MLACP 2.0: An updated machine learning tool for anticancer peptide prediction.

Area of Science:

• Immunoinformatics
• Computational Biology
• Drug Discovery

Background:

• Anticancer peptides (ACPs) are a promising class of anticancer drugs with fewer side effects than traditional therapies.
• Predicting ACPs from sequence data is a critical challenge in immunoinformatics.
• Previous machine learning methods, like MLACP, have impacted anticancer research but require improved robustness.

Purpose of the Study:

• To develop a more robust and accurate prediction tool for anticancer peptides.
• To enhance the identification of novel ACPs for therapeutic development.

Main Methods:

• Construction of the first large, non-redundant training and independent datasets for ACP research.
• Exploration of various feature encodings and conventional classifiers.
• Development of MLACP 2.0 by concatenating and training selected models through a convolutional neural network (CNN).

Main Results:

• MLACP 2.0 demonstrated excellent performance on diverse independent datasets, significantly outperforming recent ACP prediction tools.
• The tool showed superior performance in cross-validation and independent assessments compared to CNN-based embedding and conventional single models.
• MLACP 2.0 is freely available at https://balalab-skku.org/mlacp2.

Conclusions:

• MLACP 2.0 represents a significant advancement in predicting anticancer peptide activity.
• The tool is expected to facilitate experimental design and accelerate the discovery of novel ACPs.
• Enhanced prediction accuracy will aid in the development of more effective cancer therapies.