MDL-001

CEO, Daniel Haders II, PhD, to address the inference throughput barrier in AI-driven drug discovery and lead a roundtable on deploying LLMs and AI agents to navigate the multi-parameter optimization challenge in small molecule drug design San Diego, CA — 3/17/26 — Model Medicines , an AI-first biotechnology company engineering first-in-class therapeutics, today announced that Daniel Haders II, PhD, Founder and CEO, will present on the transformative power of Ultra-Large Virtual Screening (ULVS) and lead a roundtable discussion highlighting Agentic AI at the upcoming AI Convergence: Small Molecule Discovery Summit , March 19–20, 2026, in Boston, Massachusetts. Presentation & Roundtable Schedule ● Roundtable: Building Agents & Exploring Emerging LLM Use Cases for Small Molecule Discovery Funnel Applications - Thursday, Mar 19, 2026 ● Presentation: Breaking the Throughput Barrier: Ultra-Large Virtual Screening as a Precision Amplifier for AI-Driven Drug Discovery - Friday, Mar 20, 2026 Reaching Into Deep Chemical Space The featured presentation will demonstrate that inference throughput is the bottleneck impeding the full potential of AI-driven drug discovery, and that Model Medicines’ Ultra-Large Virtual Screening architecture unlocks that potential . Laboratory-based High-Throughput Screening (HTS) has the potential to evaluate one million compounds per day. [1] State-of-the-Art (SOTA) AI-Driven drug discovery campaigns are capable of screening two million (BioHive) to eight billion (Atomwise) compounds per day. [2] Together, laboratory-based and current SOTA AI-driven screenings have been constrained to a ceiling of 8E+9 in chemical space. It is estimated that all potential drug-like compounds exist in a chemical space of 1E+60. [3] The scale of compounds screened for a given therapeutic target, relative to the potential chemical space, has been described as less than a single drop of water in all the world’s oceans. The result is incremental improvements to known target-ligand chemistry and the rediscovery of known scaffolds. Novel chemistry resides in deeper, unexplored regions of chemical space that current methods cannot access due to insufficient throughput. Model Medicines’ Ultra-Large Virtual Screening (ULVS) approach overcomes this constraint and unlocks AI-driven drug discovery. The company executed a 325-billion-compound ULVS in a day in 2025 in partnership with Google. [4] This was the largest machine-learning–driven bioactivity screen publicly reported to date. [5] Touching 3E+11 chemical space, this approach was the foundation for the development of two first-in-category programs against the “undruggable” transcription factor BRD4 and the novel broad-spectrum RdRp Thumb-1 target. This year, the company announced that it is constructing a one-trillion-compound (1E+12) scale screen. “Inference throughput is the discovery variable holding back AI-driven drug discovery from revolutionizing medicine,” said Daniel Haders, PhD, Founder and CEO of Model Medicines. “Trillion-scale Ultra-Large Virtual Screening regimes fundamentally change what chemistry can be discovered, what diseases can be solved, and how many patients can be reached.” Drug Discovery as a Multi-Parameter Optimization Problem The roundtable discussion will build on previous talks Dr. Haders has delivered on AI agents and LLMs and their ability to address drug discovery’s defining challenge: Multi-Parameter Optimization. Every drug program is guided by a Target Product Profile (TPP) that maps back to the patient. The list spans indication and intended use, target population, efficacy goals, safety and tolerability, dosage form, route and frequency of administration, storage and stability, market access, patient experience, regulatory milestones, and manufacturing feasibility. Translating a TPP into a molecule means simultaneously optimizing across affinity, potency, selectivity, solubility, oral bioavailability, tissue partition, half-life, pharmacokinetics, drug-drug interactions, safety and tolerability, ADME properties, synthesizability, and chemical and physical stability. Model Medicines recently published and released AmesNet, an agent that replaces the regulatory required Ames genotoxicity test. AmesNet outperformed all literature-reported Ames agents, including the FDA’s DeepAmes, Baidu’s GROVER, and MIT’s ChemProp. [6] Using AmesNet as an example, the roundtable will examine how agent-based systems and emerging LLM applications can be deployed across the discovery funnel to navigate multi-parameter optimization complexity. Here, the objective is to accelerate decision-making, coordinate across functional disciplines, and maintain alignment with the TPP as programs advance toward the clinic. About Model Medicines Model Medicines is an AI-first biotechnology company engineering first-in-class small molecules that target the biological linchpins underlying disease. The company’s research spans infectious disease, oncology, and inflammation, with programs designed around conserved molecular choke points that drive multiple pathologies. Model Medicines has discovered a direct-acting, non-nucleoside, broad-spectrum antiviral (MDL-001) and a BRD4 inhibitor with no measurable activity against BRD2/3 (MDL-4102). Its work demonstrates how large-scale computation can uncover entirely new classes of drugs once thought unreachable. Model Medicines is advancing a new generation of therapeutics that redefine what is possible in modern drug discovery. Learn more at Media Contact: Patrick O’Neill Head of Partnerships & Investor Relations media@modelmedicines.com [1] AstraZeneca. High throughput screening at The DISC: a new era of drug discovery [Video]. YouTube. February 12, 2025 [Accessed March 16, 2026]. Available from: [2] Model Medicines. Record-Scale AI Screening with Model Medicines on Google Cloud: GALILEO™ Achieves 325 Billion Molecule Throughput for Oncology Drug Discovery. Available from: [3] Reymond JL. The Chemical Space Project. Acc Chem Res. 2015 Mar 17;48(3):722-30. doi: 10.1021/ar500432k. [4] Google Cloud. LA Tech Week - AI for Startups in Healthcare Lifesciences [Internet]. Venice (CA): Google; 2025 Oct 17. Available from: [5] Google Cloud. Google Cloud to host second-annual Cancer AI Symposium in New York City [Internet]. New York: PRNewswire; 2025 Oct 30. Available from: [6] Umansky T, Woods V, Russell SM, Haders D. AmesNet: A Task-Conditioned Deep Learning Model with Enhanced Sensitivity and Generalization in Ames Mutagenicity Prediction. bioRxiv [Preprint]. 2026 Feb 11 [cited 2026 Mar 16]:[15 p.]. Available from: