounds. Machine learning (ML) algorithms can analyze vast amounts of data, such as chemical structures and biological properties, to identify patterns and predict the efficacy of potential drug candidates. This enables researchers to prioritize the most promising compounds for further investigation, saving time and resources in the drug development process. Natural Language Processing (NLP) is a crucial component of AI and ML. NLP enables machines to understand and contextualize human language, opening up new possibilities for scientific research. “In the future, NLP-driven AI and ML will augment human researchers in the lab, enabling them to uncover new relationships between data and automate large analyses to accelerate research and increase the success of drug development,” explains Dr. Upton. While automation tools and machine learning algorithms are powerful, Dr. Buchanan notes that they are not intelligent; “For me the intelligence comes from human insight and the ability to ask the right question. Once the question is agreed, then these tools come into play and hopefully enable new more powerful insights and discoveries.” By combining human expertise with AI capabilities, scientists can leverage the full potential of AI in advancing scientific knowledge and innovation. Integrating technologies in the laboratory of the future The adoption of advanced technologies in the lab of the future is not without its challenges. Labs often face several barriers when considering the implementation of these solutions. One significant barrier is the initial cost associated with acquiring and implementing new technologies. Advanced systems such as LIMS, robotics, IoT devices and AI software may require substantial investments in infrastructure, equipment and training. Labs must carefully evaluate the costs and benefits of adopting these technologies and ensure that they align with their budget and long-term goals. Another challenge is the compatibility and integration of new technologies with existing systems and workflows. Labs may already have established processes and legacy systems in place, making it difficult to seamlessly integrate new technologies. Ensuring compatibility, data interoperability and a smooth transition from old to new systems requires careful planning and coordination. Additionally, labs must address concerns regarding data privacy, security and regulatory compliance. As technologies like cloud computing and IoT involve the collection, storage and transmission of sensitive data, labs must ensure robust cybersecurity measures and adhere to regulatory requirements to protect the integrity and confidentiality of their data. Furthermore, the adoption of advanced technologies often requires a cultural shift and changes in the mindset of lab personnel. Researchers and staff may need to acquire new skills and adapt to new ways of working. Training and education programs must be implemented to familiarize lab members with the new technologies and build their confidence in utilizing them effectively. Collaboration and knowledge sharing among labs can also be a challenge. Labs working in isolation may struggle to access shared resources, collaborate on projects, or exchange best practices. Overcoming these barriers requires establishing networks, partnerships and platforms that facilitate communication, collaboration and the sharing of resources and knowledge. Other technologies driving the lab of the future Robotics: Automated systems in the lab perform repetitive tasks, improving precision and freeing up researchers’ time. They offer advantages like increased accuracy, reproducibility and the ability to handle hazardous materials or sterile conditions. Internet of Things (IoT): IoT connects devices and sensors, enabling real-time data collection and analysis. It optimizes lab operations, improves quality control, automates routine tasks and provides critical insights into experiments and processes through data on environmental factors. Virtual and Augmented Reality (VR/AR): VR immerses users in simulated environments using head-mounted displays and interactive devices. AR overlays digital information onto the real world through mobile devices or smart glasses. Both VR and AR enhance data visualization, accelerate discovery and improve efficiency by providing immersive and interactive experiences for researchers. 7 TECHNOLOGYNETWORKS.COM Lab of the Future Credit: iStock Dr. Buchanan recommends having an expert team of automation engineers and data scientists to implement these technologies that “should integrate well with experimental colleagues, scripting existing protocols, suggesting modifications more suited to automation. These protocols need to be established, validated and maintained.” It is interesting to note that the COVID-19 pandemic has significantly impacted workflows in labs and accelerated the adoption of digitization in scientific research. With physical distancing measures and restricted access to lab facilities, researchers had to find alternative ways to continue their work and maintain productivity. This led to a greater reliance on digital tools and technologies. One key aspect of digitization that became crucial during the pandemic was remote access to data and software. Labs quickly embraced cloud-based solutions to enable researchers to access their data and analysis tools remotely, ensuring that research projects could continue even when physical access to the lab was limited. Furthermore, labs increasingly adopted LIMS to streamline data collection, storage and analysis. Digital solutions provided centralized repositories for data, improved data integrity and enabled efficient collaboration across teams, even when physically dispersed. These experiences will likely have a lasting impact, with labs continuing to leverage digital technologies to enhance efficiency, collaboration and adaptability in the postpandemic era. Dr. Upton highlights the impact of the pandemic on labs, which accelerated the field of automation and robotics, and notes that we can expect even more in the next 12 months; “The COVID-19 pandemic encouraged many labs to explore the use of automation, robotics and VR and AR. Huge benefits are to be reaped from standardizing and connecting up laboratory data, making research more efficient and less costly, while the rise in automated labs that can be operated 24 hours a day will ultimately lead to significant productivity gains. Over the next 12 months we expect to see even greater levels of automation and labs emerging that are completely driven by AI-powered robotics.” OpenAI’s ChatGPT could be a game-changer, says Upton, but “with any new technology, from the outset it’s important to take the time to understand the potential risks and implications of its use.” Transforming the laboratory for the better LIMS, robotics, IoT, AI, cloud computing and VR/AR are powerful tools for labs looking to optimize their operations, improve their data quality and streamline their workflows, making them critical technologies driving digital transformation in laboratories. By optimizing experimental design, providing real-time data and analytics, automating routine tasks, reducing errors and freeing up researchers’ time, these technologies can significantly improve laboratory efficiencies, ultimately accelerating scientific discovery and innovation. On the lab of the future, Dr. Upton insists that “technology has the potential to transform the laboratory for the better, but only if our industry is prepared to invest in the people and processes that must accompany it.” 8 TECHNOLOGYNETWORKS.COM How To Reduce Data Integrity Risk Bob McDowall, PhD Data integrity problems continue to plague the pharmaceutical industry – especially in regulated laboratories – with violations still being identified by health authority inspectors over 18 years after the Able Laboratories fraud case.1 This is despite guidance issued by various regulators2,3,4,5,6 and industry bodies7,8,9,10,11 that include advice for ensuring that data and records are protected and comply with applicable regulations. The aim of this guide is to provide practical advice on how to reduce data integrity risk and ensure GxP compliance. We will learn from those in the industry who are expert at failing to comply with data integrity regulations and who have been on the receiving end of an FDA inspection. We will use citations from a recent and extensive 483 observation issued to a single organization in December 2022.12 We will present and analyze selected citations to identify the reasons for failure. We will then suggest ways to prevent reoccurrence or, for readers who are more proactive, prevent the issue from arising in their own laboratories. Some remediation suggestions will involve laboratory automation. Before senior management has collective heart failure in response to this proposal, know that automation will be accompanied by improvements in business benefits that will more than offset the cost, including validation. The approaches outlined here could also be useful for nonregulated laboratories to improve efficiency. Manual colony counting Observation 1 item 2 in the citation targeted the practice of manual colony counting on microbiological plates. This process is slow and error prone and inevitably generates regulatory interest. The relevant section of the citation reads: Laboratory records do not include complete data derived from all tests … to assure compliance with established specifications and standards. For example, 1. Environmental monitoring samples were not counted accurately. On November 22, 2022, review of plates from QC1 that had been counted by one analyst and checked by a second analyst found the reported result to be less that the number of colonies on the plate. 2. Microbiology personnel reported the laboratory practice was to count colonies that merge together, with similar morphology, as one colony. Review of plates showed this practice resulted in under counting of colonies. 12 Under reporting of colonies means that potential out of specification (OOS) results are hidden and failing batches are passed or rooms that are meant to be clean become potential sources for microbiological contamination. After manual counting and a second-person review, the plates are thrown away after with no record other than what is written down in the analytical batch record. This provides no objective record for an inspector or an auditor to assess if the recorded value is correct. The inspectors were able to review current plate counting on a specific day before the plates were discarded. Lab of the Future Credit: iStock 9 TECHNOLOGYNETWORKS.COM Lab of the Future Credit: iStock What the citation does not state is by how much the counts were under reported or if it was deliberate. However, counting merged colonies is unscientific and deliberate falsification. A solution is to use an automated colony counter that also produces a photographic record for each plate counted. Simple manually fed colony counters can read and photograph one plate at a time. Depending on the number of plates to be read, an automated plate feeder in combination with a bar code reader to positively identify plates would be a better option as it removes labor from the process. Instrument purchase and validation costs will be offset with the gains in productivity. Lack of instrument printouts An apocryphal good manufacturing practice (GMP) saying is that if it’s not written, it’s a rumor, which can be modified for this example, to if it’s not printed, it’s a rumor, as seen in the next 483 citation from observation 1 item 3: b) pH printouts and recording of description associated with the <redacted> and <redacted> samples from <redacted> could not be provided. 12 This is gross incompetence indicating a lack of basic GMP training and a total failure of the pharmaceutical quality system. Systematic failures here were: 1. The analyst failed to document their work and failed to print the pH values obtained for an instrument check and sample analysis. 2. A failure of design in the analytical batch record, which should have prompted the analyst to document these results. 3. The second-person reviewer failed to identify the lack of documented evidence. 4. Training of the performer and reviewer has failed totally. 5. Procedures were not followed or not created in the first place. 6. Quality oversight is conspicuous by its absence. If this is the situation for pH measurement, what other data chasms exist in this laboratory, especially for more complex analyses? I’m not convinced that retraining analysts and reviewers would work and therefore automation is a prerequisite to solve the problem. Long-term remediation requires an informatics solution to automate and enforce working practices and ensure that data are collected from any interfaced instruments. The analyst would not be able to proceed with the analysis unless the instrument measurement was captured by the application. A solution should not require an analyst to press a SEND button on the instrument, as this allows them to select a “correct” result and all that is gained is electronic data falsification. Automate the whole process. Paper records would also be eliminated by taking this approach, which would speed the analysis and review but also enable verifiable quality oversight. Uncontrolled manual chromatographic peak integration In chromatography, peak integration is a critical process requiring control of manual integration parameters due to multiple instances of falsification found during inspections since 2005.1 This is an area of great regulatory interest. Inspection of chromatograms and peak integration revealed inconsistencies that led to observation 2, item 1: There is no procedure describing the use of manually entered integration events, including baseline points, tailing sensitivity and peak slice for processing chromatography data … The reviewers only review the final chromatogram and do not review the processed chromatogram to ensure that the manually entered integration events are justified. Procedure SE/BQC/00165 Interpretation of Chromatograms requires manual integration be documented clearly stating the reason the manual integration was performed and the initials of the section head for approval. But when analysts manually enter integration events to force the software to integrate in a specific way, there is no similar documented justification and approval process ... For example (this is one of several observations that has been selected for discussion here): ... Additionally, the 6-month accelerated time point for the same lot <redacted> was integrated manually by adding a fronting sensitivity and a tailing sensitivity factor to the peak for impurity <redacted> but not for the standard of the same impurity. This reduced the area of the impurity compared and gave a result of <redacted>% compared to a limit of <redacted>%. When the fronting and tailing sensitivity factors are removed to ensure integration of the impurity compared with the standard, the reportable result changes to <redacted>%, a value that would have required an investigation. 12 First and foremost, there needs to be a procedure for integration of chromatographic peaks that is applicable to all laboratory staff. Changes made by an analyst must be justified. The journey from first to last processing must be traceable and complete to ensure data integrity. While there is a procedure, it was not followed. The major problem with the laboratory work is a failure to be scientifically sound as required by 21 CFR 10 TECHNOLOGYNETWORKS.COM Lab of the Future 211.160(b): … the establishment of scientifically sound and appropriate specifications, standards, sampling plans, and test procedures …13 Chromatography is a comparative and not an absolute analytical technique, therefore standards and samples must be integrated the same way. As the inspectors found, treating samples and standards differently has hidden an out-of-specification result and avoided an investigation. This is data falsification. We have moved on from test injections in the early days of data integrity violations to subtle changes in integration that may be difficult for an auditor or inspector to identify unless they take sufficient time to look in detail at manual peak integration. Resolution may be found by using technical controls to limit manual integration for main assays within a chromatography data system. However, there is a problem with impurity analysis as we may be very close to limits of quantification and manual integration maybe the only way of measuring peaks. Training is essential along with effective review and quality oversight, coupled with management leadership about what are acceptable and unacceptable practices. However, there is much doubt if this company can achieve this. Instrument logbooks are critical for data integrity Instrument maintenance and use logbooks are mandatory under both US and EU GMP regulations13,14 and are the unsung heroes of data integrity. Logbooks provide corroborating evidence of activities performed with instrumentation and are also used to correlate activities in the controlling computerized system including the audit trail entries with printouts from the system. If the system is electronic or there are only printouts from an instrument, the role of the logbook remains the same. The 483 observation 3 item 1(a) is quoted at some length for you to understand that instrument logs must be completed and analysts must be trained to do this every time with no exceptions. Reviewers also must be trained to ensure that entries are complete and consistent. As for QA and Management they are conspicuous by their absence except when it comes to making excuses. There were no weight-specific entries in your LIMS Instrument Usage Logbook for the time recorded in any of the balance printouts. Your Associate Executive Vice-President of Corporate Quality and Compliance and Manager of QC stated that the weighing activities recorded on the Instrument Usage Log is not a true representation of samples weight with start and end time for each weighing activities for a specific lot of a product. Further, there is no consistency among QC employees in term of recording information in LIMS logbook pertaining to a total number of lots tested. Some QC employees may enter this information whereas others may not, leaving no traceability for the exact number of lots tested and their start and end time of analysis. This issue is applicable to all analytical instrument in your QC laboratory … For example, the time stamp on each of the balance printout and <redacted> spectrum did not match with your LIMS Instrument Usage Log record for samples weighed and analyzed. There follows a list of time differences, torn balance printouts, suspicious balance weights between “real” and torn printouts. 12 To have an associate executive vice-president of corporate quality and compliance state to an FDA inspector that only some analysts complete logbook entries implies that the individual knows of the problem. The individual is in a senior management role with the power to do something but has done nothing about it. The main problem is that instrument logbooks are still paper and must be completed manually. A suggested fix involves incorporating automatic instrument logs, like audit trails, into instrument data systems but software suppliers only respond to market forces. Have you got a GMP shredder? I leave the worst citation to the last as Observation 3, item 1 has the following citation: The responsibilities and procedures applicable to the quality control unit are not fully followed. As we shall see this is a vast understatement. Specifically, there is a cascade of failure in your Quality Unit’s lack of oversight on the control and management of GMP documents that are critical in ensuring the drug products manufactured and tested at your site are safe and effective. Summarizing the observation: QC, Production and Engineering employees destroying GMP documents by tearing it into pieces and disposed in scrap areas. Additionally, we found a truck full of transparent plastic bags containing shredded documents and black plastic bags containing document torn randomly into pieces…12 This is industrial-scale destruction and falsification of records for which there are no excuses. The 483 observation identifies several torn records in detail as well as attempts to hide destroyed evidence and lying to the inspectors. My resolution for this violation is very simple. Fire all employees and raze the facilities to the ground. 11 TECHNOLOGYNETWORKS.COM Lab of the Future Summary There is an apocryphal phrase never assume malice when stupidity will suffice. However, the citations in this 483 Form provide ample evidence of both malice and stupidity.12 This conclusion arises from evaluating the systematic problems with the pharmaceutical quality system, lack of record keeping, inconsistent documents, inadequate review in the lab coupled with abysmal quality and management oversight highlighted in the 483 observations. Many of the observations discussed suggest incompetence, and we have highlighted how automation can be used to remediate these processes. However, peak integration manipulation, manual colony counting, hiding destroyed records, lying to inspectors and the evidence of a truck full of shredded and torn GMP records being driven off-site shows that malice in the form of industrial scale falsification is abundantly present. Learn from this company to avoid data integrity citations. References 1. United States Food and Drug Administration. Able Laboratories Form 483 Observations. 2005. 2. Medicines and Healthcare products Regulatory Agency. MHRA GMP Data Integrity Definitions and Guidance for Industry 2nd Edition. 2015. 3. Medicines and Healthcare products Regulatory Agency. MHRA GXP Data Integrity Guidance and Definitions. 2018. 4. World Health Organization. WHO Technical Report Series No.996 Annex 5 Guidance on Good Data and Records Management Practices. 2016. 5. United States Food and Drug Administration. FDA Guidance for Industry Data Integrity and Compliance With Drug CGMP Questions and Answers. 2018. 6. Pharmaceutical Inspection Convention / Pharmaceutical Inspection Cooperation Scheme. PIC/S PI-041 Good Practices for Data Management and Integrity in Regulated GMP / GDP Environments Draft. 2021. 7. International Society for Pharmaceutical Engineering. GAMP Guide Records and Data integrity. 2017. 8. International Society for Pharmaceutical Engineering. GAMP Good Practice Guide: Data Integrity - Key Concepts. 2018. 9. International Society for Pharmaceutical Engineering. GAMP Good Practice Guide: Data Integrity by Design. 2020. 10. Parenteral Drug Association (PDA). Technical Report 80: Data Integrity Management System for Pharmaceutical Laboratories. 2018. 11. Active Pharmaceutical Ingredients Committee. Practical risk-based guide for managing data integrity, version 2 2022. 12. United States Food and Drug Administration. Intas Pharmaceuticals Limited Form 483 Observations. 2022. 13. United States Food and Drug Administration. 21 CFR 211 - Current Good Manufacturing Practice for Finished Pharmaceuticals. 1978. 14. European Commission. EudraLex - Volume 4 Good Manufacturing Practice (GMP) Guidelines, Chapter 4 Documentation. 2011. Did you know you can access details about your instruments on service contracts from your PC or mobile device? It's all about the data! Gain transparency and visibility into all aspects of your instruments enabling you to make faster and more informed business decisions to improve your lab. With PerkinElmer's OneSource Portal web and mobile app you can view all of the data relating to your instruments such as serial numbers, year of manufacture, past and scheduled service events, and contract status all through an easy to use interface. Exclusive access to the mobile app is included with your Portal subscription. Additional features included with the mobile application: • Use voice and text capabilities to describe issues well as upload images to describe service issues Some additional capabilities in the OneSource Portal are: • Quickly generate a service request • Share service details with colleagues • Receive notifications about upcoming instrument events such as PM and service activities Improve your lab operations with the OneSource® Portal and Mobile App STAY CONNECTED TO YOUR LAB Laboratory Services PORTAL Web-based access for all of your OneSource needs MOBILE APP Mobile access to your instrument repair status To learn more please contact your local PerkinElmer representative. For a complete listing of our global offices, visit www.perkinelmer.com Copyright ©2023, PerkinElmer, LLC. All rights reserved. PerkinElmer® is a registered trademark of PerkinElmer, LCC. All other trademarks are the property of their respective owners. 96600 PerkinElmer, LLC. www.perkinelmer.com • Scan instrument service tags to quickly open service requests • View upcoming service events • Create instrument 'favorites' • View service history and and generate reports for individual instruments or your entire fleet 13 TECHNOLOGYNETWORKS.COM Towards the Lab of the Future Joanna Owens, PhD In the lab of the future, researchers will be freed from manual, repetitive experimental tasks, as automated tools and artificial intelligence-powered robots carry out protocols, collect and analyze data and design subsequent experiments, freeing up time for humans to focus on interpreting what the results mean and addressing the bigger scientific questions. The lab of the future will bring together a range of different technologies, all digitally connected and seamlessly integrated. These innovations will be involved in every step of the research cycle, from managing a lab’s supply chain of scientific products and reagents – handling samples, chemicals and equipment – to sharing data within and across organizations. But the timescale for realizing this vision is slower in some research sectors than others. In this article, we look at the barriers preventing more widespread adoption of automation and digitization, and the opportunities they could bring. Automating the academic lab Anyone who has worked in a lab will be familiar with the repetitive, manual nature of many experiments, and there seems ripe opportunity for using automation to free up researchers’ time. But for academic labs, adopting automation can be daunting and cost-prohibitive, and isn’t helped by structures for funding and impact assessment. “I think the vision of the lab of the future differs in academia and industry because we have different outputs,” says Dr. Ian Holland, an engineer who moved from the automation industry to a lab focused on tissue biofabrication at the University of Edinburgh and has written about the “automation gap” in academia. “Academic labs tend to carry out a wider range of work and there’s considerable protocol variability, whereas industry uses standardized protocols for highly focused, repetitive applications, which are more amenable to automation. Academic labs cannot afford to invest in off-the-shelf technologies that aren’t more flexible to suit their needs. So, although there is appetite for the improved efficiency that automation brings, the route to the lab of the future for academic labs is less clear.” There is a shared vision though, which is a world where scientists spend more time doing science and automation carries out the manual tasks. “It’s not good having highly educated people carrying out manual tasks, and I think that happens too much in academia. I’d like to see more manual tests done by machines and let scientists do more science,” says Holland. Barriers to adopting automation The short-term nature of academic research funding does not lend itself to investing in large-scale technologies for modernizing the lab, says Holland, and although investment in major infrastructure such as robotics will improve efficiency, it is difficult to directly relate that to an increase in the output of research papers – the main metric used to measure a lab’s success – making the investment hard to justify. This is a problem also experienced by Prof. Ross King, at Cambridge University, who has been working for several Lab of the Future Credit: iStock 14 TECHNOLOGYNETWORKS.COM Lab of the Future decades on ‘robot scientists’ – semi- or fully autonomous robots that automate simple forms of scientific research, from setting new hypotheses to automatically designing and running efficient experiments to discriminate between them. This futuristic type of research seems to divide funding panels, who have tended to take a conservative view, and existing university structures don’t lend themselves to the collaborative, interdisciplinary nature of the work required. “I think it’s slowly changing, and we’re getting traction in different areas, especially now these ideas are being taken up by the pharmaceutical industry,” says King. Another challenge for academic scientists is a skills gap, because automation and robotics requires an understanding of mathematical models, machine learning and engineering – expertise not every lab has easy access to. And although automation brings efficiency, it also brings with it new challenges, such as how to manage large amounts of data. This is where having the right expertise can help, as Prof. Ola Spjuth, from Uppsala University in Sweden, explains: “We have a big focus on trying to automate our entire cell-based screening and profiling methods in the lab, and this generates a lot of images. This scale of data can scare a lot of researchers, but we have a background in managing big data and using high-performance computing clusters here, so we see large amounts of data as valuable. We’re also not the typical life scientists in that we take an engineering approach and have a multidisciplinary group with experimentalists, data scientists and engineers.” Using automation and AI to improve efficiency and reproducibility Spjuth took what he calls an unconventional strategy to automating the lab in that they did not go out and procure entire robotic installations from vendors, instead choosing to buy individual equipment components and build the system themselves using an open-source approach. “It’s a lot more challenging than buying something off the shelf, but we have full control of all steps in the protocol, and we wanted a research environment that we can grow with and update.” So far, the main efficiency gains have not been the envisioned capacity increase from using robots able to work 24/7. “We are getting there,” says Spjuth, “but our system still needs a lot of human support, and steps such as cell culture are too expensive for an academic lab to fully automate right now.” The major gain, he says, is in reproducibility – every experiment is carried out in exactly the same way. In fact, alongside efficiency, reproducibility appears to be one of the main drivers for automating research processes. One of the goals of King’s work on robot scientists is to improve the scientific method. “Machines in some ways already do better quality science than humans because what they do is recorded, explicit and clear,” says Ross. “Human beings are often unintentionally sloppy about what they do in experiments, and there’s a huge problem with scientific reproducibility because experiments are so susceptible to human error. Just like games on computers have improved over the years, we think that in science, the machines will keep progressing. Ultimately, they’ll be as good as humans at science, and maybe even better.” King has already developed two prototype robot scientists, Adam and Eve. Adam was designed to carry out functional genomics in yeast, assigning functions to the genome that was sequenced back in 1996. Eve specializes in early-stage drug design, using artificial intelligence to find compounds to treat specific diseases. “The way compound screening used to be done in industry was you would make an automated assay to tell you if a compound was likely to be good or not, and then you’d screen a large compound library – maybe one million compounds – and find a small number of hits to take forwards. Then you’d start again with another assay and library,” explains King. “But actually, that’s a missed opportunity, because you’ve learned something during the screen and you could use that insight to decide what to do next.” By using quantitative structure-activity relationship (QSAR) models and accumulating biological knowledge, Eve was trained to find hits using only a small fraction of the compounds in a library – speeding up the process and make it more cost effective. Now, King is working on the next iteration of the robot scientist – called Genesis – as part of the Nobel–Turing AI scientist grand challenge. The challenge is to develop AI systems capable of making Nobel-quality scientific discoveries autonomously at a level comparable, and possibly superior, to the best human scientists by 2050. Genesis is a scaled-up robot scientist with thousands of micro-chemostats – tiny bioreactors where nutrients are continually added to cells and metabolic end products are continually removed. These will enable Genesis to run more sophisticated experiments in parallel. “We need an AI system to plan so many experiments and especially hypothesis-led experiments, rather than just altering a component and seeing what happens,” says King. “Here, the robot is saying ‘I think change Y will do X to this model, and then it conducts the experiment to see if the hypothesis is true.” Moving towards a digitized laboratory In addition to adopting robotic solutions to improve efficiency and reproducibility in the lab of the future, many researchers are moving towards digitizing their 15 TECHNOLOGYNETWORKS.COM Lab of the Future labs, switching from paper-based systems to informatics solutions such as laboratory information management systems (LIMS) and electronic notebooks (ELNs). LIMS enable researchers to keep track of data associated with samples, experiments and instruments efficiently, as well as actively manage lab processes, while ELNs digitize note taking and can automate the data review process. Guidance on good records and data management practices from the World Health Organization (WHO) recommends that hybrid systems – a combination of manual and electronic systems – should be replaced by fully digitized systems at the earliest opportunity. Adopting informatics solutions such as LIMS can offer laboratories several benefits, including helping to improve performance, maximizing quality and ensuring compliance requirements and regulations are met. They can also remove repetitive, laborious steps in workflows and reduce human error. The time savings can empower scientists, allowing them to focus on more complex and meaningful work. Despite the benefits offered, barriers to adopting these solutions and digitizing a laboratory remain. The cost of subscriptions, new equipment and software, as well as time to implement the solutions, can be prohibitive for many laboratories, particularly in academia. “Accessibility is also a huge barrier. Many academic laboratories aren’t set up for the digital capture of laboratory information, both from a hardware and software perspective,” Dr. Samantha Pearman-Kanza, senior enterprise fellow at the University of Southampton, told Technology Networks previously. Problems with outdated equipment and software compatibility can further limit the adoption of digital technologies. In addition, “The lab can often be a hostile place for technology,” said Pearman-Kanza. Space for using laptops or tablets may be limited, and researchers may be concerned about spills and accidents occurring. Even things such as removing gloves to type notes rather than jotting them in a notebook can be seen as prohibitive. However, the continuing advancement of technologies is likely to reduce these barriers and encourage greater adoption of digital solutions in the lab of the future. “Much like smart homes have become commonplace in today’s society, so will smart labs. Users will be able to control their laboratories by voice using smart lab assistants, all of the laboratory systems will be seamlessly linked together and users will have multiple options to record their data via voice, tablets, phones or computers if they wish,” envisaged Pearman-Kanza. Taking small steps towards the lab of the future It might be another two decades before fully autonomous robots are designing and conducting experiments in the lab, but it’s never too early for academic labs to start their journey towards automation, says Holland. “As an engineer in a biology lab you can see the potential opportunities to use technology to improve processes. However, I think too often in academia, researchers strive for a magic machine that does everything. But that is never how you develop automation as an engineer, you build prototypes that carry out each part of the process.” Holland advocates starting small, by automating something simple such as fluid dispensing that can bring substantial gains in efficiency and reproducibility. In the tissue biofabrication lab, just making this change has reduced a protocol from 25 to 5 minutes, freeing up time for other tasks. Another advantage of adopting automation early is it can help researchers looking to translate discoveries from bench to bedside. “The earlier you can include automation in your process and start thinking about that, the better chance you have of convincing people to invest in your product, because they can see it will be easy to scale up quickly.” In Spjuth’s lab they are hoping for additional collaboration with other researchers working on their own robotic and automated solutions for the lab, sharing protocols and code. “With major advances in technology such as 3D printing and people now sharing code for these and other applications, it is becoming possible for researchers to do much more independently. The do-it-yourself movement is advancing and that means you can build your own microfluidic chips and microscopes, and as prices for robots come down there is an opportunity for many biological labs to adopt some sort of lab automation.” However, an important consideration as this movement advances, notes Holland, is sustainability. “There is already a real problem with automated processes generating high amounts of waste – a machine generating millions of waste pipette tips, for example. I think this needs to be considered more and certainly in the design stage, both from an environmental perspective and to ensure supply chains can meet demand.” 16 TECHNOLOGYNETWORKS.COM New Advances in the Bioprocess Pipeline Aron Gyorgypal, PhD The bioprocess landscape is an ever-advancing industry looking to optimize the manufacturing of biologics. Of key interest is addressing the challenges of high costs, time-consuming assays and processing complexities. These challenges are tackled with technological innovation meant to streamline and optimize the production pipeline. Here we delve into groundbreaking advancements throughout the bioprocess pipeline that are poised to advance bioprocessing. High-throughput upstream process screening technology Early biologic production is known to come with high costs as it takes time to evaluate an optimized process, including the medium, feed, reactor conditions and purification methodologies. A critical need is to speed up analytical times while scaling down cell culture systems, to help better assess cell lines for key attributes, while decreasing the costs of goods manufactured. This will ultimately lead Lab of the Future Credit: iStock, Gyorgypal, A. doi: 10.7282/t3-1c02-k359. Bioprocess scheme showing the overall bioprocess scheme of the upstream and downstream unit operations, including early clone clean selection and the use of model and control along the product pipeline. 17 TECHNOLOGYNETWORKS.COM Lab of the Future Credit: iStock to accelerated production of innovators and biosimilars. Both liquid handlers and microfluidics coupled with analytical equipment such as ultra-high-performance liquid chromatography (HPLC) and mass spectrometry (MS) have been showcased as effective for achieving this goal. A successful cell line has high productivity. Cell clones that produce high titers allow for the production of drugs in a higher quantity but in less time. While new technologies enable the engineering of high titer cell clones, there is still a need to screen these clones quickly.1 The current paradigm for titer analysis depends on chromatographic approaches which, even with the newest ultra-high-performance liquid chromatography (UHPLC) models, will take minutes to analyze one cell line. However, a recent innovation by D’Amico et al. used microfluidics coupled with electrospray ionization MS to improve analysis time tremendously from 3 minutes to 25 seconds per sample.2 This is achieved using a multiplate Protein A-based cleaning coupled with microdroplet entrainment and subsequent analysis by native MS. The authors also imply that this technology can measure key product quality attributes. Unsurprisingly, microfluidics has recently gained prevalence as a technology to aid in high throughput automation with both MS and HPLC techniques, as an avenue towards process automation and away from the need of manual in-process testing.3 A common product attribute that needs to be monitored in biopharmaceutical manufacturing is protein aggregation, which is known to influence a product’s safety and efficacy. During early-stage development, identifying aggregation in scale-down cell culture systems can be tedious, as the small sample volumes make aggregate detection challenging. Scientists at the University of Sheffield looked to overcome this hurdle by developing an automated small-scale analytical platform workflow, combining purification and aggregation analysis of protein biopharmaceuticals expressed in 96 deep well plate cell cultures.4Their workflow utilized a liquid handler equipped with a protein-A PhyTip column for purification and a size-exclusion chromatography (SEC) column for aggregate monitoring. The data was then analyzed, and cultures were ranked by their aggregation levels. This high-throughput method, allowing up to 384 clones to be screened in 32 hours, enables high aggregation phenotypes to be eliminated from clone selection early on in cell line screening and can be used for both monoclonal as well as bispecific antibodies alike. Technology such as this one showcases the utility of liquid handlers for highthroughput automated screening. Downstream purification – New ligands for multimodal chromatography Downstream product processing accounts for over half of the production cost for monoclonal antibodies, and significant research is devoted to developing compact and affordable purification processes. Scientists are looking to innovate new chromatographic technologies using novel ligands capable of improved separation for higherthroughput operation. Within separation science, attention is put on multimodal ligands that combine multiple techniques, such as hydrophobic, hydrogen binding and electrostatic integrations, that work orthogonally or complementary to affinity or conventional hydrophobic and ion exchange chromatography.5 The use of multimodal chromatography effectively removes process-relative impurities, such as host cell protein (HCP), DNA, aggregates and protein fragments. Scientists at the North Carolina State University developed a multimodal salt tolerant cation exchange (CEX) membrane, which is the first of its kind based on nonwoven fabrics to be published and used for product capture and polishing of biologics.6 This membrane was produced by coupling 2-mercaptopyridine-3-carboxylic acid (MPCA) to a polybutylene terephthalate (PBT) nonwoven fabric, modified by UV grafting of glycidyl methacrylate (polyGMA). The membrane achieved higher biologic binding and capacity versus commercial CEX multimodal resins and may be an effective alternative to current commercial resin columns. Future efforts by the team look to expand on the success of this technology to bring more purification applications, such as a low-cost Protein-A alternative and peptide-based resins for HCP depletion. Ultimately, technology such as this looks to disrupt the current downstream process purification paradigm to allow for low-cost, high-throughput purification of biologics. Establishing Multi-Attribute Monitoring The complexity of biologics processing comes from the multitude of heterogenous post-translational modifications that can alter protein-based therapeutic efficacy and safety. Using multi-attribute monitoring (MAM), a liquid chromatography-mass spectrometry (LC-MS) based method, allows for direct characterization and monitoring of numerous product quality attributes at the amino acid level. Simultaneously monitoring multiple attributes on the peptide level makes MAM a more informative and streamlined approach than typical LC and MS methodologies. A recent publication by Millán-Martín et al. described a comprehensive MAM workflow for biologic characterization and current good manufacturing (cGMP) practices.7 This work aims to standardize the protocols used for MAM to help with implementation and is the first protocol to establish the MAM methodology, which has not been available in the literature prior. The overview of the workflow explains the three major parts of setting up a MAM protocol: 1. Sample digest: A proteolytic digest of the protein to 18 TECHNOLOGYNETWORKS.COM Lab of the Future Credit: iStock produce peptides. 2. High-resolution accurate MS (HRAM-MS) of the peptide: Discovery phase (phase 1) to enable confident identification of peaks to produce a peptide library for the biologic. Monitoring phase (phase 2), where full MS acquisition is carried out for the predefined product quality attributes (PQAs) and critical quality attributes (CQAs) based on the phase 1 peptide library. 3. Data processing and analysis: Data processing for phase 1 looks to evaluate the sequence coverage and assess the PQAs/CQAs present for the biologic. In phase 2, during targeted monitoring, the predicted PQAs and CQAs are monitored, and detection of new peaks are screened/investigated for the presence of either product or process-related impurities that may be present during bioprocessing. This MAM method relies on a bottom-up approach. Other technologies using MAM seek to do native MS to allow for PQA/CQA analysis on the intact protein level, which can give a different level of information. Work by Bhattacharya et al. explains the use of native-MS MAM to characterize antibody titer, size, charge and glycoform heterogeneity in cell culture supernatant.8 The workflow was achieved by 2D-LC-MS in which the first dimension coupled Protein A in series with size exclusion chromatography, and the second dimension was cation exchange chromatography. The glycosylation was then characterized with Quadrupole Time of Flight Mass Spectrometry (qTOF-MS). Academic research has seen the quick adoption of the use of MAM workflows, which will help speed up workflows in industry applications once validated for use. Forecasting multistep-ahead-profiles using data-driven models Process models and simulations are increasingly used to better understand, optimize, and predict different aspects of a bioprocess. More dense datasets are now available for bioprocesses with new advancements in process analytical technology (PAT) and artificial intelligence (AI). By integrating AI-based data-driven models with an upstream bioprocess, the outcome can be forecasted with better prediction. Outputs, such as the cell viability, product titer and metabolite concentrations, such as glucose, can be predicted and better optimized for the process. These optimized models can also be implemented with a realtime digital twin. Recently, researchers at Sungkyunkwan University published a practical guideline for selecting optimal combinations of data-driven elements to predict Chinese hamster ovary cell culture’s process profile.9 This was achieved through a systematic framework for collectively evaluating various AI algorithms, forecasting strategies and model inputs. The utility of this framework was then demonstrated on an array of cell culture datasets under various conditions, selecting the most predictive algorithm for said condition. The choice of model elements that are considered will impact both computation load and accuracy of the model, the researchers found. The suitability of the model will depend on data availability, the complexity of the modeled culture conditions and model input variables. This framework should be used as a rational and practical guide on what should be considered to build the best forecasting data-driven model. References: 1. Tihanyi B, Nyitray L. Drug Discov Today Technol. 2020;38:25- 34. 2. D’Amico CI, et al. J Am Soc Mass Spectrom. 2023;34(6):1117- 1124. 3. Tiwold EK, et al. J Pharm Sci. 2023;112(6):1485-1491. 4. Lambiase G, et al. J Chromatogr A. 2023;1691:463809. 5. Zhang L, et al. J Chromatogr A. 2019;1602:317-325. 6. Fan J, et al. Sep Purif Technol. 2023;317:123920. 7. Millán-Martín S, et al. Nat Protoc. 2023;18(4):1056-1089. 8. Bhattacharya S, et al. J Chromatogr A. 2023;1696:463983. 9. Park S, et al. Biotech Bioeng. 2023;120(9):2494-2508. Considerations for HTS Artificial Intelligence and High-Throughput Technology AUTOMATION MINIATURIZATION DATA ANALYSIS The growing influence of AI across manufacturing can also be seen in biomedical settings. Let’s take the example of drug discovery. There are several points during a pharmaceutical’s life cycle where AI can have a say. Target identification The majority of targets for small-molecule drugs are proteins. Open-source software AlphaFold2 can predict the structure with incredible accuracy. AlphaFold uses an input amino acid template sequence to build up a multiple sequence alignment model that predicts molecular arrangement. This initial step is combined with neural networks – data structures that mimic the connections between cells in the human brain. Click here to download the full infographic Cubis® II Lab Balance – The Future of Lab Weighing Experience unparalleled connectivity and efficiency in your lab with the Cubis® II balance. Seamlessly connect to LIMS/ELN systems or Ingenix fleet management, ensuring the highest level of accuracy and data integrity for your weighing data. Fully compliant with Ph.Eur. Chapter 2.1.7, USP Chapter 41, and 21 CFR Part 11, the Cubis® II ensures all standards are met, no matter the industry. Connect your lab to the future with the Cubis® II. Click to Learn More 21 TECHNOLOGYNETWORKS.COM Data Reporting: Connecting the Islands of Automation Bob McDowall, PhD In this article, we explore ways analytical scientists can improve their data reporting and sharing practices, leading to improvements in data integrity and regulatory compliance. Laboratory process automation in the pharmaceutical and related industries has been a topic of discussion since the 1980s. To help achieve this, there have been many technical advances in applications such as instrument data systems (e.g., chromatography data systems (CDS)), electronic laboratory notebooks (ELN), laboratory execution systems (LES) and laboratory information management systems (LIMS). In addition, there has been convergence between applications; LIMS applications integrated with LES functionality mean only one application must be implemented. However, often the problem is not the functionality offered by an informatics application but the way a project is implemented in laboratories. In many cases, informatics application problems include: • Failure to design a fully electronic process for automation. • Not eliminating spreadsheets from a process. • Incorrect application configuration by overinterpreting regulations. • Not interfacing instruments requiring manual data entry into the application. • Partial instead of end-to-end process automation. We will discuss some ways of implementing these informatics applications to avoid running into these problems. Although the focus in this article will be on good practice (GXP)-regulated laboratories, the principles outlined here are applicable for all. Even if GXP regulations do not apply, the business benefits of effective laboratory process digitalization will enhance business effectiveness, speed of decision making and collaboration inside and outside of an automated laboratory. The aim of this article is to connect the islands of automation that exist in an ocean of paper in most laboratories to ensure an electronic environment. Process, process, process The starting point for an effective implementation of any informatics application is to understand the workflow that you are automating. Typically, this is a two-stage process: 1. Map and understand the current (as-is) process This needs to be a logical mapping of the process with a team of subject matter experts (SMEs) from the laboratory. They need to document the activities in the process, including those not in current SOPs, and understand the data inputs and outputs and the records created. Any bottlenecks in the process are identified and the causes listed. Paper printouts and data vulnerabilities that need to be resolved in the second part of the mapping process can also be documented. Lastly, improvement ideas for the process should be listed. Process maps can be drawn after the workshop for review later. Lab of the Future Credit: iStock 22 TECHNOLOGYNETWORKS.COM Lab of the Future 2. Redesign the process to work electronically (to-be process) The first part of the second phase is to review the “as-is” process maps from the first phase. Inevitably there will be changes and this is normal. Once the current process maps are agreed, the improvement ideas can be used to challenge the as-is process. This includes removing bottlenecks, eliminating paper and interfacing analytical instruments to avoid manual data transfer and the associated transcription error checking. The “to-be” process should be used as a blueprint for automating the process with the selected informatics application. Although redesigning a process takes time, the payoff is immense as the process is simplified and operates faster than currently performed. If an application automates an as-is process, there is a high likelihood that it will require more effort as the process will still be hybrid and less efficient with zero return on investment for the parent organization. Therefore, management must resource and support the process redesign effort. Process mistake one: Partial automation During training courses that I have conducted, several attendees have stated that the first phase of a LIMS project is to automate sample management alone. This is a mistake, in my view, as a complete process is not being automated and there will be no business benefit generated. There will be a lot of effort being put into sample management with pretty barcode labels but nowhere to use them during analysis. Instead, ensure sample management is integrated into an end-to-end process that generates reportable results. This is not to say “automate the whole laboratory” but rather to ensure that say, raw materials testing can be automated in its entirety from receipt to release. Then additional analytical processes such as in-process, finished product and stability testing can be added incrementally. This approach is vital for demonstrating to both users and management that the system works and delivers benefits. Process mistake two: Failure to interface analytical instruments If no analytical instruments are interfaced to an informatics application, how will analytical data be input to the software? Manually, obviously! However, it is better to now consider the impact of the updated EU GMP Annex 11, where proposals for validated software technical controls are preferred to procedural controls operated by users: Clause 2 ... Configuration hardening and integrated controls are expected to support and safeguard data integrity; technical solutions and automation are preferable instead of manual controls.1 The regulation also states that digitalization should be considered: Clause 3. An update of the document with regulatory expectations to ‘digital transformation’ and similar newer concepts will be considered.1 In addition to the business benefits of interfacing now, there is the probability of regulatory pressure to automate in the future. Therefore, when planning laboratory automation, it is essential, from both business and regulatory perspectives, that key instruments are interfaced in the first phase of any project. Process mistake three: Failure to eliminate spreadsheets Spreadsheets are ubiquitous in laboratories, easily available, easy to use and a compliance nightmare.2,3,4 They are also a hybrid system with signed paper printouts and an electronic file. The biggest problem is that as a hybrid system you can’t have an electronic process. The calculations performed should be incorporated into the informatics application to prevent printouts and manual input of the data. The exception to this rule is where a spreadsheet can be used within an application such as an ELN where the application’s audit trail can monitor changes, but the data used in the spreadsheet calculation must be loaded automatically into the spreadsheet to avoid printing and transcription error checking. However, it is critical to check if the application audit trail can track individual actions within the spreadsheet because if interim changes between an open and save event cannot be tracked this is a data integrity failure. You are just changing a hybrid problem into an electronic problem. Process mistake four: Failure to eliminate blank forms The FDA has required quality control (QC) laboratories to control blank forms since 19935 and this has been iterated in regulatory guidance on data integrity from MHRA, FDA and PIC/S.6,7,8 If uncontrolled blank forms are used it is impossible to determine how many times the work has been performed before an “acceptable” result has been obtained. If blank forms are used the administrative controls required to ensure that data cannot be falsified result in a high overhead for the QA department to issue and reconcile these forms. 23 TECHNOLOGYNETWORKS.COM Lab of the Future Credit: Bob McDowall Process mistake five: Failure to use electronic signatures Although a process may be automated, some laboratories don’t take the final step of using electronic signatures by the performer and reviewer of the analysis. This creates a hybrid system but also leaves the electronic records unlocked, at which point post-reporting changes could be made. This is unacceptable and therefore the use of electronic signatures is a natural outcome of an electronic or digitalised process. Turning principles into practice That’s the theory for laboratory automation. To see how this is put into practice, we interviewed Roger Shaw of Digital Lab Consulting (DLC). His principles for enabling laboratory connectivity, data sharing and regulatory compliance are to: • Ensure analytical instruments are connected to instrument data systems or informatics applications. • Connect informatics systems together and communicate electronically between them. • Ensure each informatics system supports applicable regulatory requirements including data integrity. From these principles, there are further requirements. Instrument standardization is important as it simplifies interfacing as well as a reduced validation approach that can be taken after the first connection has been verified. Unlike pharmaceutical manufacturing, where there are well-established data standards and protocols, there is a lack of data standardization and interface standards for laboratory instruments. For some projects Roger’s team have used instrument interfacing standards developed by the SILA Consortium and data standards such as ISA-88 when combining process development and manufacturing data. For other projects, the data system from the instrument supplier provides a simple interface for connection. Roger described a case study for a QC department in one site of a multinational pharmaceutical company, shown in Figure 1. The components of the project consisted of the following: • Gas and liquid chromatographs connected to a networked CDS. • pH meters and analytical balances connected to the instrument supplier’s data system which is equivalent to an LES in that workflows can be defined to link instruments together for assays. • LIMS connected to the CDS and instrument data system. Chromatography Data System HPLC Analytical Balance pH Meter Laboratory Information Management System Instrument Data System Figure 1: Outline of a laboratory automation project. 24 TECHNOLOGYNETWORKS.COM Lab of the Future The workflows between the three informatics applications are: • The LIMS is responsible for sample management and reporting of results to production. • The pH meter and balances are interfaced to the instrument data system which acquires data from standards and samples and transfers the results of this work to the LIMS. • The sample identities and weights, and the reference standard purities are imported from the LIMS into the CDS for each batch of analysis. • Following chromatographic analysis and interpretation, all post-run calculations are incorporated into the CDS workflows and electronic signatures are used by the performer and reviewer of the batch. • When a reviewer signs the CDS report, all records are locked and can be viewed by authorized users but not changed unless unlocked. • The reportable result is transferred to the LIMS. • All sample test results are collated in the LIMS and a certificate of analysis (CoA) is generated electronically and electronically signed. Business benefits as well as regulatory compliance are obtained with this approach. Summary This article explores ways to improve data reporting and sharing practices coupled with improving data integrity and ensuring regulatory compliance. To achieve these goals, it is imperative that the current process is analyzed to identify bottlenecks, use of spreadsheets and blank forms and identify data vulnerabilities. The redesigned process should eliminate as many these problems as much as possible by working electronically with electronic signatures. To ensure success the applications selected to automate the new process must have adequate technical controls to ensure data integrity and regulatory compliance. Rather than be standalone, each application should be interfaced to transfer data and information electronically to enable effective scientific collaboration. References 1. European Medicines Agency. Concept Paper on the Revision of Annex 11 of the Guidelines on Good Manufacturing Practice for Medicinal Products – Computerised Systems. 2022. 2. US Food and Drug Administration. FDA Warning Letter Tismore Health and Wellness Pty Limited. 2019. 3. McDowall RD. LCGC Europe. 2020;33(9):468–476. 4. McDowall RD. Spectroscopy. 2020;35(9):27-31. 5. US Food and Drug Administration. Inspection of Pharmaceutical Quality Control Laboratories. 1993. 6. Medicines and Healthcare products Regulatory Agency. MHRA GXP Data Integrity Guidance and Definitions. 2018. 7. US Food and Drug Administration. FDA guidance for industry data integrity and compliance with drug CGMP questions and answers. 2018. 8. Pharmaceutical Inspection Convention/Pharmaceutical Inspection Cooperation Scheme. PIC/S PI-041 Good Practices for Data Management and Integrity in Regulated GMP/GDP Environments Draft. 2021. 25 TECHNOLOGYNETWORKS.COM The Future of Digital Health With Professor Michael Snyder Kate Robinson & Lucy Lawrence Michael Snyder, Stanford W. Ascherman professor of genetics, is a leader in the field of functional genomics and proteomics. Snyder has combined different state-of-the-art “omics” technologies to perform the first longitudinal detailed integrative personal omics profile (iPOP) and has used this to assess disease risk and monitor disease states for personalized medicine. Technology Networks invited Snyder to an Ask Me Anything session to answer your questions about the latest technologies and innovations that are shaping the future of digital health. These are just some of the questions that we asked Snyder, click here to watch the full AMA. Q: What sparked your interest in lab digitalization? A: Well, I think the healthcare system is broken. We tend to go to a physician only when we’re ill rather than try and keep ourselves healthy. There are many steps involved in healthcare. You have to travel, usually at a very inconvenient time, to a doctor’s office, which pretty much looks the same as it did 40 years ago. When you’re there, they stick a needle in you and draw blood, and from that blood, they’ll make very few measurements. Then they’ll make decisions about your health based on population averages, calculating a mean of your measurements, and compare it with everyone else’s. So, we think every one of those steps can be changed. And I think one big part of it is pulling in big data. Q: How can wearable technology bridge the gap between lab research and clinical applications? A:Right now, this is all research, but a small number of wearable devices are approved by the US Food and Drug Administration (FDA). The reason we’re very keen on wearables is because they’re measuring you 24/7. In a physician’s office, you get 15 minutes, then they make a measurement, and most people are anxious, so the measurements aren’t always accurate. Wearables run in the background and measure heart rate and heart rate variability, which are important health monitors, and because they can take your resting heart rate first thing in the morning, they’re quite accurate. So, if something is off, it’s likely that you’re either ill mentally or physically. When I had presymptomatic Lyme disease, my smartwatch showed that my heart rate jumped up and my blood oxygen dropped, and after blood tests I was given a diagnosis. In a real-time detection study, we have also shown that you can tell someone with COVID-19 is unwell. Our realtime alerting system can detect illness three days prior to symptom onset, and it’s more sensitive than an antigen test in some cases. Lab of the Future Credit: iStock 26 TECHNOLOGYNETWORKS.COM Lab of the Future Credit: iStock If you’d like to sign up to a study, follow this link: https:// innovations.stanford.edu/wearables Q: What do you see being the future of digital health and lab tech, relating to diabetes? A: Historically, glucose monitors have mainly been useful for type one diabetics and insulin dependent type two diabetics, but we started putting these monitors on people without diabetes and those with prediabetes. In doing this, we discovered that a lot of “normal” people’s glucose spiked just as high as those with diabetes, which could be an indication of those on their way to becoming diabetic. Glucose monitors are powerful, because what spikes an individual’s glucose is very personalized, meaning some people spike in response to bananas, others to potatoes, etc. Nearly everybody spikes in response to rice and cornflakes. These differences are partially due to the microbiome, along with genetic and epigenetic factors. Apps have been developed that utilize continuous glucose monitors to make personalized recommendations on what foods to eat, what to avoid, what times to eat and how to exercise. In Europe, glucose monitors can be bought over the counter, but in the United States, you still need a physician to order them. I think monitors will soon be accessed more easily for everyone. The ultimate goal is to get your glucose levels under control early. While this may not prevent you from becoming diabetic, it could at least delay things for a few years. That’s what health monitoring is all about, trying to keep yourself healthy as opposed to trying to fix yourself when you’re ill. Q: Do you see virtual reality playing a role in digital health? A: We’ll certainly see a role in health in general. I think that people would prefer not to go to a physician for most things, but rather, want to know what’s going on right there and then. Virtual reality could allow people to choose a time convenient for them to have a checkup, and they could go to a physician in person for more serious things. Using virtual health doctors mediated through virtual reality could help people relate to things better. Right now, we’re used to thinking in two dimensions and that works fine, but as these new platforms come out, we’ll likely get better about thinking of them in three dimensions. From a research lab standpoint, virtual reality would be highly useful and more accurate for training than watching a video. Q: Will data commons, with consent of the individual, be a good direction to go in? A: Sharing data and aggregating data – with consent of the individual – is for the good of all. There are millions of people with health records who have used medications. If you have a health condition, it would be good to know about the other people with the same condition, what drugs they have taken and the outcomes of treatment. There’s a lot of data out there on that, but it’s not shared. Companies that pull in these kinds of data do exist. They try and use the data for health recommendations, something that’s been done a lot in an academic environment, as well. These days, most things are being done in a federated fashion, meaning that each hospital doesn’t share their data, but they give access to it. So, you can run a search on the data, and they can tell you what their outcomes look like, which is better than nothing. It’s not as good as if everyone shared their data, but we’re still working our way through this. Professor Michael Snyder was speaking to Lucy Lawrence, Senior Digital Content Producer for Technology Networks. SOLUTIONS DETECTION FOR PFAS PerkinElmer’s portfolio supports your laboratory identifying and quantitating PFAS in complex matrices, meeting stringent regulations and reaching the lowest detection limits. The QSight Triple Quad LC/MS/MS system consistently delivers the throughput and productivity you need in your analytical testing laboratory. LEARN MORE For more information visit www.perkinelmer.com/category/pfas 28 TECHNOLOGYNETWORKS.COM Lab Sustainability: A Holistic Approach Srividya Kailasam, PhD Laboratories are complex spaces that cater to different disciplines and are designed to carry out activities ranging from wet chemical reactions and microbiological studies to analysis using sophisticated instruments. They may be operated as independent commercial establishments (e.g., testing labs), or they could be part of educational or research institutions, companies and hospitals. Given this diversity, the need for heating and cooling, and the presence of energy-intensive instruments, it is hardly surprising that labs consume inordinate amounts of resources and generate large volumes of waste. These could be both solid and liquid and possibly infectious. In the past few years, there has been an increasing awareness of the environmental footprint of labs and a growing trend toward “sustainable laboratories”.1 The scientific community has started to focus on areas that make labs sustainable. In a study published in 2022, My Green Lab and Intercontinental Exchange Inc. (ICE), identified that many pharma and biotech companies have adopted zero carbon goals. In this listicle, we explore how a holistic approach that includes optimal lab design, environmentally friendly equipment and instrumentation, and awareness about the impact of labs, coupled with training and adoption of sustainable practices, is required to make labs “green”. Lab design A laboratory’s design can have a vast impact on its sustainability. Careful consideration when building, setting up or renovating a laboratory can help to reduce wastage and lead to improved energy efficiency. A lab should be designed to increase the amount of natural lighting and energy-efficient artificial lighting should be used as a supplement to the daylight. Increasing natural lighting is known to not only improve productivity, but also save costs. A white or nearly white ceiling is recommended for the proper distribution of natural lighting. Dark bench tops and reagent shelves that make the room seem darker must be avoided. While a well-lit, air-conditioned lab is essential for the proper running of sophisticated instruments, it is important to turn off the lights and air conditioning when the instruments are not in use. To ensure compliance, team members can take turns switching off the lights and equipment not in use. Regular light bulbs can be replaced with LED bulbs. Temperature control can also be achieved using chilled beams. When designing a sustainable lab, factors such as its size, layout and infrastructure flexibility must be borne in mind to minimize energy consumption and adapt to future technological changes. The use of modular and reconfigurable furniture also helps in creating a sustainable lab. Biosafety labs must ensure unidirectional flow of materials and personnel and incorporate controls such as Lab of the Future Credit: iStock 29 TECHNOLOGYNETWORKS.COM Lab of the Future negative pressure zones and airtight doors and windows to prevent highly infectious pathogens from escaping into surrounding areas. Since all materials entering a biosafety level 3 lab (BSL3) are considered biohazardous, minimizing the materials (paper, media bottles, plastics, etc.) brought into biosafety labs will reduce the necessity to autoclave and/or dispose of these materials that could otherwise be reused. Several resources are available to help with planning the creation of a sustainable lab. For example, a strategic design framework, Green Lab, has been developed to help create labs with minimal environmental impact using recycled materials. These labs should be capable of generating energy from renewable sources.2 Involving a wide range of stakeholders, applying the highest possible sustainability standards and balancing lab sustainability goals with the health and safety of personnel are suggested as the three keys to designing a sustainable lab.3 Equipment Labs depend on a huge variety of equipment, and careful consideration during its purchase, use and disposal is crucial in ensuring sustainability goals are met. When upgrading or replacing existing instruments, certified environmentally friendly options, such as those with ENERGY STAR® and ACT (Accountability, Consistency and Transparency) labels, should be purchased. The Environmental Impact Factor (EIF) criteria form the basis of creating the labels for life science products. These labels provide scores for parameters such as renewable energy used during manufacture, water consumed during use and sustainability impact at the end-of-life stage. Lower overall score indicates lesser overall environmental impact. In addition, My Green Lab certification helps the labs get their sustainability practices audited. Choosing models that consume fewer resources, including energy and solvents, helps to reduce the environmental impact. For instance, smaller autoclaves with more efficient vacuum generation systems consume less water and can be used instead of a larger model. Periodic maintenance of all lab equipment and instruments must be carried out to ensure optimal performance and energy efficiency. Keeping incubators, freezers, refrigerators and cold rooms clean and frost-free will make them more energy efficient and will increase the lifespan of these products. When possible, refurbished instruments can be used to minimize the build-up of hazardous substances and other products that cannot be recycled in the environment. Manufacturers are increasingly investing in building a circular economy for their instruments, offering incentives for labs to return their old equipment, which can be refurbished and sold on. Operation Laboratories can contain a wide range of energy-intensive equipment that often requires continuous operation. Subsequently, compared to office buildings, they typically consume 5 to 10 times more energy per square foot. Coupled with this, many processes depend on singleuse plastics and use vast amounts of water. Although it is challenging for laboratories to try to match the resource consumption and waste output of a space with such different needs, several steps can be taken to improve their operational efficiency. Instruments such as chromatographs, mass spectrometers and spectroscopes can be shut down when not in use. Similarly, other equipment such as pH meters, balances, centrifuges and water baths can also be turned off when not in use. Equipment that doesn’t require continuous use can be fitted with programmable outlet timers, which could help to reduce equipment energy consumption by up to 50%. When in use, the sash of fume hoods should be kept in the lowest possible position to improve exhaust efficiency and thus save energy. Energy can also be saved by lowering the airflow volume when the fume hoods are not in use. When possible, ductless hoods that filter contaminated air before recirculating it can be used. Operating ultralow temperature freezers at -70 °C instead of -80 °C saves energy without being detrimental to the samples. Labs should reduce, recycle and reuse plastic products whenever possible. Consolidating orders for plastic ware, glassware and chemicals for the different projects/ experiments running in the lab or for multiple labs, buying smaller pack sizes or buying just pipette tips instead of boxed tips can reduce the plastic waste in the lab. Routine lab activities such as filtration, which is an integral part of sample preparation, generates plastic waste such as single-use syringes and filter cartridges. These and other plastic wastes, such as pipette tips, tip boxes, syringes, tubes and nitrile gloves, can be recycled instead of sending them to landfills or incineration. When recycling, care should be taken to disinfect biowastes and segregate these along with other hazardous wastes from recyclable, nonhazardous plastic wastes. A lab case report documented various approaches, such as reusing decontaminated plastic tubes and using sustainable materials like reusable wooden sticks for patch plating and metal loops for inoculation that helped reduce and reuse single-use plastics in a microbiology lab. After implementing these strategies, the lab demonstrated a 43 kg reduction in plastic waste in 4 30 TECHNOLOGYNETWORKS.COM Lab of the Future weeks, in addition to reducing costs.4A report from MIT demonstrates the feasibility of recycling clean lab plastics, stating that the Environmental Health and Safety (EHS) office collected nearly 280 pounds of plastic waste from participating labs every week in 2022 for recycling. To minimize water consumption, taps and other water outlets can be fitted with spray nozzles or low-flow aerators, and recirculating water baths used for cooling reactions that can be carried out on a smaller scale. To reduce water wastage, leakage from faucets, autoclaves, water baths and any seepage from water pipes should be attended to as soon as possible. Washing and autoclaving should be done with minimal wastage of water. Scientists, students and lab personnel should cooperate to ensure that autoclaves are run at full capacity to reduce water as well as energy consumption. Less hazardous substitutes for chemicals and cleaning supplies should be used whenever possible. Labs and workspaces must be kept clean to prevent crosscontamination that leads to wastage. Simple actions like proper labeling and storage of chemicals will enhance their shelf-life and proper usage. Experiments should be designed to derive maximum information with minimum consumption of resources as well as minimum waste generation. This will also reduce the number of experiments and eliminate erroneous experiments. Replacing paper-based lab notebooks, operating procedures and test protocols with electronic records will help minimize paper consumption. Computers and other office equipment can be turned off when not in use. Training Educating stakeholders on the importance of sustainability is a crucial requirement for achieving a green laboratory. In addition to training the scientists on guidelines and regulations, they must be sensitized towards the ecological impact of indiscriminate usage of chemicals, glassware, plastic ware and consumption of resources, especially electricity and consequent generation of waste in the laboratories. The importance of education and capacity building in “green chemistry” and “sustainable chemistry” for creating greener and more sustainable processes has been described by Zuin et al.5 Sustainability in Quality Improvement (SusQI), developed by The Centre for Sustainable Healthcare (CSH) adopts principles of Education for Sustainable Development (ESD) such as “future thinking” and “systems thinking” and “thinking creatively” in the training programs for clinical laboratory professionals.6 Several resources are available online for learning and disseminating information regarding lab sustainability. LEAF or Laboratory Efficiency Assessment Framework, an independent standard for good environmental practice in labs designed by University College London, provides training, a toolkit, resources and strategies for improving the sustainability and efficiency of labs. Behavior As behavioral changes take time, training on lab sustainability must be augmented with monitoring, reminders and encouragement to implement practices that will lead to smaller carbon footprints. Educational institutions have started initiatives such as “Unplug” and “Shut the Sash” competitions to encourage lab users to adopt energy-saving practices. In the “Sustainable Laboratories” report published by the Royal Society of Chemistry, recognizing and rewarding initiatives and actions that promote sustainability in the labs has been listed as the first plan of action along with providing resources, funds and advocating for change. Leaders can make a difference by setting the right goals and performance indicators on sustainability, encouraging collaboration and, most importantly, walking the talk. The “Green Labs Guide” published by the University of Pennsylvania and sustainable research practices on Princeton University’s EHS website provide a number of strategies and checklists that can help lab managers to make their labs sustainable. Conclusion Sustainability in a lab should become a way of life. It will not only reduce the environmental impact, but also save money in the longer term, offsetting the initial investments for achieving sustainability. A “green lab” should be the collective responsibility of all the team members. A well designed and well maintained lab also contributes to higher efficiency and productivity. Awareness and taking simple steps can go a long way in improving lab sustainability. References 1. Durgan J, et al. Immunol Cell Biol. 2023;101(4):289-301. 2. Belibani R, et al. Progress in Sustainable Energy Technologies Vol II. 2014:273-283. 3. Hersh E. Harvard School of Public Health. Three keys to sustainable lab design to improve health and safety. 2019. 4. Alves J, et al. Access Microbiol. 2021;3:000173. 5. Zuin V, et al. Green Chem. 2021;23:1594-1608. 6. Scott S. Clin Chem Lab Med. 2023;61(4):638-641. 31 TECHNOLOGYNETWORKS.COM Failing Faster, Succeeding Sooner: Technologies That Are Accelerating Drug Discovery Anthony King Most drug hopefuls fall flat. Approximately 90% of drug candidates fail clinical development, costing hundreds of millions to billions of dollars. That’s a gargantuan waste of resources thrown into compounds that never reach patients. Drug discovery scientists at university labs and in big pharma, along with patients and governments, want more winners. One solution is to have drugs fail faster, fail earlier and fail completely – while it sounds contradictory, this could reduce costs and efforts invested into compounds that ultimately never make it onto a patient’s prescription. Researchers now have some helpful levers that they can pull to fail better and boost success rates. One is laboratory automation, where automated technologies take the place of human hands in manual and time-consuming tasks, making the research and development of possible drugs quicker and easier. A second is artificial intelligence approaches (AI), which divulge patterns in the myriad sand grains of data available, such as linking nebulous data patterns to the intervention of a potential drug compound, that no human could realistically wrap their head around. Both are being combined with an array of innovative techniques in the world of drug R&D to open new treatment avenues, which we’ll explore in this article. In silico methods for early compound evaluation The early stages of drug discovery benefits most from AI right now, says Professor Andreas Bender. “There’s lots of machine learning in early-stage ligand discovery, but the problem is always translation to the clinic,” meaning taking the research from laboratory to patient bedside. An analogy for ligand discovery is finding a key that fits and turns a lock, usually a protein involved in a disease. How the key and lock interact is pure physics and chemistry, and this is something that algorithms and computers handle well. There are treasure troves of public data that can be reached into, such as ChEMBL, a curated database of bioactive molecules with drug-like properties. Bender taps in silico computer methods, essentially experiments performed via a computer simulation, to evaluate compounds early on. Such calculations can instantly reveal that a compound’s chemical structure will lead to its rapid clearance from the body, meaning dosing will be a problem in patients. “Despite having huge amounts of data, predictive models are always fallible. You need experimental data to validate,” says Professor Miraz Rahman, medicinal chemist at King’s College London, UK. Lab of the Future Credit: iStock 32 TECHNOLOGYNETWORKS.COM Lab of the Future Credit: iStock But biology gets messier as you move towards humans, which is why the big hurdle is often Phase II clinical trials, when a drug is tested to see if it carries efficacy and can make a difference to patients’ lives. Often the information fed into an algorithm to try to improve on patient efficacy comes from lab animals, but the algorithm usually does not know about underlying conditions that impinge on the results from the animals, such as the age, genetics or sex of the organism. This situation is set to improve. “Experimental datasets will increase over the next 5 to 10 years probably,” says Miraz. “This will enrich AI models and will likely make the success rate significantly higher.” That is not to say machine learning cannot help now. AI crossed with folk medicine Bender investigates natural compounds believed to have medicinal qualities for their influence on the formation of blood vessels, known as angiogenesis. He used machine learning to identify such plant compounds from texts online, where they are often described as molecules of interest in folk medicines. Such botanical products might prep the pace of blood vessel formation and benefit heart attack patients, or they might choke off blood supply to hungry tumors, helping cancer patients. “We tested the compounds to see if they had an effect on blood vessel formation,” recalls Bender, who collaborated with colleagues in China. Zebrafish embryos – less than a week old – were placed in trays with 96 wells before varying doses of the compounds were added to each well. Zebrafish embryos are transparent, and automation in the lab allowed each embryo to be imaged using a microscope and the data fed directly into an algorithm. This revealed which compounds, and at what dose, influenced angiogenesis. “You get good bang for your buck and quite a lot of return from your time and effort,” Bender comments on the pairing of AI and automation in this context. It’s an approach that is well established, and advances at this stage in drug discovery can help move compounds closer to use in patients. “What matters is getting better compounds into the clinic. That’s where the real impact comes from,” says Bender. “Meaning the right compound and the right dose into the right patients.” When combined with automated and cell imaging and other laboratory techniques, AI can also make great inroads in identifying patient subgroups that might benefit the most from a specific treatment. This has been most notable for cancer patients, where researchers have developed successful treatments of patients by pinpointing drug targets on their tumors. This tailored approach requires an in-depth understanding of their individual disease, something machine learning can assist with. For example, clinical data and computer tomographic images have been combined to identify which patients with colorectal cancer are likely to also develop metastatic liver cancer, and how best to treat them. Beyond cancer, there is also hope that machine learning can aid the discovery of drugs for central nervous system diseases, as noted in a recent review. Conversely, in the 1990s, high-throughput screening of compound libraries pinpointed many interesting candidates, but after testing and moves towards patients, the approach turned out to be largely disappointing in terms of patient impact. Lack of disease understanding was often to blame. “If you poke in the dark, you will reduce your chances of success,” says Bender. Instead, AI works Bufobufo gargarizans. Cinobufotalin is the primary component of Chan-Su, an extract from the paratoid glands of B. gargarizans. Bender and colleagues found cinobufotalin could inhibit endoethelial tube formation in vitro, but promoted angiogenesis in zebrafish. The findings suggest the active ingredient has unknown pharmacological effects, which should be explored in more detail. 33 TECHNOLOGYNETWORKS.COM Lab of the Future Credit: iStock best if there is a working hypothesis on how to treat a disease, as well as relevant data to feed into an algorithm; then, in silico predictions can deliver testable hypothesis which, followed by experiment, can then identify diseasemodifying compounds. Beating drug resistance One area of drug R&D that has particularly struggled is the discovery of new classes of antibiotics. Microbes that are resistant to existing drugs are deemed one of the top global public health threats facing humanity by the World Health Organization. With effective antimicrobials, medical procedures such as cesarean sections, hip replacement, cancer chemotherapy and organ transplantations will become far riskier. “Globally, recent data indicates that close to one million deaths per year are due to these types of infections,” says Dr. Jose Bengoechea, a professor of biomedical sciences at Queen’s University Belfast, UK. Bengoechea has just embarked on a project funded by the Medical Research Council in the UK that will use AI and machine learning to find new ways of turning off microbial defenses against antibiotics. Previously, AI has been tapped to identify how microbes gained resistance in healthcare settings such as hospitals. Often this relies on sequencing genomes of the microbes. But Bengoechea’s new project seeks previously unknown protein targets on a troublesome bacterial species, Klebsiella pneumoniae. This is a common bacteria found in the environment and naturally inside and outside the human body, but some strains can cause pneumonia, wound infections and blood infections in hospital patients, which are difficult to treat with standard antibiotics. As part of his project, Bengoechea seeks new ways to block mechanisms of resistance in bacteria by combining AI and lab experiments. “This will allow us to make discoveries faster than traditional approaches,” he says. He explains that the problem with traditional approaches is that they keep turning up the same drug targets: “We are getting the same hits all the time.” AI brings the opportunity to discover new ways to make Klebsiella susceptible to existing antibiotics and our own defenses, which will be cross-checked with existing drugs. “We will interrogate libraries of drugs that already have received approval,” says Bengoechea, “dramatically cutting the cost and time it takes for a drug to be tested in patients.” Results can then be checked in lab experiments. If AI with other experiments riding in tandem succeeds in reducing the defenses of K. pneumoniae, the lab will explore whether the strategy can be deployed against other troublesome organisms resistant to our antimicrobials. Fewer animal tests, and mini-me cancers Newer approaches that lean on AI and automation can reduce the reliance on animal testing in drug discovery and developments. In principle, animal tests have not been accepted for cosmetics in Europe for some time, while recently the US removed the rule whereby tests on two animal species were needed before drugs could go to human trials. “I see this as a step forward, because the bar is so low,” says Bender. “Animal tests are not sufficiently predictive.” Results from drug tests on rats and mice often differ from each other, never mind from people, and it is an oft used quip that we have drugs that cured many diseases in mice, but failed where it matters – humans. Now, in a move away from animals, many scientists tap previous experiments for running in silico tests, coupled with new lab experiments that don’t involve rats, mice or guinea pigs. For example, an anti-cancer compound might be added to cells in a dish or mini-organs (organoids) in 96-well or 384- well plates. An automated imaging system tracks changes to the mini-organ or monitors what happens inside of the cells, to see if expected changes occur. Once upon a time, potential cancer drugs were mostly tested on human cancer cells by applying them to immortalized tumor cell Nuclei, microfilaments and membrane particles in HeLa cells, a prolific cell line that has been central to advances in modern medicine. 34 TECHNOLOGYNETWORKS.COM Lab of the Future Credit: iStock lines organized in single layers on a dish or suspended in a flask. One of the most famous is HeLa, a prolific line of cells that originated from cervical cancer cells taken from a US woman – Henrietta Lacks – who died of cancer in 1951. Biologists could keep these cells alive beyond a few days in the lab, and experiments using them contributed greatly to medical advances. However, a problem is that such cells adapt to life in the lab. Worse, perhaps, layers of cells are not the same as the 3D milieu of cells in most human tumors, so that cell lines do not mirror real-life cancer – and there is lots of variation within tumor types. This is where organoids enter the story. Cancer organoids are tiny three-dimensional (3D) tissues that can be grown in a lab from patient tissue obtained during surgery or biopsy. They are predicted to revolutionize our understanding of patient-specific tumours. “If you want to find a drug that works for a patient’s tumour, then you make sure the organoids you are testing are as close as possible to the patient,” says Dr Alice Soragni, who heads a lab at the University of California, Los Angeles, that uses organoids to better understand rare cancers. She seeds 96- well plates with organoids taken directly from patients at surgery. Handling error and variabilities in techniques are kept to a minimum because of automation use and how the organoids are placed into the wells as a ring. Once the “mini tumors” are in place, Soragni hits the button on an automated system that monitors them while they are exposed to different doses of off-the-shelf drugs over several days. In one set-up, she can use an imaging technique called interferometry that allows her to weigh and image the cells in real-time, to see how they are responding to a treatment. The Soragni lab is now also printing cells from patients with rare sarcoma tumours, embedded in a support matrix. “This is a high throughput platform powered by machinelearning based tools,” Soragni explains, “which allows us to extract as much information as possible from the images.” Her team seeks to leverage machine learning tools to link tell-tale characteristics of tumor organoids, which would not be picked up by eye, to their response to therapy. Organoids – tied with automated approaches – are making headway in other areas too. In a Chinese study, researchers profiled the chromosomes of 84 pancreatic cancer organoids to analyze their response to hundreds of chemicals and five chemotherapies. Meanwhile, scientists in Germany developed a human midbrain organoid to screen 84 drugs, pesticides and other chemicals for toxicity. Their fully automated set-up identified a flame retardant and a pesticide as toxic to dopaminergic neurons through analyzing microscope images. These neurons, which release the neurotransmitter dopamine, are of interest partly because their deterioration is the hallmark of Parkinson’s disease. Back in California, Soragni is also interested in a genetic condition, neurofibromatosis type 1, which causes tumors to grow under the skin of patients. The tumors do not spread or become cancerous, but they can cause a patient pain and distress. Surgery is one option, but Soragni has obtained patient tumor samples to grow into organoids in a quest for treatments. She is testing existing drugs on these patient-derived organoids to try to help these patients. “Automation makes this possible, because manual screening would be technically challenging, time consuming and labor-intensive, and could also suffer from operator issues,” explains Soragni, referring to the fact that individuals in a lab can vary in their techniques and introduce variability. “Once you have automation, everything becomes not only faster but also more robust and easier,” she adds. She is excited about the vistas that lab automation and organoids open in terms of discovering new drugs, but also in trying them out against tissue from an individual patient. “We don’t even need that much information about the tumor per se,” enthuses Soragni. “We can take a tumor, shower it with different drugs and let the tumor tell us what works best.” The ultimate beneficiary of AI and automation should be patients, with newer therapeutics brought to market that are more tailored to them as individuals, boosting effectiveness and diminishing side effects – the right drug for the right patient. For Research Use Only. Not for use in diagnostic procedures. © 2023 Thermo Fisher Scientific Inc. All rights reserved. All trademarks are the property of Thermo Fisher Scientific and its subsidiaries unless otherwise specified. AD001871-EN 0223M Built for stability. Built for performance. Why compromise? Thermo Scientific™ Chromeleon™ 7.3.2 Chromatography Data System delivers vast improvements in performance, compliance, and overall usability while increasing the resilience of network operations. With greater processing power and additional network structure, the entire CDS performance is amplified while providing IT with all the tools for security, stability and maintenance without compromising the needs of the lab. It’s the best Chromeleon CDS yet. Learn more at thermofisher.com/chromeleon 36 TECHNOLOGYNETWORKS.COM How To Future-Proof Your LIMS: Handling Software Updates Bob McDowall, PhD The shortest part of a commercial LIMS life cycle is the time between the selection, implementation (including computer validation for regulated laboratories), user training and roll-out for operational use. This was covered in two earlier articles for Technology Networks covering LIMS selection and implementation. More recently, an article on How to Future-Proof Your Laboratory Informatics Environment looked at a wider informatics environment in a laboratory. In this article, we continue the journey of how-to future-proof a LIMS by ensuring that it is kept current with software updates. This is the longest part of the system life cycle and the one that must be managed proactively to get the most out of your investment in the system and ensure that it is futureproofed. Here, the focus will be on updates to the LIMS application itself rather than IT infrastructure and support e.g., operating system patches or changes to the computing platform. The assumptions made in this article are: • Laboratory processes have been mapped and redesigned to eliminate paper and spreadsheet calculations with use of electronic signatures • Key instruments and laboratory computerized systems are interfaced to the LIMS • There is a current written Requirements Specification(s) (User and Configuration) that documents the LIMS processes and application configuration • A laboratory has access to a number of LIMS environments e.g., DEV (development), TEST or VAL (validation) and PROD (production) to enable upgrades to be evaluated before they are implemented LIMS installations Before we can start discussing LIMS application software releases and upgrades we have to consider that a LIMS has been implemented in one of two ways: • In-house or on-premises installation: The LIMS can be installed on physical or virtual hardware located either in the company data center or in an off-site data hotel. The system, IT infrastructure and the data stored on it is under the direct control of the company. • Software as a Service (SaaS) Installation: The software is leased from the LIMS company and is operated on virtual infrastructure using a cloud service provider. Put simply: your data on someone else’s computer. There is no direct control by the laboratory. The only control is via the performance metrics defined in the agreement with the SaaS provider. These two different implementations lead to two vastly different approaches to handling updates: one is voluntary, the other mandatory. On-premises LIMS installations tend to be static and SaaS LIMS are dynamic, as we shall see now. Lab of the Future Credit: iStock 37 TECHNOLOGYNETWORKS.COM Lab of the Future Credit: iStock Types of LIMS upgrades All software needs to be maintained and updates can be of two main types: • Error or bug fixes: If not for users, all software would be perfect! During operational use, software defects, errors or bugs will be found and reported to the LIMS supplier. These will be classified by the supplier from fatal to cosmetic. Fatal and high impact errors will be resolved relatively quickly, as a software patch or hotfix will be released for users to install to resolve the problem. Whilst implementing patches or hotfixes needs to be controlled, there is no need for major retesting or revalidation activities as the error is typically localized; however, read the release notes first to understand the impact of the fix. • Increases in functionality: These upgrades are in response to existing customer input for new or improved functions or may also arise from potential new customers who will only select the system if additional specified functionality will be available. These upgrades need to be handled more carefully as changes in the way the system operates now may impact current processes that could have been extensively configured or customized. Again, read the release notes and understand the changes and their impact on the system. On-premises installation and LIMS upgrades An on-premise LIMS installation is the easiest to manage and control software upgrades as the laboratory, quality and IT are all involved with the process and it can take place at a time to suit everybody. However, this can be an Achilles heel as it can lead to deferring several updates as the laboratory might be too busy or it is not seen as adding value. In a GXP regulated laboratory the maxim is we have spent so much time validating the system – DON’T touch it! However, this is a complacent attitude as all commercial software, including LIMS, has a defined lifetime. Many software companies will support the current major release and the previous one. Therefore, after a time software maintenance is limited to known bugs only support. I have known several laboratories that have left their system static without any changes for a number of years and either they have to support the system internally or have a major upgrade and data migration project on their hands. Neither option is an ideal situation. It is much better to have frequent incremental upgrades than take the Big Bang option to upgrades. Assessment of the release notes and evaluation of the changes in a DEV environment allow a laboratory to assess whether any retesting or validation is actually needed. A LIMS that has been configured (turning functions on or off and parameterizing others) is easier potentially to upgrade than a customized LIMS (using either a vendor supplied language or a recognized software language). The latter is more difficult especially if the customization rules of the supplier have been ignored as regression testing should be conducted to see if the upgrade has impact any of the customized functions. SaaS installation and LIMS upgrades The situation with LIMS application upgrades is completely different with a SaaS installation. Remember this is your data on someone else’s computer and you cannot control directly. Under the agreement you will be required to take ALL upgrades to the system. You will read in the agreement that there is no opt-out. Given an Agile development process, this could mean a new upgrade every 3–4 months. If you are a regulated laboratory, this is a major a problem as you have to consider revalidating the system with each new upgrade. Welcome to the validation treadmill. Regardless of the environment a laboratory works in, regular enforced upgrades must be managed effectively and the overall process is shown in Figure 1. Before signing the agreement, you did read and understand what was involved with the acceptance and timing of upgrades. You did, didn’t you? Let’s assume that there are quarterly upgrades per year not counting hot fixes, the following process should apply as shown in Figure 1: • At the start of the development cycle (shown in blue), the new features for the release will be selected from a backlog and developed (coding, testing and resolution of errors) over the course of several two-week sprints. • Towards the end of the second month, the new features for the next version will be known and information for the release notes prepared. These notes must describe what has been changed in the new release and these must be given to customers in good time so that they can determine if they want to use the new feature and, if so, understand the impact on their processes. This and the test environment are shown in yellow in Figure 1. • A test environment must be available for customers to assess the new release and decide if they will use any of the new features offered. Note: you will get these features regardless, whether you want them or not, as this is a condition of the agreement signed earlier. • The LIMS SaaS provider also needs to allow enough time for a customer to evaluate the new features and decide if they want to use them or not and to see that 38 TECHNOLOGYNETWORKS.COM Lab of the Future Credit: Bob McDowall they work correctly. The period available for this will be defined in the agreement between the two parties. • A major problem with GXP regulated laboratories is that normally the change control process is slow plus the time required to generate and execute the validation documentation requires a complete rethink to stay within the agreement timelines. Validation needs to get smarter and more flexible as this is repeated every three months and you don’t get time off for good behavior. Now do you understand the term validation treadmill? • After a defined period in the test environment, the release will be pushed to the operational environment, shown in green, whether you like it or not (remember the agreement you signed with the supplier?) I am not trying to put you off from a SaaS LIMS, I want you to be prepared for reality. Summary Software upgrades are a way of life: use software – you must manage upgrades proactively. You can take the ostrich approach and ignore them all or just implement bug fixes but if you find an unknown error when your software is out of support this potentially could result in a major project to upgrade. This can be avoided by small incremental upgrades that require smaller and simpler retesting or revalidation meaning that your LIMS is futureproofed. Testing Coding Release Backlog LIMS Supplier Software Development SaaS LIMS Test Environment SaaS LIMS Production Environment Release Notes Roll out of Version To Customers TIME CRITICAL Figure 1: Process flow of a SaaS LIMS upgrade or hotfix release.