PICS: Probabilistic Inference for ChIP-seq

ChIP-seq, which combines chromatin immunoprecipitation with massively parallel short-read sequencing, can profile in vivo genome-wide transcription factor-DNA association with higher sensitivity, specificity and spatial resolution than ChIP-chip. While it presents new opportunities for research, ChIP-seq poses new challenges for statistical analysis that derive from the complexity of the biological systems characterized and the variability and biases in its digital sequence data. We propose a method called PICS (Probabilistic Inference for ChIP-seq) for extracting information from ChIP-seq aligned-read data in order to identify regions bound by transcription factors. PICS identifies enriched regions by modeling local concentrations of directional reads, and uses DNA fragment length prior information to discriminate closely adjacent binding events via a Bayesian hierarchical t-mixture model. Its per-event fragment length estimates also allow it to remove from analysis regions that have atypical lengths. PICS uses pre-calculated, whole-genome read mappability profiles and a truncated t-distribution to adjust binding event models for reads that are missing due to local genome repetitiveness. It estimates uncertainties in model parameters that can be used to define confidence regions on binding event locations and to filter estimates. Finally, PICS calculates a per-event enrichment score relative to a control sample, and can use a control sample to estimate a false discovery rate. We compared PICS to the alternative methods MACS, QuEST, and CisGenome, using published GABP and FOXA1 data sets from human cell lines, and found that PICS’ predicted binding sites were more consistent with computationally predicted binding motifs.

💡 Research Summary

The paper introduces PICS (Probabilistic Inference for ChIP‑seq), a statistical framework designed to extract transcription‑factor binding information from aligned ChIP‑seq reads. The authors begin by highlighting the advantages of ChIP‑seq over ChIP‑chip—greater sensitivity, specificity, and spatial resolution—but also note the new analytical challenges posed by sequencing biases, variable fragment lengths, and the presence of repetitive regions that cause read loss. PICS tackles these issues in a multi‑stage pipeline. First, it scans the genome for local enrichments of directional reads, exploiting the fact that forward and reverse reads should converge toward the protein‑DNA complex. This directional clustering yields candidate binding regions. Second, a Bayesian hierarchical t‑mixture model is fitted to each candidate. The model incorporates a prior distribution on DNA fragment length, allowing per‑event fragment‑size estimation and the removal of events with atypical lengths. By using a truncated t‑distribution, the method also adjusts for missing reads in low‑mappability regions, leveraging pre‑computed whole‑genome mappability profiles. Parameter inference is performed via Gibbs sampling, producing posterior distributions for location, fragment length, and dispersion; the resulting uncertainties are used to define confidence intervals and to filter low‑confidence calls. Finally, PICS computes an enrichment score for each event relative to a control (input) sample and estimates a false discovery rate (FDR) across the dataset. The authors benchmark PICS against three widely used tools—MACS, QuEST, and CisGenome—using publicly available GABP and FOXA1 ChIP‑seq data from human cell lines. Across multiple metrics, PICS demonstrates superior performance: it identifies more sites that contain the known consensus motifs, it separates closely spaced binding events more accurately, and it reduces false positives by discarding regions with abnormal fragment length distributions or poor mappability. The study concludes that the explicit modeling of fragment length priors, directional read patterns, and genome‑wide mappability within a Bayesian framework yields higher resolution and reliability than existing peak‑calling algorithms, making PICS a valuable addition to the computational toolbox for transcription‑factor and chromatin‑modification studies.