Abstract
Chromatin immunoprecipitation followed by next-generation sequencing (ChIP-seq) is a powerful technology for studying genetic and epigenetic regulation. However, ChIP-seq data can be heavily affected by variations in chromatin amount and composition, ChIP enrichment, library preparation, and sequencing depth, affecting its overall reproducibility across biological replicates. Here, we evaluated four ChIP-seq normalization methods utilizing triplicate Foxa1 ChIP-seq data performed in prostate cancer tissues from three mice. We found that count-per-million (CPM) normalization, although not affecting peak calling in individual samples, is very useful for visualization and comparison of peak distribution and intensity across samples. By contrast, equal-read normalization improves both peak identification and intensity comparison. Moreover, spike-in normalization took advantage of spike-in chromatin ChIP to correct technical variations in ChIP-seq, including ChIP enrichment, sample preparation, and sequencing. Lastly, input-adjusted spike-in normalization further accounts for differences in input chromatin amount across samples, which is especially crucial for tissue ChIP-seq that often starts with different amounts of input chromatin. Overall, our study demonstrated that appropriate normalization is essential to improve the reproducibility and comparability of ChIP-seq experiments and highlighted the importance of input-adjusted spike-in normalization for tissue ChIP-seq.