ACP-ML: A sequence-based method for anticancer peptide prediction.

Area of Science:

• Biochemistry and Molecular Biology
• Computational Biology and Bioinformatics
• Oncology

Background:

• Cancer remains a significant health challenge, with chemotherapy causing severe side effects.
• Anti-cancer peptides (ACPs) offer targeted cancer cell destruction, driving new drug development.
• Experimental screening of ACPs is costly and inefficient due to protein complexity.

Purpose of the Study:

• To develop an effective computational model, ACP-ML, for predicting anti-cancer peptides.
• To identify optimal feature extraction and selection methods for ACP prediction.
• To evaluate the performance of ensemble learning models for ACP identification.

Main Methods:

• Utilized multiple feature descriptors (DPC, PseAAC, CTDC, CTDT, CS-Pse-PSSM) for ACP prediction.
• Employed a two-step feature selection process (MRMD, RFE) to identify key features.
• Developed and validated a voting-based ensemble learning model (ACP-ML) using cross-validation and independent test sets.

Main Results:

• The ACP-ML model achieved high prediction accuracy on independent test sets (90.891% and 92.578%).
• The proposed feature processing and ensemble methods demonstrated superior effectiveness compared to existing algorithms.
• The ACP-ML model exhibited strong generalization capability and enhanced accuracy in predicting ACPs.

Conclusions:

• ACP-ML provides a highly accurate and efficient computational tool for identifying anti-cancer peptides.
• The study highlights the potential of ensemble learning and advanced feature selection in drug discovery.
• This approach facilitates the development of novel, targeted anti-cancer therapeutics.