By Matthew Luchette
May 9, 2012 | BOSTON—Manolis Kellis, professor of computer science at MIT, addressed the Bio-IT World Conference on efforts to revolutionize the study of human disease by bridging the gap between genetics and epigenetics.
Kellis uses genomic datasets to discover relationships between single nucleotide polymorphisms (SNPs) and epigenetic regulation. Previously, these data sets of genomic variation were used to derive associations between SNPs and specific diseases in so-called genome-wide association studies (GWAS). Kellis’s research goes one step further. “We wanted to find out what the molecular mechanism is for this association,” he said.
Over the past five years, there has been an explosion of research on the statistical associations between specific SNPs and the likelihood an individual would get a disease. This paradigm led researchers to look for specific genes that caused a disease phenotype. The list of contributing genes for complex diseases, such as Crohn’s disease or type 2 diabetes, can run into the dozens, yet only explain a fraction of the heritability of the diseases.
What Kellis’s research shows, though, is that diseases are actually caused by subtle contributions from a huge number of regulatory variants. “If we rank all the SNPs based on the association with a disease,” Kellis said, “we find that there are not 10 or 20, but thousands of regions that weakly contribute to the disease.”
Instead of correlating SNPs with the likelihood of disease, Kellis’s research focuses on the association between SNPs and chromatin states. These chromatin states are epigenetic modifications to the genome that can affect the expression of a genetic element.
A gene’s chromatin state can then be linked to upstream regulators, such as activators or repressors, to show how the expression of the gene is affected. SNPs that are associated with high activator binding or inhibition, for example, can provide researchers with clues to how the gene is regulated in the disease. “The goal is to have a systems-level understanding of genomes and gene regulation,” said Kellis.
Kellis predicts his research could make the genome a more powerful tool to understand disease pathology. To test this, he followed hundreds of patients for more than ten years, half of whom had Alzheimer’s disease, and took brain samples after they died.
Using an mQTL analysis, a statistical program used to determine the distance between methylation sites and SNPs, Kellis found that Alzheimer’s patients had more methylated genomes and a high correlation between specific SNPs and methylation states.
However, Alzheimer’s patients could not be predicted de novo from their methylation states alone. “The predictive power is better than genome wide association studies, but is not yet actionable,” he said. Using his results to identify regulatory motifs in patients with Alzheimer’s, however, could suggest potential pathways for the disease. This same analysis could be applied to a host of human diseases to help researchers create a pathway for the disease and identify new drug targets.
“We’ve developed methods to tests hundreds of variants to see how global SNP changes affect diseases,” he continued. Currently researchers rely heavily on in vitro cell culture to delineate disease pathways and identify drug targets. Finding this information de novo from the genome could decrease both the time and money required to produce a treatment for the disease.