Feature Selection Combined with Neural Network Structure Optimization for HIV-1 Protease Cleavage Site Prediction

Area of Science:

• Biochemistry
• Computational Biology
• Drug Discovery

Background:

• Understanding the specificity of Human Immunodeficiency Virus type 1 (HIV-1) protease is essential for developing effective protease inhibitors.
• Identifying key determinants of HIV-1 protease cleavage site specificity can significantly improve drug design strategies.

Purpose of the Study:

• To develop and evaluate a novel feature selection method combined with neural network structure optimization for analyzing HIV-1 protease specificity.
• To identify critical positions within octapeptides that dictate HIV-1 protease cleavage.
• To assess the efficacy of different feature fusion techniques for enhancing prediction performance.

Main Methods:

• Utilized two types of novel features from the Amino Acid Index database alongside traditional orthogonal encoding features.
• Incorporated physiochemical and sequence information for comprehensive analysis.
• Employed combination fusion and decision fusion methods for feature representation and prediction improvement.
• Applied feature selection combined with decision fusion for HIV-1 protease cleavage site prediction.

Main Results:

• Feature selection identified p2, p1, p1', and p2' as the most important positions in octapeptides for HIV-1 protease cleavage.
• Decision fusion of selected feature subsets achieved excellent prediction performance.
• The proposed method demonstrates effectiveness and utility for HIV-1 protease cleavage site prediction.

Conclusions:

• Feature selection combined with decision fusion is a highly effective strategy for predicting HIV-1 protease cleavage sites.
• The identified important positions provide valuable insights for the rational design of HIV-1 protease inhibitors.
• This approach offers a promising direction for future HIV-1 protease inhibitor development.