K-Pax

K-Pax applies a Bayesian model-based approach to classify aligned protein sequences into functionally divergent subgroups and identify subgroup-specific residues.

Key Features:

• Automated Group Identification: Infers the number of evolutionary groups directly from aligned sequence data without pre-specification.
• Model-Based Approach: Uses a Bayesian model to distinguish conserved sequence regions relevant for clustering from noisy regions.
• Stochastic Optimization Algorithm: Implements a fast stochastic optimization algorithm to maximize posterior probability for clustering.
• High Specificity and Sensitivity: Demonstrates high specificity and sensitivity in simulated and real-world clustering tasks.

Scientific Applications:

• Evolutionary Analysis: Identifies evolutionarily related subgroups to elucidate relationships among proteins.
• Functional Annotation: Highlights residues associated with subgroup-specific functions and residues near active sites to inform annotation.
• Data-Driven Insights: Provides objective subgrouping from sequence data without arbitrary group specification to support reproducible analyses.

Methodology:

K-Pax employs a Bayesian framework on aligned sequence data, models conserved versus noisy regions, automatically infers subgroup numbers, and uses a fast stochastic optimization algorithm to maximize the estimated posterior probability.