In drug discovery, interest in harnessing the power of AI ramped up significantly with breakthroughs like AlphaFold, where AI predicted protein structures with astounding accuracy. AI’s initial focus was analyzing existing data, with machine learning systems excelling at tasks like predicting new drug interactions, molecular behaviors, and even biological pathways, based on troves of experimental data. ML can also aid in identifying promising drug targets by using natural language processing to analyze scientific literature.
But AI’s role is rapidly evolving. In 2024, AI is poised to transition from analyzing existing data to a more proactive drug discovery role as a predictor and collaborator. Shifts fueling the trend include the rise of generative AI, which can create novel molecular structures and predict their properties. Another factor driving the shift is AI-based optimization of wet lab experiments, suggesting tweaks to protocols and experimental conditions and, in the long run, the development of scores of novel drugs and protein structures designed entirely by AI algorithms.
AI set to accelerate drug discovery and synthesis in 2024
“In the past, converting an AI-designed compound into a tangible one in a lab was a long process, often spanning months or years,” said Peter Madrid, cofounder and chief scientist, Synfini, an AI-enabled drug discovery startup that spun out from SRI International. In 2024, however, the field is poised to take the next step, in which AI transitions from a predominantly lab observer to a more proactive force. “These AI tools will design laboratory procedures automatically, leading to new models that facilitate automated lab procedures,” Madrid noted. The result could be fewer lab delays and inefficiencies.
While AI seems to be omnipresent on some level, “what that actually means, I think, can vary massively,” Madrid added. Surveys involving drug discovery and development professionals tend to indicate that AI projects are often exploratory. “What’s more interesting is where [AI is] fundamentally changing how you work and leading to entirely new different workflows and processes for how you’re operating,” Madrid said.
AI: From observer to active collaborator
Advances in automation and the rise of powerful chemistry-specific AI models could transform how molecules are designed, synthesized, and tested. While fully “self-driving labs” may still be some years away, more industry observers and respected journals are seeing a growing role for intelligent robotic lab assistants in the relatively near term. While initial investment in these systems might be significant, the potential gains in efficiency and discovery yield could quickly outweigh the costs, transforming the economics of drug development.
This progress sets the stage for a research era in where the integration of AI with lab operations is a core component of the research and development process. In that vein, Nathan Collins, cofounder and head of strategic alliances and development, Synfini, sees the emergence of what he calls large chemistry models (LCMs), a sort of answer to the large language models like ChatGPT that have received so much attention over the past roughly 15 months. “Similar to interacting with ChatGPT, scientists will be able to collaborate with an integrated large chemistry model and AI driven lab automation to both design and run experiments of increasing complexity,” Collins said.
Automation and AI-powered chemistry
Collins envisions that LCMs will give chemists the opportunity to ask specific queries such as, “I need a drug candidate that effectively binds to target A with a potency of X, while avoiding any binding to target B that could result in toxicity.” The LCM will then rapidly design and execute complex synthesis and testing experiments, analyzing results to recommend and optimize the search for a solution.
While the potential is considerable, the crowded marketplace and difficulty in comparing AI offerings sharpens the need for experienced professionals who can help drug discovery professionals avoid unproductive experimentation and make good on AI’s potential to reshape drug discovery workflows. But Collins also notes a catalyst for change: a new generation of scientists more likely to embrace intuitive AI tools. “It’s been challenging for the more experienced middle-career chemists to adopt these tools and bring them into being mainstream,” he said. Collins notes that the more junior employees that Synfini has brought on board are having no problems adopting AI tools. He said, “There’s no barrier for them.”
Filed Under: Data science, Drug Discovery and Development, machine learning and AI