HTSmix

HTSmix analyzes high-throughput perturbation screens to detect significant phenotypic changes while modeling biological and technical variation.

Key Features:

• Statistical modeling framework: Employs linear mixed-effects models tailored for experimental designs that include at least two control samples profiled throughout the experiment.
• Normalization: Implements phenotype normalization to standardize measurements across samples.
• Variance estimation: Estimates variance components to separate stochastic variation from true phenotypic effects.
• False discovery rate control: Reports significant hits with control of the False Discovery Rate (FDR).

Scientific Applications:

• Evaluation on Saccharomyces cerevisiae screens: Validated on screens comprising 4,940 single-gene knock-out haploid mutants, 1,127 single-gene knock-out diploid mutants, and 5,798 single-gene overexpression haploid strains.
• Integration with experimental designs: Applicable to realistic high-throughput experimental setups that include distributed control samples.
• Extension to alternative workflows: Framework allows adaptation to varied experimental workflows.
• Sensitive detection of biologically meaningful changes: Demonstrates improved noise reduction and sensitive discovery of significant biological changes compared with existing procedures.

Methodology:

Linear mixed models to represent signal structure with fixed and random effects; phenotype normalization and summarization; estimation of variance components; control of the False Discovery Rate (FDR).