Structural characterization of length-varying peptide sequences for peptide quantitative structure-activity...

Area of Science:

• Computational chemistry
• Bioinformatics
• Peptidology

Background:

• Peptide quantitative structure-activity relationship (pQSAR) models predict bioactive peptide function.
• Amino acid descriptors (AADs) characterize peptides but create inconsistent data for length-varying peptide sequences (LVPSs).
• Existing methods like auto-cross covariance (ACC) are limited for LVPSs in pQSAR.

Purpose of the Study:

• To introduce and validate a novel multivariate method, Residue Descriptor-Distance Vector (RDDV), for pQSAR analysis.
• To overcome the limitations of length-dependent descriptors in pQSAR for LVPSs.
• To provide a consistent descriptor representation for LVPSs in pQSAR modeling.

Main Methods:

• Developed the Residue Descriptor-Distance Vector (RDDV) method based on inter-residue pseudo-interaction potentials.
• Characterized peptide sequences using a fixed number of descriptor parameters regardless of length.
• Validated RDDV using an in-house pQSAR dataset with various AADs and regression tools.
• Compared RDDV performance against the traditional auto-cross covariance (ACC) method.

Main Results:

• RDDV successfully generated an invariable number of descriptor parameters for LVPSs.
• The RDDV method demonstrated effectiveness in pQSAR analysis within the tested dataset.
• Comparative analysis showed RDDV's potential as an alternative to ACC for specific pQSAR applications.
• Systematic exploration revealed optimal combinations of AADs and regression tools for RDDV.

Conclusions:

• RDDV is a promising second-generation multivariate method for AAD-based pQSAR.
• The method effectively addresses the challenge of descriptor inconsistency in LVPSs.
• RDDV offers a valuable new approach for computational peptidology and drug discovery.