Model Medicines is a biotechnology company focused on accelerating drug discovery and early-stage development through artificial intelligence–based platforms that model chemistry and human biology. Founded in 2019 to realize the vision of AI-driven drug discovery, the company is headquartered in La Jolla, California. Model Medicines reports that it has produced 67 validated assets and 192 compounds across 12 therapeutic targets through its multimodal platform, GALILEO™, which integrates multimodal data, modeling, and drug design. The company develops patent-pending therapeutic candidates in oncology, virology, inflammation, neurological and metabolic/weight disorders, and gastrointestinal diseases.

History and Organization

Model Medicines was established in 2019 and operates with a multidisciplinary team specializing in artificial intelligence, machine learning, chemistry, and pharmaceutical development. The leadership and research team includes Daniel Haders II (Founder, CEO), Sean M. Russell (Founder, COO), Patrick O’Neill (Partnerships and Investor Relations), Tyler Umansky (Head of Machine Learning), and Virgil Woods (Senior Scientist). The clinical and preclinical advisory board features Davey Smith, Launa Aspeslet, Philippe A. Gallay, Adolfo Garcia-Sastre, Kris White, Tushar Menon, David Garvey, William Brubaker, Rubén Darío Flores Saaib, and James Weber, who provides expertise in health economics.

Technology Platform: GALILEO™

GALILEO™ is an artificial intelligence platform designed to efficiently scan the vast chemical and biological space by combining multimodal data, advanced modeling algorithms, and molecular design tools.

The data layer aims to generate scalable, first-principles-based biochemical signals derived from three-dimensional protein structures, referred to as Constellation™. This data production pipeline integrates hypothesis-driven insights from primary literature and pharmacophore modeling to produce purpose-specific datasets. According to the company, these datasets surpass commercial reference databases in terms of data source diversity, QSAR bioactivity volume, biological coverage, unique chemical structure variety, and number of potent bioactive compounds.

The modeling layer incorporates generative models such as autoencoders, variational autoencoders, and generative adversarial networks, which interact with predictive and learning components like CHEMPrint™ and Constellation™. CHEMPrint™ is designed for zero-shot activity prediction in unexplored chemical spaces, while Constellation™ focuses on analyzing binding modes and target regions by learning atomic interaction patterns from X-ray crystallography and cryo-EM data.

The design layer applies the Tanimoto similarity metric to quantitatively evaluate new chemical entities with low similarity to training data and high drug-likeness. Maintaining a similarity threshold below 0.4 is intended to preserve chemical novelty.

Model Medicines has also proposed a framework called “The AIDD Code” for assessing the maturity of AI-driven drug discovery (AIDD) ventures. This framework includes criteria such as novel biology/target identification, new chemical entity generation, zero-shot hit rates, low Tanimoto similarity, broad multi-disease pipelines, and preclinical proof-of-concept in gold-standard animal models. Within this framework, the company reports identifying the Thumb-1 region of RdRp as a universal RNA antiviral target family and developing MDL-001, a broad-spectrum polymerase inhibitor candidate.

Research and Development

In virology, the company develops direct-acting agents with broad activity profiles and advanced antiviral properties. The lead candidate MDL-001 targets the Thumb-1 domain of viral RNA-dependent RNA polymerase (RdRp) and is positioned to achieve a clinical-stage safety profile. Multiple compounds have been validated in vitro across targets such as RdRp, 3CLpro, BIRC2, CCND3, KDM5B, DAGLA, MAPK11, and KDM1A.

The oncology program has produced new candidates for the broad-spectrum targets BRD4 and AXL. The BRD4-targeted compound MDL-4101 is being advanced toward late-stage preparation with a focus on thyroid cancer, and additional assets have been validated for both BRD4 and AXL across various indications.

In emerging therapeutic areas, the inflammation program has identified more than 50 compounds through computational discovery aimed at anti-neutrophil cytoplasmic antibody–associated vasculitis. In the context of longevity biology, the company investigates links among gastrointestinal dysmotility, mental health, metabolism, and neurological disorders, conducting discovery at both the target and compound levels.

Scientific Outputs and Validation

Model Medicines has published preprints and technical reports demonstrating the performance of GALILEO™ in end-to-end compound generation and candidate prioritization, the generalization capacity of its AmesNet deep learning model in Ames mutagenicity prediction, and single-shot identification of broad-spectrum antiviral chemical libraries.

The company reports that its data have been independently validated in leading academic and institutional laboratories. Specifically, MDL-001 has undergone preclinical evaluations in gold-standard animal models at the Garcia-Sastre Laboratory of the Icahn School of Medicine at Mount Sinai. Research disclosed for the 2024–2025 period highlights the characterization of the RdRp Thumb-1 target family, the development of broad-spectrum inhibitors, and methodological details of AI-enabled chemical space expansion.

Corporate Structure

Model Medicines is headquartered in La Jolla, California, with research spanning virology, oncology, inflammation, and longevity biology. The company emphasizes its guiding principle that “humanity cannot wait decades,” focusing on the rapid discovery and validation of candidates aligned with target product profiles and high clinical success potential. It reports reaching 67 validated assets and 192 compounds across 12 therapeutic targets.

Through its GALILEO™ platform, Model Medicines builds a portfolio spanning multiple disease domains by integrating scalable protein-structure-based data generation, coordinated use of generative and predictive models, and systematic in-silico/in-vitro validation cycles. Reported findings demonstrate capabilities in novel target identification, proposal of low-similarity chemical entities, and high hit rates in zero-shot experiments. Corporate statements indicate that the platform’s outputs have been independently evaluated in academic and institutional settings, with leading candidates advancing toward preclinical stages.