HTSFilter

HTSFilter implements a data-driven filtering procedure based on the Jaccard similarity index to remove lowly expressed genes from replicated RNA sequencing (RNA-seq) count data and improve detection power while aiding false-discovery rate control in differential expression analyses.

Key Features:

• Jaccard Similarity Index-Based Filtering: Uses a global Jaccard similarity index to determine an optimal filtering threshold by quantifying overlap of expressed genes across replicates.
• Data-Driven Thresholding: Dynamically adjusts the filtering threshold according to dataset-specific expression characteristics to retain relevant genes for downstream analysis.
• Increased Detection Power: Removes genes with consistently low expression across experimental conditions to enhance detection of moderately to highly expressed differentially expressed genes.
• Adaptability Across Experiments: Produces experiment-specific thresholds that reflect unique expression profiles and replicate structures of each dataset.
• Implementation: Implemented in R.

Scientific Applications:

• Differential Expression Analysis: Pre-filters RNA-seq count data to improve statistical power in identifying differentially expressed genes.
• False-Discovery Rate Management: Supports control of false-discovery rates by removing low-information features prior to testing many genes.
• Studies with Subtle Expression Changes: Beneficial for experiments and complex biological systems where small but biologically relevant expression changes are critical.

Methodology:

Calculate the Jaccard similarity index of expressed genes across replicates to derive a global filtering threshold that excludes genes with low, constant expression; performance was evaluated through comparisons with alternative data filters.