SamPler

SamPler optimizes parameter selection for gene functional annotation to improve the accuracy and reliability of genome-scale metabolic annotations and model reconstruction within the merlin framework.

Key Features:

• Semi-automated approach: Integrates manual curation with automated evaluation to balance curator input and algorithmic assessment.
• Parameter optimization: Performs prior analysis of annotation algorithm parameters by evaluating automatic annotations across all possible parameter combinations.
• Sample-based curation: Uses a randomly selected subset of genes/proteins that are manually curated to serve as a reference for parameter evaluation.
• Confusion matrix analysis: Generates confusion matrices for each parameter set and derives performance metrics including accuracy, precision, and negative predictive value.
• Integration with merlin: Implemented as a plugin for merlin to operate on genome-scale metabolic annotation and model reconstruction.
• Empirical demonstration: Applied to four different genome annotations within merlin.

Scientific Applications:

• Rapid annotation of newly sequenced genomes: Increases confidence in gene function assignments for newly assembled genomes by optimizing annotation parameters.
• Large-scale and diverse-species projects: Enables systematic parameter tuning when manual curation alone is impractical across many genomes or taxa.
• Genome-scale metabolic model improvement: Enhances the quality of annotations used for metabolic model reconstruction within merlin.

Methodology:

Manually curate a randomly selected subset of genes/proteins, evaluate automatic annotations across all parameter combinations, generate confusion matrices per parameter set, compute metrics such as accuracy, precision and negative predictive value, and select optimal parameter values; implemented as a merlin plugin for genome-scale metabolic annotation and model reconstruction.