Pauling puts the full screening pipeline in your hands: screen millions of compounds yourself, iterate in real time, and validate hits with MD and ADMET without waiting on anyone. Atomwise pioneered deep-learning virtual screening (AtomNet) with strong pharma partnerships, but it's a service -- you send them targets, they send back hits.
| Category | Pauling | Atomwise |
|---|---|---|
| Access model | Self-serve platform -- run jobs on demand | Service model -- submit targets, wait for results |
| Screening method | Physics-based docking (UniDock/Vina) with explicit scoring functions | Deep learning (AtomNet CNN) trained on co-crystal structures |
| Screening scale | 10M+ compounds via Cloud Dataflow auto-scaling | Large library (proprietary), but you don't control scope |
| Turnaround time | Minutes to hours depending on library size | Weeks to months per engagement |
| Iteration speed | Modify parameters and re-run immediately | New round requires new service request |
| MD validation | Full GROMACS MD with MM-PBSA binding energy | Not included -- docking scores only |
| ADMET profiling | Integrated ADMET pipeline on every hit | Not part of standard deliverable |
| Pose quality control | PoseBusters + MolProbity validation on all poses | |
| Data ownership | Full ownership -- private catalogs, all files yours | Atomwise retains rights to screen data |
| Transparency | Every parameter, input, and output logged in jobs dashboard | Black-box predictions -- no scoring function visibility |
| Hit rate enrichment | Explicit Vina scoring with known physics | Strong enrichment from AtomNet in benchmarks |
| Pharma partnerships | Early-stage -- focused on computational chemists | Extensive pharma collaborations (Bayer, Merck, etc.) |