SAR-Atlas: Drug Design, Body Distribution Modeling, and Candidate Screening in Rust
SAR-Atlas (v3.8.0) is an open-source medicinal chemistry and pharmacokinetics workbench built to help drug-discovery researchers and students evaluate whether promising candidate compounds carry the balanced physical properties, oral absorption, brain penetration, and binding stability needed to succeed as real medicines. Many small molecules bind strongly to a purified target in test tubes yet fail later in development because they dissolve poorly in gut fluids, clear too quickly through the liver, or are expelled at the protective blood-brain barrier. SAR-Atlas integrates four complementary screening pipelines inside one offline desktop application: multiparameter drug-likeness scoring (CNS MPO and QED) alongside matched-molecular-pair activity-cliff detection, 24-hour oral blood-and-brain physiologically based pharmacokinetic (PBPK) modeling, maximum-likelihood relative binding free energy (RBFE) thermodynamic cycle closure, and pool-based Bayesian active learning. Evaluated across 26 CNS and peripheral reference drugs, SAR-Atlas identified 11 balanced Pareto leads and 5 sharp activity cliffs; in 24-hour oral dosing it separated brain-entering Haloperidol (Kp,uu = 0.923, 82.41% receptor occupancy) from excluded Lapatinib (Kp,uu = 0.062, 0.15% occupancy); and it reconciled 6-cycle Dopamine D2 thermodynamic loop errors down to 0.000 kcal/mol. CRV Science Apps Hub: https://crvscience.com/appsCanonical Preprint URL: https://crvscience.com/apps/preprints/sar-atlas-multiparameter-pbpk-rbfeMaster Concept DOI: 10.5281/zenodo.24222390Repository: https://github.com/bpwhite/crvscience-sar-atlas
Authors
- Bryan White (ORCID: https://orcid.org/0000-0003-4191-3511)
Institutions
- CRV (Netherlands) (NL)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-14
- DOI
- https://doi.org/10.5281/zenodo.24222390
- Primary Topic
- Computational Drug Discovery Methods
- Type
- preprint