Active Semisupervised Model for Improving the Identification of Anticancer Peptides.

Area of Science:

• Biotechnology
• Computational Biology
• Oncology

Background:

• Cancer poses a significant global health threat, necessitating the development of effective anticancer agents.
• Accurate identification of anticancer peptides (ACPs) is crucial for designing novel cancer therapeutics.
• Existing machine learning algorithms for ACP identification often struggle with limited labeled data, impacting their accuracy.

Purpose of the Study:

• To develop a novel computational method, ACP-ALPM, integrating active learning (AL) and label propagation (LP) for enhanced ACP identification.
• To address the limitations of current machine learning models in ACP detection, particularly concerning data scarcity.

Main Methods:

• Developed an efficient feature representation for peptide sequences using descriptor and coding information.
• Employed an active learning (AL) strategy to select informative data for model training.
• Utilized label propagation (LP) iteratively to build a robust classifier.

Main Results:

• The proposed ACP-ALPM method demonstrated significantly superior performance compared to state-of-the-art and classic methods.
• Experimental comparisons confirmed the effectiveness of the AL strategy over random selection across three public datasets.
• Visualization experiments validated that AL enhances model performance by effectively utilizing unlabeled data.

Conclusions:

• ACP-ALPM offers a powerful and data-efficient approach for identifying anticancer peptides.
• The integration of AL and LP provides a promising framework for improving machine learning model performance in bioinformatics.
• This method has the potential for broader application in peptide identification and related research areas.