| license: mit | |
| tags: | |
| - protein-ligand | |
| - binding-affinity | |
| - dsmbind | |
| library_name: pytorch | |
| # DSMBind_modal | |
| Modal-ready checkpoint repository for DSMBind. | |
| ## Credit | |
| This repository redistributes DSMBind checkpoints from the DSMBind project: | |
| - Jin et al., *DSMBind: SE(3) denoising score matching for unsupervised binding energy prediction and nanobody design*, bioRxiv 2023 | |
| - Jin et al., *Unsupervised protein-ligand binding energy prediction with Neural Euler's Rotation Equations*, NeurIPS 2023 | |
| ## Included checkpoints | |
| - `model.drug.allatom` (protein-ligand all-atom scoring) | |
| - `model.antibody.allatom` (antibody-antigen) | |
| - `model.skempi.allatom` (protein-protein) | |
| - `cross.ckpt` | |
| - `dock.cdr1.ckpt`, `dock.cdr2.ckpt`, `dock.cdr3.ckpt`, `dock.sc.ckpt` | |
| - `PDL1.ckpt`, `aaratio.ckpt` | |
| ## Score semantics | |
| For DSMBind inference, **higher predicted DSMBind score means better predicted binding**. | |