Structure-based classification of FAD binding sites: A comparative study of structural alignment tools

Area of Science:

• Structural bioinformatics
• Computational biology
• Protein structure analysis

Background:

• Flavin adenine dinucleotide (FAD) is a crucial cofactor in numerous biological processes.
• Accurate identification and classification of FAD binding sites are essential for understanding protein function and designing new therapeutics.
• Existing structural alignment tools vary in their performance for analyzing protein-ligand interactions.

Purpose of the Study:

• To assess the performance of six structural alignment tools in discriminating FAD from non-FAD binding sites.
• To evaluate the ability of these tools to classify a large set of FAD binding sites through all-to-all comparisons.
• To characterize the structural diversity within FAD binding sites based on alignment and clustering results.

Main Methods:

• Six structural alignment tools (TM-Align, TriangleMatch, CLICK, ProBis, SiteEngine, GA-SI) were employed.
• Methods were tested for discriminating FAD vs. non-FAD binding sites using Matthews correlation coefficient.
• An all-to-all comparison of 883 FAD binding sites was performed, followed by various clustering techniques (Single-linkage, UPGMA, Complete-linkage, SPICKER, k-Means).

Main Results:

• All tested alignment methods successfully distinguished FAD from non-FAD binding sites with high accuracy.
• Significant differences were observed among the methods in classifying FAD binding sites, indicated by poor correlation in results.
• GA-SI produced distinct alignments compared to other methods.
• Clustering analysis revealed unique structural features and patterns characterizing different groups of FAD binding sites.

Conclusions:

• Structural alignment tools can reliably differentiate FAD binding sites, but their performance varies for classification tasks.
• The study highlights the structural diversity of FAD binding sites, with distinct clusters identified.
• The findings provide insights into selecting appropriate tools for analyzing protein-ligand interactions and understanding binding site variability.