Knowledge-driven multimodal mutual learning for cell line-targeted anticancer peptide prediction.

TargetPC achieves over 20% higher accuracy than existing methods in cell line-targeted anticancer peptide prediction, enabling precise tumor-specific drug discovery.

• Why it matters: Accurate prediction of tumor cell-specific anticancer peptides is crucial for developing targeted therapies with minimal side effects, yet current methods struggle with limited annotated data.
• What they did: Researchers developed TargetPC, a deep learning framework that integrates multimodal representations from peptide and cell line data using pretrained models and mutual learning, trained on a hybrid dataset from five peptide databases.
• The result: TargetPC outperforms state-of-the-art models, demonstrates strong generalization to unseen peptides and cell lines, and successfully identifies novel peptides targeting breast cancer cells, facilitating efficient anticancer drug development.