We're bringing Om's STAT6 discovery work to OpenLabs: using agentic small-molecule discovery to turn predictions into...

We're bringing Om's STAT6 discovery work to OpenLabs: using agentic small-molecule discovery to turn predictions into experimental evidence.

Om Therapeutics (omtx.ai) connects AI-guided molecule selection to physical experiments. Our team includes Scripps Research alumni. We built library-by-library encoded screening technology that can screen hundreds of encoded proteins against hundreds of millions of encoded small molecules. Encoding traces interactions back to each protein and molecule; sequencing reads out the multiplexed experiment.

That data underpins LULA. The training program uses approximately 500 million protein–small-molecule interaction data points from Om and public sources, covering about 8% of the human proteome.

LULA-2 combines ESM-2 protein representations and ChemBERTa molecule representations through learned cross-attention. Its interaction model learns from binding data without requiring a solved protein structure. Given an amino-acid sequence and molecule SMILES, it returns a ranking score—not a measured affinity.

On a public benchmark of 503 proteins without close experimental PDB matches and 14,139 protein–ligand pairs, LULA-2 reported AUROC 0.819 versus 0.543 for Boltz2. Some ligands overlap training; prospective experiments remain essential. https://huggingface.co/omtx/lula-2

Agentic discovery connects the steps. Through Om's MCP, API and SDK, agents can help research targets, run supported models, organize candidates and prepare experimental submissions. Om Accessible Space connects molecule selection to orderable compounds. Researchers review the science and authorize spending.

Our STAT6 workflow: • Prioritize eligible molecules with LULA-2 or another method, considering diversity, known chemistry and controls. • Procure and test accepted molecules by affinity-selection mass spectrometry (AS-MS). • Confirm promising interactions and assess selectivity and function. • Use experimental evidence to improve the next selection round and, with sufficient licensed data, evaluate model improvements.

Why STAT6? It is a transcription factor downstream of IL-4 and IL-13, central to type 2 allergic inflammation. A 2023 study identified gain-of-function STAT6 variants in 16 patients from 10 families with severe early-onset allergic disease, providing human genetic evidence for its disease role. https://pubmed.ncbi.nlm.nih.gov/36884218/

Small-molecule engagement is experimentally established, including the STAT6–AK-1690 co-crystal structure. We aim to find useful chemical starting points and generate prospective binding data. A binder could support mechanistic research or inhibitor/degrader development; therapeutic benefit requires further evidence. https://www.rcsb.org/structure/9BIG

This connects to our existing STAT6 Open Pipeline Challenge. On September 8, 2026, its public page reported 257 accepted molecules, all listed in validation, with no Binder or Non-binder outcomes reported yet.

Participants can select eligible molecules with LULA or their own methods. Om procures and tests each accepted molecule; contributors own their Result Data and control publication. Current STAT6 terms are 5,000 Wallet Credits per accepted molecule and $1,000 per payout-selected Qualified Binder, initially up to 25. Om's training license activates only for qualifying data selected within the payout cap after payment posts. https://www.omtx.ai/discovery/challenges/open-pipeline-stat6

STAT6 sits within a broader planned challenge slate: • BRD4: a well-studied BET bromodomain target with established ligand chemistry—a tractable reference for evaluating a discovery workflow. • TBXT/Brachyury: a transcription factor that chordoma cells depend on—a harder cancer target for examining how experimental data supports discovery. • STAT6: an inflammatory-disease target with human genetic evidence and an active prospective validation challenge.

BRD4 and TBXT are planned extensions, not announcements of open challenges.

We'll share the reasoning behind the STAT6 experiments, results we can disclose, and what succeeds or fails. We welcome collaborators on STAT6 biology, assay controls, medicinal chemistry and fair evaluation of agent-driven discovery.

More on Om's screening technology and LULA: https://www.omtx.ai/technology