Histone modification levels are predictive for gene expression

Abstract

Histones are frequently decorated with covalent modifications. These histone modifications are thought to be involved in various chromatin-dependent processes including transcription. To elucidate the relationship between histone modifications and transcription, we derived quantitative models to predict the expression level of genes from histone modification levels. We found that histone modification levels and gene expression are very well correlated. Moreover, we show that only a small number of histone modifications are necessary to accurately predict gene expression. We show that different sets of histone modifications are necessary to predict gene expression driven by high CpG content promoters (HCPs) or low CpG content promoters (LCPs). Quantitative models involving H3K4me3 and H3K79me1 are the most predictive of the expression levels in LCPs, whereas HCPs require H3K27ac and H4K20me1. Finally, we show that the connections between histone modifications and gene expression seem to be general, as we were able to predict gene expression levels of one cell type using a model trained on another one.high CpG content promoter | low CpG content promoter | regression analysis | transcription T he DNA of eukaryotic organisms is packaged into chromatin, whose basic repeating unit is the nucleosome. A nucleosome is formed by wrapping 147 base pairs of DNA around an octamer of four core histones, H2A, H2B, H3, and H4 (1-5) which are subject to a number of posttranslational covalent modifications [(6); for review, see ref. 7]. These modifications can alter the chromatin structure and function by changing the charge of the nucleosome particle, and/or by recruiting protein complexes either individually or in combination (8). Hence, histone modifications are thought to constitute a "Histone Code," which is read out by proteins to bring about specific downstream effects (9, 10).Histone modifications have been linked to a number of chromatin-dependent processes, including replication, DNA-repair, and transcription. The link between histone modifications and transcription has been particularly intensively studied. It has been found that individual modifications can be associated with transcriptional activation or repression. Acetylation and phosphorylation generally accompany transcription; sumoylation, deimination, and proline isomerization are usually found in transcriptionally silent regions; methylation and ubiquitination are implicated in both activation and repression of transcription (8). Furthermore, the establishment of some modifications is dependent on the presence of other modifications, e.g., the catalysis of H3K4me3 requires the presence of H2BK120ub1 (the so-called trans-tail pathway) and the phosphorylation on serine 5 on the C-terminal domain of RNA polymerase II (pol II) (for review, see ref. 11, which also reviews other examples for the combinatorial action of histone modifications).Transcription proceeds in a series of steps, also referred to as transcription cycle, starting with preinitiation complex form...

Uniform, optimal signal processing of mapped deep-sequencing data

“…Our signal estimation method, EFilter, consistently outperformed the sum-of-tags approach 20 at predicting mRNA levels from histone ChIP-seq data. EFilter learns systematic gene-specific effects that are correlated across different cell lines, and suffers minimal loss of accuracy (1-2%) when trained on data from unrelated cell types.…”

“…Consider, for example, the problem of using histone ChIP-seq data to predict gene expression 20 . This is a 'signal estimation' problem, in which the observable variables (data) are used to estimate the value of a continuous hidden variable.…”

“…Expression was calculated as log-transformed RNA-seq fragments per exonic kbp per million (log-FPKM (fragments per kbp per million)) 21 . For comparison, we used an existing linear regression method 20 and predicted mRNA levels based on the total tag count of each ChIP-seq data set within the 6-kbp window around TSS. Training and test data were derived from the same cell line, using twofold cross-validation.…”

Application of machine learning methods to histone methylation ChIP-Seq data reveals H4R3me2 globally represses gene expression

“…They count tags only in a region surrounding the transcription start site (i.e., ±1 kbp [35] or ±2 kbp [36] of the TSS). A major problem with this method is that many marks do not have promoter/5' end biased enrichment patterns.…”

Unifying model for molecular determinants of the preselection Vβ repertoire

“…To examine these combinatorial relationships, we used classification and regression analyses comparing chromatin features and predicted RSS quality with Vβ use. These analyses were guided by recent computational strategies devised to predict gene expression levels based on patterns of histone modifications (50,51). We applied one validated approach (50) to study whether chromatin features, predicted RSS quality, and spatial proximity are predictive of the observed Vβ repertoire.…”