CATALYST program to fast-track safer medicines from lab to patients 

Program’s data and computational tools will shorten drug development timelines, lower costs, and improve patient safety  

The Advanced Research Projects Agency for Health (ARPA-H), an agency within the U.S. Department of Health and Human Services (HHS), today announced the research and development teams receiving awards from its Computational ADME-Tox and Physiology Analysis for Safer Therapeutics (CATALYST) program. Leveraging artificial intelligence and machine learning, CATALYST intends to develop computer models that mimic real human biology to predict safety and effectiveness for Investigational New Drug (IND) candidates, ensuring that only the most promising and safest medicines move forward to patients.  

For decades, drug developers have relied on animal models to predict how new medicines will work in people. But these models fall short, leading to long, costly development cycles and missed opportunities for patients—especially children and pregnant women, who are frequently excluded from clinical trials. CATALYST will transform how drug safety is tested in both preclinical and clinical studies and build a pipeline of tools that drug developers and regulators can use. If we can make truly predictive in silico models of drug safety and toxicity, we would shift the paradigm of drug development and testing, bringing therapies to patients faster and at lower cost; animal models will no longer be the proxy for how a drug behaves in humans.  

“We are committed to modernizing our regulatory processes to lower costs, shorten timelines, and increase our confidence in the safety and efficacy of pharmaceuticals,” said FDA Commissioner Marty Makary, M.D., M.P.H. “Advancements in drug safety testing have given us better predictive insights and allowed us to radically reduce or eliminate animal testing — it’s critical that we continue to advance in this field.” 

The program brings together top performer teams to build and validate computer models to predict how drugs behave in the human body, with specific attention to pharmacokinetics, including absorption, distribution, metabolism, and excretion (ADME), and pharmacodynamics. These human-based in silico models will be tested in real-world scenarios, with input from regulators and drug developers, to ensure they are ready for use in IND applications. CATALYST will drive adoption of these predictive tools for use in regulatory filings by supporting their use in preclinical pharmaceutical drug development. This distinctive approach brings product sponsors (pharmaceutical companies) in at the beginning to enable such an early demonstration of the technology.    

“Too many promising medicines fail late, after years of work and enormous cost, because our best tools still don’t reliably predict how a drug will behave in people. With today’s CATALYST awards, ARPA-H is backing cutting-edge teams to build human-based, AI-enabled models that can forecast drug safety and efficacy long before the first clinical trial,” said Alicia Jackson, Ph.D., ARPA-H Director. “This is an ARPA-level push to move beyond over-reliance on animal models, cut years and cost from development, and give regulators and product sponsors better evidence to protect patients — including children, during pregnancy, and others who are too often left out of trials.”  

The CATALYST program total award amount is up to $125 million over 4.5 years. Performer contract awards* (not procurement contracts, grants, or cooperative agreements) vary in funding amount per awardee and are contingent upon each team meeting aggressive and accelerated milestones.  

“The time and cost to get a new drug from discovery, through preclinical testing and clinical trials, and, finally, to people who need it, are enormous. The majority of candidates fail during this process because we can’t accurately predict drug effects in humans,” said CATALYST Program Manager Andy Kilianski, Ph.D. ”CATALYST’s data and computational tools will make shorter development timelines, less expensive therapies, and better patient safety a reality. We are excited to work with our partners at the FDA, NIH, and across government, academia, and industry to make this vision a reality.”   

CATALYST performer teams are led by:  

• The Charles Stark Draper Laboratory in Cambridge, Mass. seeks to develop a patient-centered, predictive model using the patient data layer, human organ data layer, biopsy data layer, and microphysiology data layer. This team will build a human data stack that can be used to understand how different patients respond to novel therapies.
• Deep Origin in San Francisco intends to build an FDA-qualifiable, self-improving in silico ADME-Tox and dosage prediction platform aimed at reducing and eventually replacing animal models, so drug developers can test how a medicine moves through and affects the human body on a computer before it ever reaches a person.
• Inductive Bio in Brooklyn, N.Y. aims to transform preclinical drug safety assessment for small molecules by building in silico organ toxicity models for liver and heart that integrate the latest AI advances with physiology-based mathematical models., helping identify potential heart and liver side effects earlier and more accurately during drug development.
• The Jackson Laboratory in Bar Harbor, Maine intends to replicate broad human cardiovascular variation to comprehensively characterize non-arrhythmic cardiotoxicity using transcriptomic, metabolic, and phenotypic analyses. By building digital cardio models based on mouse and human physiology, drugs can be evaluated against a wide range of heart genetic profiles and physiology differences based on patient populations.
• Black Mesa Technology, Inc., in Bedford, Mass. Intends to establish and document a good experimental practice (GxP) environment to assure quality and security of data and AI methods framework, ensuring the AI tools behind CATALYST are trustworthy, secure, and meet regulatory-quality standards.
• Cedars-Sinai Medical Center in Los Angeles will leverage its extensive biobank samples to derive patient-specific cell-based models of cardio- and neurotoxicity, which can be exposed to real-world stressors like aging to develop clinically accurate computational toxicity predictions. This will make safety assessments in older patients and patients with comorbidities more accurate before clinical trials.
• Peptilogics, Inc., in Pittsburgh seeks to develop ADME-Tox predictive models and methods for therapeutic peptides while including aspects related to off-target protein-ligand complexes and post-translational modifications, helping designers of next-generation peptide drugs spot unintended targets and side effects earlier in development.
• University of North Carolina at Chapel Hill aims to build antibody drug-customized organ-on-chip models and three in-silico platforms focused on improving ADME-Tox predictions in humans including pregnant individuals, so developers of antibody-based medicines can better predict how these drugs work in the body and assess safety for populations often left out of trials, including during pregnancy. 

For more on CATALYST, visit the program page. 

*Awarded in September 2025