doepipeline

doepipeline optimizes bioinformatics processing pipelines using statistical Design of Experiments (DoE) methodologies to select and tune software parameters that affect analytical outcomes.

Key Features:

• Design of Experiments (DoE): Applies statistical DoE principles for systematic and efficient parameter selection.
• Subset designs for screening: Uses subset designs to span the entire parameter search space during an initial screening phase.
• Response surface and OLS modeling: Employs response surface designs and fits Ordinary Least Squares (OLS) models for parameter optimization.
• Multi-tool pipeline optimization: Optimizes parameters for single tools and multiple sequential tools within a pipeline while accounting for parameter interactions.
• Python implementation: Implemented in Python as the computational framework.
• Empirical performance gains: Identified parameter settings that outperformed default values in measured outcomes across tested scenarios.

Scientific Applications:

• De-novo assembly: Parameter optimization for de-novo genome assembly with demonstrated improvements over default settings.
• Scaffolding of fragmented genome assemblies: Optimization of scaffolding parameters for fragmented genome assemblies.
• K-mer taxonomic classification of Oxford Nanopore Technologies MinION reads: Tuning parameters for k-mer–based taxonomic classification of ONT MinION sequencing reads.
• Genetic variant calling: Optimization of parameters in genetic variant calling workflows.

Methodology:

Performs a screening phase using subset designs to explore the parameter space followed by an optimization phase using response surface methodologies and Ordinary Least Squares (OLS) modeling to fine-tune parameters.