Pauling runs a full end-to-end pipeline -- from PDB download through receptor prep, pocket detection, docking at scale, ADMET profiling, quality control, and molecular dynamics -- all through a single conversational interface on cloud infrastructure. Potato is building AI tools for specific comp chem tasks; the end-to-end pipeline from structure to validated candidates isn't there.
| Category | Pauling | Potato |
|---|---|---|
| End-to-end pipeline | Full: structure prep, pocket detection, docking, MD, ADMET, QC -- all integrated | Focused AI tools for specific comp chem tasks |
| Docking | UniDock (GPU-accelerated Vina) -- 10M+ compound virtual screens | Docking capabilities in development |
| Molecular dynamics | Full GROMACS MD with automatic ligand parameterization (ACPYPE), checkpoint/resume, MM-PBSA | Limited MD offering |
| ADMET profiling | Integrated into pipeline -- run ADMET alongside docking and MD results | ADMET predictions available |
| Pocket detection | P2Rank integrated -- automatic binding site identification from PDB structures | Binding site tools available |
| Quality control | MolProbity (structure validation) + PoseBusters (pose QC) -- every result is auditable | Limited QC tooling |
| Cloud scaling | Google Cloud Dataflow with Apache Beam -- auto-scales to hundreds of workers | Cloud infrastructure available |
| File format handling | Automatic: PDB, PDBQT, SDF, MOL2, CIF, SMILES -- all conversions handled internally | Format support varies by tool |
| Interface | Conversational AI -- describe experiments in natural language, get results back | AI-assisted interface |
| Jobs & tracking | Full dashboard: every parameter, input file, output file tracked per job | Job tracking available |
| Private molecule catalogs | Owner-scoped catalogs (owner_id) with data isolation | Data management varies |
| Cost | Pay-per-use cloud compute, no license fees | Pricing model varies |