Singular value decomposition of protein sequences as a method to visualize sequence and residue space

Area of Science:

• Bioinformatics
• Computational Biology
• Structural Biology

Background:

• Singular Value Decomposition (SVD) is a powerful mathematical tool for analyzing multiple sequence alignments (MSAs).
• Understanding SVD is crucial for extracting meaningful biological insights from sequence data, including conservation and covariance.
• Many protein scientists find the mathematical underpinnings of SVD challenging to access.

Purpose of the Study:

• To provide an intuitive and comprehensive explanation of SVD analysis for MSAs.
• To demonstrate how SVD can identify sequence subgroups and their defining features.
• To correlate identified sequence clusters with protein structure, function, stability, and taxonomy.

Main Methods:

• Described the underlying mathematics of SVD in an accessible manner.
• Applied SVD to analyze sequences from a controlled model and two protein families (homeodomain and Ras superfamilies).
• Utilized k-means clustering to group sequences and identify distinguishing residues.

Main Results:

• SVD analysis successfully identified sequence clustering within the homeodomain and Ras superfamilies.
• Distinct sequence clusters were correlated with known taxonomic and functional groups.
• Python scripts were developed for SVD analysis, visualization, and cluster identification.

Conclusions:

• SVD is an effective method for subgroup identification and feature extraction in MSAs.
• The approach links sequence characteristics to biological properties like function and taxonomy.
• Freely available Python scripts facilitate the application of SVD in protein sequence analysis.