bioRxiv2022
BackgroundComplex diseases have multifactorial etiologies making clinically actionable diagnostic markers difficult to identify. Novel tools with higher diagnostic yield and utility in driving personalized care are needed. MethodsWe utilized Illumina methylation array data from over 2400 samples to assess DNA methylation patterns in 20 distinct cell types ranging from sperm to brain as well as various disease states. We generated a simple analysis pipeline for DNA methylation data that focuses on intra-individual methylation variability within gene promoters. The analysis is designed, not to identify single causative gene alterations but instead focuses on any movement away from "healthy" methylation. This approach identifies altered regulation across multiple genes in related pathways thus enabling us to detect shifts in gene regulatory activity associated with distinct tissues and phenotypes. We explored three distinct questions in our assessment. 1) Are patterns of epigenetic variability tissue specific? 2) Do diseased tissues exhibit altered variability patterns compared to normal tissue? 3) Can epigenetic variability be detected in complex disease? ResultsUnsupervised clustering analyses established that patterns of epigenetic variability are tissue specific and that these patterns are at least as predictive of tissue type as differential methylation analysis. We demonstrated the ability to use these patterns to differentiate between healthy and diseased tissue with unsupervised clustering even in cases of complex multifactorial diseases. We applied this method to the clinical use case of male infertility and found that men undergoing intrauterine insemination (IUI) with the lowest number of epigenetically dysregulated promoters in their sperm were almost twice as likely to father a child than men participating in IUI with the highest number of dysregulated promoters (p=0.011). We saw no significant difference in birth rates between groups of men with high and low numbers of dysregulated promoters undergoing in vitro fertilization (IVF), indicating IVF as a better treatment than IUI to achieve live birth in the presence of multi-pathway dysregulation in sperm. ConclusionsThis study demonstrates that patterns of epigenetic variability can differentiate between tissue types. While intuitive, this finding has never been demonstrated previously and suggests that specific epigenetic variability patterns may be used to predict phenotypic changes in disease states as these are, by definition, functional changes to cellular phenotypes. We demonstrate that the variability of gene regulatory marks are distinct between healthy and diseased tissue. This is particularly apparent at genes known to be important to cell function of the tissue of interest. While in some cases these regional alterations can be seen across the entire genome, more often the regulatory alterations that define a pathological phenotype are restricted to genes of known importance to a particular tissue. Importantly, in the case of sperm, we found that these patterns of variability did have utility in predicting infertile patients who would conceive through intrauterine insemination (IUI). We would propose that this discriminatory ability is due to the fact that the signature can be assessed in an n-of-1 context and that the patterns of variability identify any shift away from regulatory normalcy in pathways known to be impactful in the tissue of interest, and not only assessing the presence or absence of rare genetic variants. While the data presented here are encouraging, more work needs to be performed in other tissues to determine when, and in what context, these findings could be clinically actionable.