November 14, 2024
by Sanji Bhal, Director, Marketing & Communications, ACD/Labs
Interest & Investment in AI/ML Continues to Grow
The McKinsey Technology Trends Outlook report has highlighted generative AI investigation and adoption to be a priority over the last few years. Innovation in generative AI, applied AI, and industrializing machine learning has accelerated, even in response to a decrease in private equity investment. A recent survey conducted by the Pistoia Alliance echoes continued interest in AI with participants reporting an increase in AI/ML investment and a decrease in budgets for infrastructure technologies (LIMS, ELNs, Cloud data platforms).
These are strong indicators that scientists should be preparing to leverage artificial intelligence and machine learning in their daily work. The reality, however, is that while there are pockets of R&D where practical use of machine learning and artificial intelligence has begun, these tools have yet to make a broad impact in research labs. The majority of organizations are still trying to wrangle data.
Stepwise Digital Transformation (Dx)
Change management is an important factor in effective digital transformation. The broad and diverse scope of digital needs in an R&D organization requires careful consideration and planning. A good understanding of the current Design-Make-Test-Analyze (DMTA) cycle is necessary to help identify where digitalization can improve efficiency and execution. A stepwise transformation process, therefore, helps to ensure a sustainable digital experience.
Process and materials research relies heavily on analysis and measurement data. Analytical datasets are key assets that inform and advance product lifecycles across industrial, consumer, and life sciences commercialization. Inclusion of this data is, therefore, critical in digital transformation.
The adapted Maslow pyramid can help you think about the key areas of strategic focus for digital transformation of analytical data, from the foundational “necessities” to more aspirational goals.
What is your Current State of Digital Maturity?
The Basic Needs for a Digitalized Analytical Laboratory—Stage 1
At the very minimum, a digitalized analytical laboratory requires enabling infrastructure:
- Network priority Instrument Acquisition Systems that produce file- or folder-based data structures, allowing users to access analytical datasets from outside the lab space.
- Dedicated software applications for analysis experiments which require Instrument data to be processed or annotated for post-experiment reporting.
Interpreting data and the resulting decision-making process may be paper-based, but instruments that are networked and connected to software for data analysis and review (either at the instrument or remotely) are the minimum requirement for data digitalization. This is a progression from the typical lab of the 1970s and 80s where analytical instruments were connected to plotters and strip charts; and these were the only results you walked away with, for your sample.
Open access analytical labs in pharmaceutical medicinal/process chemistry are a classic example of these basic needs. Scientists can not only walk up to instruments to add physical samples to an autosampler queue, they can process data and generate reports from their office or lab computer.
Risk Mitigation Through Established Integrity & Security Infrastructure—Stage 2
Progression from the basic needs of a digitalized analytical laboratory requires secure data with integrity. Traceability from the sample to the final data file assures users that their data is error-free and accurate and that they can make confident decisions about next steps. The organization must, for example, be able to ensure that a sample run out of sequence or misinterpreted data can be identified. Processes and measures to establish a risk mitigating infrastructure include:
Data and Application Access Controls
Identifiers for the sample request (e.g., bar codes, sequence files), the method, and experiment metadata establish traceability between the sample and data. Furthermore, passwords or other controlled access measures ensure that only those with the correct permissions can open and manipulate data.
Processing and Analysis Audit Trail
Audit trails are compulsory in a GxP environment but establishing them for all processing and analysis routines ensures that changes to data are traceable. Including a reason for changes made to data, if necessary, is also valuable.
Knowledge Management
Manually constructed and curated data with processed and analyzed results can further help mitigate risk. A searchable repository where data is stored with chemical context or interpretations ensures it can be re-used for future decisions or to answer the right questions.
Productivity & Collaboration. Scientists Get to Do More Science—Stage 3
Automation is key to establishing a digitalized infrastructure that drives productivity and effective collaboration. Progressing to Stage 3 of digital transformation means making data accessible and consistent, removing human dependencies, and building fault tolerance for your digitalized processes (being able to detect and identify system failures easily and quickly).
Automated data source monitoring, processing and contextualization of data (metadata association), and routing of processing/analysis results to a searchable knowledgebase is an example of a Stage 3 analytical organization.
R&D milestones require a significant amount of effort from cross-functional teams that work collaboratively. The work of one researcher can have a significant impact on downstream work or be direct input for a colleague’s studies. Digitalized, well-managed data enables seamless collaboration and efficient execution of project tasks. Forced degradation studies are one example of complex studies that benefit from collaboration and productivity tools. Whether for statistically significant study design, sample-condition exposure, longitudinal sampling, analytical method selection (and/or development), and final comparative analysis; digitally mature organizations enable collaboration via integrated scientific productivity tools.
Full Digital Maturity—Stage 4
A digitally mature organization has machine augmented decision-support. Analytics, machine learning, and artificial intelligence applications continually help scientists to optimize and accelerate strategies, decisions, compliance, and governance. While scientists with domain expertise establish goals and outcomes, their decisions are supported by digitally compiled data.
While software and informatics tools have supported R&D for decades, the priority has been the scientist. Software developers and IT organizations have focused on creating an optimal user experience for humans. Today, machine centric use—artificial intelligence (AI), machine learning (ML), and other data science projects—has become a priority for most organizations. Insights from ML-friendly data can be used to assess platforms, assets, staffing allocations, and help organizations implement more efficient manual or digital procedures.
Digital maturity in a drug development organization means that informatics systems can capture and digitally represent the DMTA lifecycle. In turn, these digital datasets can be leveraged to answer new questions in the future, or used for comparative analysis. One of the most challenging problems for analytical data is normalizing the heterogeneous data while retaining context so that it can be leveraged effectively in future. Organizations at Stage 4 have achieved this data engineering hurdle and are able to apply their data to glean new insights.
Effective Decision-Support Leveraging Digital Twins—Stage 5
The aspirational pinnacle of the digitalization pyramid is Digital Twins. At the peak of digitalization information is synchronized across the organization with full digital representations of experiments and scientific studies. Robust data pipelines connect prediction systems to digital representations of DMTA cycles. These prediction systems include classic regression models, ML models, and most recently large language models.
Digital twins can be varied and numerous. Examples of digital twins that leverage analytical data might include:
- Using formulations data to predict how mixing time impacts particle size.
- Leveraging chromatographic data viewed as a resolution map to visualize the impact of changing a particular parameter on resolution.
- Digital, assembled analytical data representing analysis experiments for stability studies, metabolite identification studies, catalyst screening studies, etc.
- Contextualized analytical and chemical data representing process chemistry projects (process maps, batch genealogy), biotransformation maps for DMPK studies, plate maps for high-throughput experimentation, etc.
The Future of Digitalization
Technology Development and Digital Innovation is ever dynamic: innovators continue to offer new and exciting components, but implementation can disrupt current processes. While using digital twins is a top aspiration for Dx right now, some envision a new ultimate goal—autonomous, generative AI-guided experimentation. If this goal is realized digital systems will work to serve organizations to fully manage DMTA cycles, iteratively making decisions to push initiatives forward.
To realize either current or future aspirations, digitalization teams must understand that different parts of the organization may be at different stages of digital maturity. Progression in Dx may be achieved by identifying where each is today and identifying goals that provide return on investment (ROI). Macro scale ROI from digitalization may take years to identify. Smaller near-term goals will keep you moving in the right direction.
Contact us to discuss how we can help you on your digitalization and AI-readiness journey.
About the Author
