Drug discovery has always been a slow and laborious process by nature, with researchers finding dead ends much more frequently than successful routes to the clinic. The advent of AI promises to chart paths through that maze much faster, and it’s kicked off a gold rush of projects and collaborations as biopharma companies aim to reach their goals first.
One public-private partnership could help illuminate the way forward for others: The ATOM consortium, for Accelerating Therapeutics for Opportunities in Medicine, encompasses a suite of projects to hasten precompetitive drug discovery practices—with the objective of getting a potential drug from a target to patient testing in less than a year.
Founding members of ATOM, which was launched two years ago in part through the U.S. government’s cancer moonshot efforts, include Lawrence Livermore National Laboratory, the Frederick National Laboratory for Cancer Research, the University of California, San Francisco, and GlaxoSmithKline.
The consortium intends to make new tools, models and processes broadly available, with a large focus on solutions geared toward cancer patients. It’s working on multidisciplinary approaches to incorporate supercomputing simulations, data science, machine learning and AI into a team-based setting that could be applied to other disease areas.
Researchers hope that ATOM, which is built on publicly available data—as well as data provided by GSK and its growing number of members, plus funding from the 21st Century Cures Act—will enable them to find ways to better predict how molecules will act once they’re inside the body.
The big pharma is bringing chemical and in vitro data for more than 2 million compounds to the table, plus preclinical and clinical information on 500 molecules that have failed at different points along their development.
Lawrence Livermore, meanwhile, will supply its supercomputing hardware and software, while the Frederick lab, on behalf of the National Cancer Institute, and UCSF will provide their expertise in cancer biology, drug development and patient care.
One ATOM project looks to predict a drug’s impact and toxicity on the liver as early as possible, by using computer models, cell-based in vitro assays and organoids. Such a tool could save the pharmaceutical industry millions and avoid patient harm by reducing the rate of clinical trial failures, since drug-induced liver injuries rank as a common cause of safety-related drug withdrawals.
By gauging the potential for liver injury not just in an individual patient but among the wider population as well, ATOM hopes to make its methods for safety testing a regular part of drug discovery and molecular design programs.
Other ATOM research includes developing machine learning-based screening of new molecular entities for heart-based toxicity, building in silico pipelines for predicting drug pharmacokinetics, and working to train deep neural networks using three-dimensional and dynamic molecule models.
Earlier this year, ATOM launched a collaboration with Nvidia, maker of graphics cards and processing units, to help advance its computing platforms and make GPU-accelerated supercomputing more accessible to researchers.
“Drug discovery is in the perfect position to take full advantage of AI, with advancements in high-throughput technologies, AI algorithms and high performance computing,” Nvidia’s healthcare VP, Kimberly Powell, said in the April announcement. “Working with the ATOM team of experts in drug discovery, precision oncology and supercomputing, we will democratize a data-driven, scalable, AI drug discovery workflow that reduces the time and cost to discover new drugs and accelerate the journey to precision medicine.”