ECOD domain classification of 48 whole proteomes from AlphaFold Structure Database using DPAM2

Area of Science:

• Structural biology
• Bioinformatics
• Computational biology

Background:

• Protein structure prediction models like AlphaFold generate vast datasets.
• Large-scale domain annotation is crucial for extracting biological insights from these predictions.
• The Evolutionary Classification of Protein Domains (ECOD) provides a framework for domain classification.

Purpose of the Study:

• To detect and catalog protein domains from 48 whole proteomes in the AlphaFold Database.
• To analyze the distribution and characteristics of protein domains across different organisms (eukaryotes and bacteria).
• To compare domain populations from structure predictions with existing databases and identify enriched groups.

Main Methods:

• Utilized the Evolutionary Classification of Protein Domains (ECOD) classification system.
• Applied domain detection and cataloging to 48 whole proteomes from the AlphaFold Database.
• Analyzed residue coverage, domain populations, and homologous group distributions.

Main Results:

• Successfully classified domains or identifiable regions for 90% of residues across all proteomes.
• Cataloged 746,349 domains from 536,808 proteins, totaling over 226 million amino acid residues.
• Identified distinct domain patterns in eukaryotes (higher disordered regions, repeats) versus bacteria, and enumerated shared domains like Rossmann and TIM barrels.

Conclusions:

• Large-scale domain annotation of predicted protein structures is feasible and yields significant biological insights.
• Comparative analysis reveals evolutionary differences in protein domain composition between eukaryotes and bacteria.
• The findings inform strategies for future protein classification and target selection for large protein sets.