PICS

ChIP-seq is a method for identifying genome-wide transcription factor-DNA association. It uses chromatin immunoprecipitation with massively parallel short-read sequencing, which allows for high sensitivity, specificity, and spatial resolution. However, analyzing ChIP-seq data presents new challenges due to the complexity of the biological systems and variability in sequence data.

The authors developed PICS (Probabilistic Inference for ChIP-seq) to address these challenges. PICS is designed to identify regions bound by transcription factors from aligned reads. It does this by modeling local concentrations of directional reads and using DNA fragment length prior information to discriminate closely adjacent binding events via a Bayesian hierarchical t-mixture model.

One of the key features of PICS is that it estimates uncertainties in model parameters, which can be used to define confidence regions on binding event locations and to filter estimates. Additionally, PICS calculates a per-event enrichment score relative to a control sample, which can be used to estimate a false discovery rate.

Several alternative methods exist for analyzing ChIP-seq data, including MACS, QuEST, CisGenome, and USeq. However, PICS has been shown to outperform these alternatives. Using published GABP and FOXA1 data from human cell lines, PICS' predicted binding sites were found to be more consistent with computationally predicted binding motifs. Furthermore, a simulation study confirmed that PICS is robust to model misspecification.