Sajal Kumar, B.Tech.
Joseph I. Said, Ph.D.
Joe Song, Ph.D.
New Mexico State University.
To quantify the effect of repressive chromatin states on gene expression.
Setting:Â Epigenetic factors have been qualitatively shown to heavily influence transcription regulation. The quantitative prediction of gene expression is however still an open question. Transcription factors, non-coding RNAs and chromatin states all play essential roles in regulating gene expression. In this study, we examine how epigenetic factors can quantitatively model transcription control using promoter-level gene expression atlas from the FANTOM5 project and the epigenome profiles in corresponding tissues and cells from the Epigenome Roadmap. Using these two datasets we were able to find associations between epigenetic factors and transcription levels of genes in specific tissue types. Such studies will enable us to provide possible causal clues to how cells function at the gene expression level.
Methods: We extracted the chromatin states for each transcript by its chromosome, position and strand orientation in each tissue type. We mapped these chromatin states of each gene to its expression level in FANTOM5. In a preliminary study we randomly picked 100 transcripts across 112 tissue types representing major cell types such as cancer cell lines, muscle cells, blood and T-cells, neurons, etc.
Results: The boxplots in Fig. 1 show gene expression as a function of chromatin state obtained from a total of 11,200 data points. Chromatin states 1 to 8 are known to enhance transcription, while 9 to 15 are known to suppress transcription of the nearby gene.
Figure 1. Gene expression as a function of chromatin state. The horizontal axis represents 15 chromatin states: 1_Active TSS, 2_Flanking active TSS, 3_Transcr. at gene 5â€² and 3â€², 4_Strong transcription, 5_Weak transcription, 6_Genic enhancers, 7_Enhancers, 8_ZNF genes + repeats, 9_Heterochromatin, 10_Bivalent/poised TSS, 11_Flanking bivalent TSS/Enh, 12_Bivalent enhancer, 13_Repressed Polycomb, 14_Weak repressed Polycomb and 15_Quiescent/low.
Interpretation: The states 9-15 on the boxplot show strong support to our hypothesis, despite some exceptions. We are working on creating a generalized logical network model to predict discrete gene expression levels without parametric assumption from the 15 chromatin states. This work will deepen our understanding of quantitative transcription control by epigenetics via chromatin remodeling