A physical basis for quantitative ChIP-sequencing

Abstract

Chromatin immunoprecipitation followed by next-generation sequencing (ChIP-seq) is a key technique for mapping the distribution of histone posttranslational modifications (PTMs) and chromatin-associated factors across genomes. There is a perceived challenge to define a quantitative scale for ChIP-seq data, and as such, several approaches making use of exogenous additives, or 'spike-ins', have recently been developed. Herein, we report on the development of a quantitative, physical model defining ChIP-seq. The quantitative scale on which ChIP-seq results should be compared emerges from the model. To test the model and demonstrate the quantitative scale, we examine the impacts of an EZH2 inhibitor through the lens of ChIP-seq. We report a significant increase in immunoprecipitation of presumed off-target histone PTMs after inhibitor treatment, a trend predicted by the model but contrary to spike-in based indications. Our work also identifies a sensitivity issue in spike-in normalization that has not been considered in the literature, placing limitations on its utility and trustworthiness. We call our new approach the sans-spike-in method for quantitative ChIP-sequencing (siQ-ChIP). A number of changes in community practice of ChIP-seq, data reporting, and analysis are motivated by this work.

H3K36 dimethylation shapes the epigenetic interaction landscape by directing repressive chromatin modifications in embryonic stem cells

“…In our samples, we used exogenous Drosophila (Orlando et al 2014) and synthetic nucleosome (EpiCypher) spike-ins (Supplemental Table S3). However, because of discrepancies (space limitations prevent us from discussing them here, but some of the problems are outlined in Dickson et al 2020), we eventually opted for mass spectrometry normalization. In our experience, we found this method to perform most reliably, but more research on optimizing quantitative normalization is necessary, particularly to distinguish small changes in the levels of epigenetic modifications.…”

Theoretical and practical refinements of sans spike-in quantitative ChIP-seq with application to p300/CBP inhibition

“…In our previous work[11], we built the inherent ChIPseq quantitative scale by first noting that the total number of reads available in a given IP can be written as where is the total depth of the IP and R is the total possible depth if the full IP mass were sequenced. ℱ L substituted into α to produce is the fraction of library sequenced, ρ is the total library concentration divided by the theoretical library concentration (what we call the library efficiency ), and ℱ is the fraction of IP’d material taken into library prep.…”

“…[17] Meanwhile, under the same treatment and IP conditions, A485 produced a 6.1-fold reduction in H3K18ac IP mass and a 4.9-fold reduction in H3K27ac IP mass. We carried the 1.6 and 2.5 µ g antibody points into sequencing, because this is where we would predict the highest antibody specificity [11].…”

“…The formal model of the IP binding process at the heart of siQ-ChIP explicitly expands chromatin into a list of all possible chromatin modification states[11], where we call each state a species. Below we explore the possibility that genomic annotations may provide a means of grouping several distinct species into larger classes.…”

“…Along with siQ-ChIP, we previously introduced the fractional composition of the IP as a way to present the distribution of IP products, and we studied this distribution through simulations to show that it can behave in some counterintuitive ways. [11] The fractional composition is identical to the distribution of fragments over the annotations, . The use of annotations as a proxy for species allows us to examine the fractional composition of actual IPs, and to visualize how the distribution of IP products responded to different p300/CBP inhibition paradigms.…”