A flexible statistical framework capable of identifying regions of enrichment across a wide variety of DNA-seq data types, enrichment patterns, and experimental conditions. ZINBA's flexibility in modeling background and enrichment regions with sets of covariates allows for the identification of enriched regions in difficult modeling conditions, such as in datasets with complex local CNVs or lacking a matching input control sample. ZINBA can identify both broad and sharp regions of enrichment, and we demonstrate this capability in differentiating RNA Pol II elongation status. In addition, the statistical framework used is applicable to both high signal-to-noise data such as from CTCF ChIP-seq, as well as to low signal-to-noise data such as from FAIRE-seq. ZINBA produces peak calls that are consistent with known biological patterns, and performs favorably relative to existing specialized methods over a broad range of signal patterns and data types.