Using a Classifier Fusion Strategy to Identify Anti-angiogenic Peptides

Area of Science:

• Biochemistry
• Computational Biology
• Bioinformatics

Background:

• Anti-angiogenic peptides play vital roles in physiological functions and are potential therapeutic agents for angiogenesis-related diseases.
• Accurate identification of these peptides is essential for understanding angiogenic homeostasis and developing effective antineoplastic therapies.

Purpose of the Study:

• To develop and validate an accurate ensemble predictor for identifying anti-angiogenic peptides.
• To investigate the effectiveness of different feature spaces and selection methods for improving prediction performance.

Main Methods:

• An ensemble predictor was created by combining classifiers with optimal sensitivity and specificity.
• Bi-profile Bayes (BpB) features were utilized, and feature selection was performed using the Relief algorithm and Incremental Feature Selection (IFS).
• The performance of individual and ensemble classifiers was evaluated using accuracy and Matthew's Correlation Coefficient (MCC).

Main Results:

• The ensemble classifier trained with Bi-profile Bayes (BpB) features achieved an accuracy of 0.822 and an MCC of 0.649.
• Using discriminative features identified by Relief and IFS, the ensemble classifier reached a sensitivity of 0.776, specificity of 0.888, accuracy of 0.832, and MCC of 0.668.
• The proposed prediction method demonstrated superior performance compared to previous studies.

Conclusions:

• The developed ensemble predictor, particularly when utilizing discriminative features, offers a highly accurate and reliable method for anti-angiogenic peptide prediction.
• This advancement holds significant potential for improving the understanding of angiogenesis and the development of novel cancer treatments.