Gclust

Gclust performs trans-kingdom protein sequence clustering to identify homologous protein groups across eukaryotic and prokaryotic genomes while accounting for transit peptides in organellar proteins.

Key Features:

• Heuristic-Based Clustering: Selects similar protein sequences by determining appropriate similarity thresholds for homolog detection in large-scale, divergent datasets.
• All-Against-All BLASTP Analysis: Uses all-against-all BLASTP results and single-linkage clustering to construct minimal homolog groups.
• Domain Structure Estimation: Estimates protein domain structure to inform clustering decisions.
• Exclusion of Multi-Domain Proteins: Excludes multi-domain proteins from clustering to focus on simpler homologous relationships.
• Consideration of Transit Peptides: Accounts for transit peptides in organellar proteins during clustering.
• Entropy-Optimized Organism Count Method: Applies an entropy-optimized organism count heuristic to estimate similarity thresholds for homologs.
• Evaluation Against Power Law: Evaluates resultant protein clusters using power law metrics to assess cluster quality and distribution.

Scientific Applications:

• Trans-kingdom protein clustering: Constructs protein clusters across up to 95 organisms for trans-kingdom comparisons.
• Comparative genomics and evolutionary analysis: Identifies homologous proteins between eukaryotes and prokaryotes to support evolutionary studies.
• Functional annotation: Groups homologous proteins to aid inference of protein function and underlying molecular mechanisms.

Methodology:

Performs all-against-all BLASTP, applies a heuristic including an entropy-optimized organism count to set similarity thresholds, constructs minimal homolog groups by single-linkage clustering, estimates domain structure, excludes multi-domain proteins, accounts for transit peptides in organellar proteins, and evaluates clusters using power law metrics.