How AI and ML empower a new world of drug discovery

David Gosalvez, Head of Product Strategy and Zev Wisotsky, Senior Principle Marketing Manager for Biologics at Revvity Signals, highlight how three powerful AI and ML capabilities are driving lab innovation: enhancing data accessibility, enabling intelligent chemical reaction recommendations, and automating complex cellular imaging. Enhancing data accessibility with generative AI A common experience in almost every research project is spending more time finding data than analysing it. While traditional search tools have greatly improved productivity, it is all too easy to miss critical insights. The new approach, using generative AI-driven semantic search powered by large language models (LLMs), not only returns standard keyword results but also interprets the user’s intent and context, pulling up relevant documents even without exact phrasing matches. AI’s interpretation capabilities completely change the game. Semantic search uses the LLM to analyse a natural-language question and returns results based on the model’s knowledge rather than specific keywords. In addition, because AI-powered search interprets the intention of the question, not just the literal wording, it can find results across multiple languages. This enables users to catch resources that traditional tools might miss. Those same AI capabilities can summarise and even translate resources even as they are collected, automatically condensing journal articles, experiment reports, and protocols. Using AI like this is truly transformative: scientists can find data rapidly and easily, and then drag and drop tables, spreadsheets, and documents complete with a synopsis or bullet points straight into an informatics system, such as an electronic lab notebook (ELN). Enabling intelligent chemical reaction recommendations Beyond text and search applications, AI also extends into highly specialised tasks such as chemical reaction planning. By integrating deep neural networks into research workflows, AI can analyse reactions and recommend optimal conditions for synthesis. For example, AI can suggest reaction mapping, solvents, catalysts, and by-product management based on a curated set of data from known reactions. AI recommendations can minimise trial-and-error by providing a range of options, enabling researchers to spend more time on the science. Reaction recommendations offer a window into the potential of AI to automate portions of chemical research, pushing forward insights and accelerating development timelines. In the long term, this kind of automation empowers chemists to experiment more efficiently, iterating faster and narrowing down paths towards successful drug candidates. AI technologies are already enhancing human expertise by acting as a super-powered assistant, providing a greater range of choices even as scientists make the ultimate decisions. Enhancing antibody research with AI-driven developability tools As an example of a specific area, AI-powered advanced solutions are transforming how antibody research is conducted, both simplifying and accelerating complex analyses. Traditionally, evaluating antibody developability requires intensive bioinformatics expertise, with bench scientists relying on specialists to process their data— which could often take weeks. Now, with AI-driven platforms, these analyses are becoming more accessible, enabling bench scientists to independently conduct in-depth assessments and make faster, data-driven decisions. For individual researchers, these solutions reduce the time and effort needed to evaluate antibody candidates. Capabilities such as sequence liability identification, 3D structural modeling, and accessible surface area calculations are integrated into streamlined workflows, supported by intuitive dashboards. Researchers can interpret results and compare developability metrics directly, eliminating delays caused by reliance on specialised teams, enabling scientists to devote more time to experimental design and decision-making, driving efficiency and effectiveness in the lab. At a broader level, these capabilities hold significant implications for drug discovery. They provide a systematic way to evaluate and prioritise antibody candidates, ensuring that only the most promising options advance to later stages of development. By improving data accessibility and reducing the dependency on specialised computational expertise, smaller research teams and organisations are able to access and harness sophisticated analytics, leveling the playing field in therapeutic development. The wide availability of these extraordinarily powerful AI solutions contributes to accelerating the pace of antibody research while reducing associated costs, fostering greater innovation across the field. By equipping scientists with accessible, AI-driven solutions, the industry can address critical challenges in developability, paving the way for more effective and safer therapies to reach patients in need. At almost every level, AI solutions can enhance workflows and decision-making without changing the underlying science, ensuring researchers remain at the forefront of discovery. Automating complex cellular imaging As well as the broad picture, AI brings new possibilities to individual techniques; for example, bright-field cellular imaging – first developed in the 17th century – is making a powerful comeback. Traditionally, cellular imaging requires precision and consistency from an experienced researcher to detect and analyse features — a labor-intensive task requiring manual segmentation and analysis. Now, AI models can deliver dramatically improved productivity, particularly in live cell analysis. Using deep learning, AI can detect nuclei and cytoplasm across an entire imaging area. The AI algorithm scans, analyses and annotates the image based on its vast library of existing images and enables scientists to distinguish structures within live cells in real-time, reducing the hours previously required for manual segmentation. In addition, AI can capture cell motion in real-time, allowing researchers to observe how cells respond to external stimuli. By handling these highly complex tasks, AI enables scientists to analyse cellular behaviour at a speed, scale and precision previously unattainable. Particularly for oncology and virology, where understanding how cells interact, respond to drugs, or change over time is crucial, these new AI-powered solutions give researchers a clear picture of cellular dynamics, enabling insights that can directly impact therapeutic strategies and interventions. Bringing AI to life in the lab A critical element for maximising the opportunities created by AI is to make the applications available from within the research workflow. Constant context-switching — for example, hopping out to use ChatGPT and other AI assistants — is a known drag on productivity. Further, only users with detailed knowledge of how each AI platform works will be able to ensure effective, usable results. The favoured methodology is to enable access to the AI applications from within the research platform, ready for productive use without specialised expertise. Additionally, the best-practice approach is to train the AI and LLM using integrated, curated datasets, to help avoid spurious or unreliable results. The combination of these two factors greatly increases the ease of access to the power of AI solutions, enhancing the value for scientific, clinical and drug discovery researchers. As a relatively new technology, AI is maturing rapidly. With suitable guardrails and well-designed implementation, based on curated datasets, AI offers extraordinary new capabilities. From DDW Volume 26 – Issue 1, Winter 2024/25 Read the digital issue About the authors: With 25 years in scientific software and analytics, David Gosalvez combines technical and domain expertise with broad knowledge across scientific use cases. He is a passionate advocate for improving the efficiency and quality of science via innovative software solutions. As head of Product Strategy, Gosalvez guides the Revvity Signals Informatics portfolio, collaborates with pharma and chemical scientists, and integrates AI/ML with other vendors’ capabilities. Zev Wisotsky graduated from UC Santa Cruz in molecular and cellular biology and earned his PhD in Neuroscience from UC Riverside, researching taste detection in fruit flies. Transitioning from academia, he joined startups to streamline scientific processes. With expertise in tailoring software solutions for both scientific and technical needs, he's now the Senior Principle Marketing Manager for Biologics at Revvity's Signals Suite.