123 validated targets across DUD-E and DEKOIS2. IsoScreen universal scorer — zero per-target parameters. 0.912 mean AUC on property-matched decoys. No black-box AI. Every weight traces to a physics term.
*AMR docking, ~100 lig/sec on RTX-series GPUs. Scoring: ~10–15K lig/sec. ? DEKOIS2 mean AUC, 76 targets, property-matched decoys.
A complete molecular docking platform built on physics first principles. Target-specific calibration of interpretable energy terms. No external dependencies.
Zero per-target parameters. 11 physics archetypes with strain-adaptive descriptor blending. Works on any target without calibration. AMR docking recommended.
Rank compounds by binding feasibility without docking. 8 molecular descriptors, instant results. Screen first, dock the survivors.
GPU-accelerated pose search with AMR, GA, and FFT engines. AMR recommended for IsoScreen compatibility. ~100 lig/sec on RTX-series GPUs.
Automated receptor preparation. Clean PDB, assign protonation, define binding site, generate lattice grid. Ready to dock in minutes.
GPU-accelerated rescoring of pre-docked poses. 14 calibrated HBQ energy terms. Per-target profiles with cross-validated AUC. Batch scoring for hit lists.
With labeled active/decoy data, the optimizer finds the right physics feature weights for your target. Automated cross-validation and AUC reporting.
Validated on two independent benchmarks with property-matched decoys. Zero per-target parameters.
Why two benchmarks? DUD-E is the industry standard â every tool reports on it. DEKOIS2 uses property-matched decoys that defeat statistical shortcuts, testing whether your scorer understands real binding physics. IsoScreen is the only tool that excels on both.
Per-target calibrated profiles. All AUCs are holdout-validated.
| Target | Family | Engine | Method | Validated AUC |
|---|---|---|---|---|
| ACE | Metalloprotease | FFT | E2E | 0.983 |
| HMGR | Reductase | FFT | Rescore | 0.980 |
| HIVRT | RT | FFT | Rescore | 0.944 |
| NA | Glycosidase | FFT | Rescore | 0.942 |
| PPARG | Nuclear Receptor | GA | Rescore | 0.934 |
| EGFR | Kinase | FFT | E2E | 0.933 |
| P38A | Kinase | AMR | E2E | 0.894 |
| ACHE | Enzyme | FFT | E2E | 0.880 |
| UROK | Serine Protease | GA | E2E | 0.872 |
| MDM2 | PPI | FFT | E2E | 0.855 |
| COMT | Transferase | FFT | E2E | 0.840 |
| ADRB2 | GPCR | AMR | Rescore | 0.834 |
From first-principles physics to GPU-accelerated scoring in four steps.
Scoring grounded in calibrated sterics and electrostatics. No parameters trained on binding data. Every weight traces to a physical interaction term.
14 calibrated energy terms: dispersion, repulsion, Coulomb, H-bonds, pi-stacking, burial penalties, and more.
123 validated targets with calibrated profiles, plus IsoScreen universal scoring for any target with zero setup.
GPU-accelerated scoring via WebGPU compute shaders. Score thousands of molecules per second on RTX-series GPUs, directly in the browser.
Zero per-target parameters. Docking strain and physics terms determine the scoring pathway automatically. Works on any target without calibration. 0.91 mean AUC on DEKOIS2.
Common docking failures addressed by physics-first design.
Physics-derived scoring grounded in calibrated sterics and electrostatics. No parameters trained on binding data. No decoy bias.
Docking-calibrated sR softening derived from potential matching, not AUC optimization. Reduces clash artifacts while preserving pose discrimination.
Build your own per-target profile with labeled active/decoy data. The optimizer finds the right physics feature weights for your target with automated cross-validation.
Most ML scoring functions need active/decoy labels for every new target. IsoScreen scores any target with zero calibration — docking strain and physics terms determine the scoring pathway automatically.
Full platform access during open beta. No credit card required.
LatticeZero