STARRPeaker

STARRPeaker performs statistical analysis of STARR-seq data to identify active regulatory regions (enhancers) while correcting for sequencing biases and overdispersion.

Key Features:

• Negative binomial regression framework: Implements a negative binomial regression model tailored for STARR-seq to model counts, mitigate overdispersion, and accommodate non-uniform coverage.
• Bias correction with covariates: Incorporates covariates such as GC content, mappability, and conservation to adjust for sequencing and genomic biases.
• Mitigation of confounders: Addresses confounding effects that arise in STARR-seq data, including biases related to GC content and library complexity from deep sequencing.
• Genome-wide enhancer identification: Applied to whole-genome STARR-seq datasets to identify enhancers across the genome, including analyses in human cell lines HepG2 and K562.

Scientific Applications:

• Enhancer discovery: Identification of active regulatory regions (enhancers) from STARR-seq experiments on a genome-wide scale.
• Gene regulation and chromatin studies: Facilitates interpretation of STARR-seq results for studies of gene regulation and chromatin dynamics by correcting for technical and genomic confounders.

Methodology:

Integration of covariates (GC content, mappability, conservation) into a negative binomial regression model to adjust for sequencing biases and overdispersion.