Coexpression patterns define epigenetic regulators associated with neurological dysfunction

Abstract

Coding variants in epigenetic regulators are emerging as causes of neurological dysfunction and cancer. However, a comprehensive effort to identify disease candidates within the human epigenetic machinery (EM) has not been performed; it is unclear whether features exist that distinguish between variation-intolerant and variation-tolerant EM genes, and between EM genes associated with neurological dysfunction versus cancer. Here, we rigorously define 295 genes with a direct role in epigenetic regulation (writers, erasers, remodelers, readers). Systematic exploration of these genes reveals that although individual enzymatic functions are always mutually exclusive, readers often also exhibit enzymatic activity (dual-function EM genes). We find that the majority of EM genes are very intolerant to loss-of-function variation, even when compared to the dosage sensitive transcription factors, and we identify 102 novel EM disease candidates. We show that this variation intolerance is driven by the protein domains encoding the epigenetic function, suggesting that disease is caused by a perturbed chromatin state. We then describe a large subset of EM genes that are coexpressed within multiple tissues. This subset is almost exclusively populated by extremely variation-intolerant genes and shows enrichment for dual-function EM genes. It is also highly enriched for genes associated with neurological dysfunction, even when accounting for dosage sensitivity, but not for cancer-associated EM genes. Finally, we show that regulatory regions near epigenetic regulators are genetically important for common neurological traits. These findings prioritize novel disease candidate EM genes and suggest that this coexpression plays a functional role in normal neurological homeostasis.

Promoter CpG density predicts downstream gene loss-of-function intolerance

“…First, the predicted highly LoF-intolerant genes show a 27.6-fold enrichment for genes heterozygous lethal in mouse (p = 1.03 · 10 −12 ), when compared against those predicted as non-highly LoF-intolerant. Second, they exhibit a 12.7-fold enrichment for transcription factors (p < 2.2 · 10 −16 ), consistent with the known dosage sensitivity of these genes (Jimenez-Sanchez et al, 2001;JG Seidman, C Seidman, 2002;Boukas et al, 2019). Third, they show a total depletion (odds ratio = 0) of olfactory receptor genes (p = 2.5 · 10 −5 ).…”

“…The fact that taking promoter EZH2 binding into account improves our ability to recognize LoF-intolerant genes on top of CpG density, implies that this mapping can be learned with even greater accuracy by incorporating information about other regulatory factors as well. However, a current barrier to achieving this is the relative paucity of genome-wide binding data across the full repertoire of transcription factors: the human genome encodes approximately 1500 transcription factors (Vaquerizas et al, 2009;Barrera et al, 2016;Lambert et al, 2018) and at least 300 epigenetic regulators (Boukas et al, 2019). In contrast to these numbers, currently ENCODE has profiled only 330 regulatory factors in K562 cells, the most extensively characterized cell line.…”

Co-expression analysis is biased by a mean-correlation relationship

“…We have previously observed that highly expressed genes tend to be more highly co-expressed (Boukas et al, 2019). Related, differential expression has been shown to confound differential co-expression (Farahbod, Pavlidis, 2019).…”

“…removing the top 4 principal components (PCs). The number 4 was chosen based on our previous analysis (Boukas et al, 2019), where we used positive and negative control genes to assess this. An alternative approach suggested by Parsana et al (2019) is to remove the number of PCs according to the estimated number of surrogate variables using SVA (Leek, Storey, 2007;Leek, Storey, 2008;Leek, Johnson, et al, 2012) with the number of PCs ranging from 10 to 30 in these same tissues -the impact of choosing a different number of PCs will be examined below.…”

“…Amongst popular methods for gene co-expression analysis is WGCNA which works on the correlation matrix and the graphical LASSO (Friedman et al, 2008) which works on the precision matrix, the inverse of the covariance matrix. These methods have been successfully used many times to gain biological insight (Ghazalpour et al, 2006;Oldham et al, 2008;Willsey et al, 2013;Saha et al, 2017;Boukas et al, 2019).…”