File size: 2,492 Bytes
7efee70 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 |
# EntangledSBM for Transition Path Sampling ⚛️
This part of the code is for running the transition path sampling (TPS) experiments. We demonstrate that EntanlgedSBM generates physically plausible transition paths for fast-folding proteins at an **all-atom resolution**. This codebase is partially built off of the [TPS-DPS repo](https://github.com/kiyoung98/tps-dps/tree/main).
## Environment Installation
```
conda env create -f environment.yml
conda activate entangled-tps
```
## Running Experiments
All training scripts are located in `entangled-tps/scripts/`. Each script is pre-configured for a specific experiment.
**Before running experiments:**
1. Set `HOME_LOC` to the base path where EntangledSBM is located and `ENV_PATH` to the directory where your environment is downloaded in the `.bash` files in `scripts/`
2. Replace `/path/to/your/home` to the base path where EntangledSBM is located in the `.slurm` files in `scripts/`
3. Create a path `entangled-tps/results` where the simulated trajectory figures and metrics will be saved. Also, create `entangled-tps/logs` where the training logs will be saved.
4. Activate the conda environment:
```
conda activate entangled-tps
```
5. Login to wandb using `wandb login`
**Run experiment using `nohup` with the following commands:**
```
cd entangled-tps/scripts
chmod aldp.sh
nohup ./aldp.sh > aldp.log 2>&1 &
```
**Run experiment using `slurm` with the following commands:**
```
cd entangled-tps/scripts
sbatch aldp.slurm
```
Evaluation will run automatically after each sampling step and logged with wandb.
We report the following metrics in our paper:
1. Root-Mean Squared Distance **(RMSD)** of the Kabsch-aligned coordinates averaged across 64 paths.
2. Percentage of simulated trajectories that hit the target state **(THP)**
3. Highest energy transition state along the biased trajectories averaged across the trajectories that hit the target **(ETS; kJ/mol)**
## Overview of Outputs
**Training outputs are saved to experiment-specific directories:**
```
entangled-tps/results/
├── aldp_ce/
│ └── positions/ # Generated trajectory
│ └── paths.png # Figure of simulated trajectories
│ └── policy.pt # Model checkpoint
```
**Training logs are saved in:**
```
entangled-tps/logs/
├── <DATE>_entangled-aldp-ce.log
├── <DATE>_entangled-chignolin-ce.log
├── <DATE>_entangled-trpcage-ce.log
└── <DATE>_entangled-bba-ce.log
``` |