https://github.com/keiserlab/chemprobe
https://github.com/keiserlab/chemprobe
Last synced: 6 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/keiserlab/chemprobe
- Owner: keiserlab
- License: mit
- Created: 2022-11-30T21:03:34.000Z (over 2 years ago)
- Default Branch: main
- Last Pushed: 2024-08-28T00:03:22.000Z (8 months ago)
- Last Synced: 2024-11-01T00:08:57.898Z (6 months ago)
- Language: Python
- Size: 215 KB
- Stars: 1
- Watchers: 2
- Forks: 0
- Open Issues: 3
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# chemprobe
[](https://zenodo.org/doi/10.5281/zenodo.13381833)
## installation
Requires `python<=3.11`
To access model & dataset modules:
```
pip install chemprobe
```
To access scripts for preprocessing/training/inference you must install from source:
```
git clone https://github.com/keiserlab/chemprobe.git
cd chemprobe
# use a virutal env...
# conda activate env
pip install .
```
## download data
Required to run inference
```
bash download_data.sh
```
## preprocess
Required to run inference
```
python preprocess.py \
--data_path ../data
```
## train
```
python train.py \
--name TEST \
--exp film \
--fold 0 \
--n_blocks 4 \
--data_path ../data/preprocessed \
--batch_size 16384 \
--gpus 0,1,2,3 \
--num_workers 4 \
--lr 1e-3
```
permuted label model
```
python train.py \
--study_path /scratch/wconnell/danger/chemprobe/optuna/exp=film/fold=0/ \
--data_path ../data/preprocessed \
--name perm-fold=0 \
--exp film \
--fold 0 \
--max_epochs 5 \
--batch_size 16384 \
--gpus 3, \
--permute_labels
```
## optimize
```
python optimize.py \
--study_path /srv/danger/scratch/wconnell/chemprobe/optuna/ \
--data_path ../data/preprocessed \
--exp film \
--fold 0 \
--n_trials 20 \
--prune \
--batch_size 16384 \
--gpus 1,
```
## predict
```
python predict.py \
--cpds neratinib \ # Do not specify to run on all compounds
--data_path ../data/your_data/ \
--attribute \ # optional to run attribution
--batch_size 128 \
--gpus 2,
```