onestep-sepal-ppi

Inference-only one-step CLI for SEPAL-PPI.

The PyPI package name is onestep-sepal-ppi. After installation, the command name is sepal-ppi.

Installation

Python 3.12 is recommended:

conda create -n sepal python=3.12 -y
conda activate sepal
pip install torch==2.4.1 torchvision==0.19.1 torchaudio==2.4.1 --index-url https://download.pytorch.org/whl/cu124
pip install --no-build-isolation flash-attn
pip install esm-efficient
pip install onestep-sepal-ppi

Quick Start

1. Initialize the configuration:

sepal-ppi setup --ini-path /path/to/nsp3_env_config.ini

2. Prepare features from a FASTA file:

sepal-ppi prepare -f /path/to/input.fasta --output-dir /path/to/mutifeature

3. Run prediction:

sepal-ppi predict \
  --esm-precision bf16 \
  --esm-model esm2_15b \
  --length 5000 \
  --predict-batch 12800 \
  --emb-batch 1 \
  --multifeature-dir /path/to/mutifeature \
  --interaction-list /path/to/ll.csv \
  --fasta /path/to/input.fasta \
  --output-dir /path/to/output

PDB Mode

sepal-ppi prepare -p /path/to/pdb_dir --output-dir /path/to/mutifeature

PDB mode requires dssp or mkdssp to be available in the environment.