De novo design of bitter petides workflow

A computational workflow for generating artificial peptides using ZymCTRL and predicting their bitterness with BitterPep-GCN. Includes a KNIME workflow for sequence preparation and physicochemical analysis of the generated peptides.

Requirements

• KNIME Analytics Platform
• Python ≥ 3.8
• HuggingFace Transformers
• PyTorch
• A CUDA-compatible GPU is recommended for sequence generation

Usage

Step 1 — Prepare peptide sequences

Use the first node of the included KNIME workflow to convert your peptide sequences into FASTA format suitable for ZymCTRL input.

Step 2 — Generate peptides with ZymCTRL

Follow the instructions on the ZymCTRL HuggingFace model page to generate novel enzyme sequences. The model supports both zero-shot generation (given an EC number) and fine-tuning on custom sequence sets.

Step 3 — Predict bitterness with BitterPep-GCN

Use BitterPep-GCN to predict the bitterness of the generated peptides.

Step 4 — Analyse physicochemical properties

Use the KNIME workflow to compute physicochemical descriptors for the generated and predicted peptides.

Creator

Alexandra Steuer

References

Srivastava, P.*, Steuer, A.*, Ferri, F., Nicoli, A., Schultz, K., Bej, S., Di Pizio, A.†, & Wolkenhauer, O.† (2024). Bitter peptide prediction using graph neural networks. Journal of Cheminformatics, 16, 111. https://doi.org/10.1186/s13321-024-00909-x

Munsamy, G., Illanes-Vicioso, R., Funcillo, S., Nakou, I. T., Lindner, S., Ayres, G., Sheehan, L. S., Moss, S., Eckhard, U., Lorenz, P., & Ferruz, N. (2024). Conditional language models enable the efficient design of proficient enzymes. bioRxiv. https://doi.org/10.1101/2024.05.03.592223

License

This project is licensed under the MIT License.