EnsemPred-ACP: Combining machine and deep learning to improve anticancer peptide prediction.

Area of Science:

• Computational biology
• Bioinformatics
• Drug discovery

Background:

• Anticancer peptides (ACPs) show promise for cancer therapy due to selective cancer cell targeting.
• Accurate computational prediction of ACPs is challenging due to complex cancer mechanisms.

Purpose of the Study:

• To introduce EnsemPred-ACP, an ensemble framework combining machine learning (ML) and deep learning (DL) for enhanced ACP prediction.
• To improve the identification of potential therapeutic peptides for cancer treatment.

Main Methods:

• Developed EnsemPred-ACP, an ensemble framework integrating ML and DL.
• Introduced binary profile features (BPF) to augment pre-trained protein embeddings.
• Employed a dual-pipeline architecture processing sequence features, embeddings, and BPF-enhanced embeddings.

Main Results:

• EnsemPred-ACP achieved high prediction performance: 0.863 accuracy, 0.897 sensitivity, and 0.830 specificity.
• The model demonstrated strong generalization with an AUC of 0.93.
• BPF significantly improved prediction accuracy by 2.5% (ESM2) and 11.1% (ProtT5).

Conclusions:

• EnsemPred-ACP effectively identifies potential therapeutic peptides, outperforming existing methods.
• The integration of BPF with protein embeddings enhances ACP prediction accuracy.
• This approach contributes to advancing peptide-based cancer therapeutics.