An Evaluation of Narrowpeak Quality Metrics in Chip-Seq Datasets: Signalvalue, Q-Value, and Their Relationship with Chromatin Markers

Chromatin Immunoprecipitation followed by Sequencing (ChIP-Seq) is a popular technique for detecting areas of protein-DNA binding. Transcription Factors (TFs) are one such type of DNA-binding protein. Typically, regions from ChIP-Seq are evaluated based on the observed significance of the amount of overlapping tag reads in a locus (e.g., via corrected p-value). However, one often-overlooked feature of called binding peak is the signalValue, which is the average tag read density in a location. SignalValue information is more granular than the significance level and might allow for more informative relations between binding peaks and regions of interest to be found. In this study, we first discuss the association between signalValue and q-value in bound regions (as detected by ChIP-Seq). We show that the correlation is very high (∼80%). Secondly, we investigate whether either measure follows expected association trends with markers of open or closed Chromatin (such as H3K4ME3 and DNase). Surprisingly, we show a negative association between signalValue (or q-value) and DNase in regulatory regions (via Mann-Whitney U test), as well as with H3K4ME3 (although less so). We corroborated our results using underlying read counts from processed bam files, suggesting that “high-significance” ChIP-Seq regions might not associate with open chromatin as biological intuition would suggest, while chromatin repressors such as H3K27ME3 might be more reliable in the general case.