ApexSP

ApexSP is a signal peptide design and screening workflow for cargo-specific protein secretion design.

This repository contains the inference scripts and pretrained weights used by the ApexSP pipeline.

Contents

tuili_123/
  ApexSP_tuili.py   # End-to-end inference workflow
  tuili_1.py        # Sec/Tat signal peptide generation
  tuili_2.py        # Multitask biological filter
  tuili_3.py        # Signal peptide-cargo compatibility scoring

pth_123/
  model1.pth        # Generation model weights
  model2.pth        # Multitask biological filter weights
  model3.pth        # Compatibility/ranking model weights
  pth_123.sha256    # SHA256 checksums

Check Weights

After downloading the repository, verify the weights with:

cd pth_123
sha256sum -c pth_123.sha256

On macOS, use:

cd pth_123
shasum -a 256 model1.pth model2.pth model3.pth
cat pth_123.sha256

Notes

The inference scripts are preserved from the original server workflow. Some paths in the scripts point to the original runtime environment under /home/zkyd/data1/ycy/ApexSP/. If running outside that server, update the local model paths, ESM2 path, and conda environment path before inference.