RUV-III-NB

RUV-III-NB corrects unwanted technical variation in raw sequencing count data using Negative Binomial and Zero Inflated Negative Binomial models to preserve true biological signals for downstream analyses.

Key Features:

• Model-based correction: Implements Negative Binomial and Zero Inflated Negative Binomial distributions to model overdispersion and excess zeros in sequencing count data.
• Data types supported: Applicable to single-cell RNA-seq and shotgun metagenomics datasets where overdispersion and zero inflation are common.
• Raw count input: Operates on raw sequencing count data without requiring platform-specific preprocessing.
• Technical-variation characterization: Identifies major sources of unwanted variation by analyzing datasets with deliberately introduced technical and biological factors.
• Comparative performance: Demonstrated superior removal of unwanted technical variation relative to ComBat, ComBat-seq, RUVg, and RUVs while preserving true biological signals.
• Implementation: Provided as an R package for analysis within the R environment.

Scientific Applications:

• Microbiome studies: Corrects technical confounders such as storage conditions and freeze–thaw cycles that can differentially affect taxa (for example, class Bacteroidia).
• Single-cell RNA-seq: Adjusts for overdispersion and excess zeros to improve detection of true biological variation.
• Shotgun metagenomics: Reduces technical variation in count-based taxonomic profiles to enhance downstream ecological and comparative analyses.

Methodology:

Uses Negative Binomial and Zero Inflated Negative Binomial modelling and analysis of datasets with deliberately introduced technical and biological variation to identify and remove major sources of unwanted variation.