NEW YORK (GenomeWeb) – Epigenetic profiling is poised for use in the clinic, including to help determine cancer subtypes and predict patient outcomes, according to a series of papers out today.
In four papers appearing in both Nature Biotechnology and Nature Communications, researchers from the European Blueprint project and the International Human Epigenome Consortium examined the accuracy of a number of epigenetic testing approaches as well as described their validation of genomewide DNA methylation sequencing technologies.
In addition, the researchers described a new bioinformatics approach for uncovering disease-specific DNA methylation patterns and applied chromatin mapping to the study of chronic lymphocytic leukemia.
"Epigenetic tests have a key role to play for making precision medicine a clinical reality. Epigenetics captures part of each cell's individual history, and it can predict how cancer cells will react to drug treatment," Christoph Bock, a principal investigator at the CeMM Research Center for Molecular Medicine of the Austrian Academy of Sciences, said in a statement. "This can be very useful for personalized therapy."
To first establish that DNA methylation approaches might be primed for clinical use, the Blueprint consortium, which includes CeMM's Bock, conducted a benchmarking study, which they presented in Nature Biotechnology today. In this study, the team sent 32 reference samples to 18 laboratories in seven different countries in order to gauge how well various methods could measure DNA methylation.
The assays tested included 16 absolute methylation assays based on technologies like AmpliconBS and Pyroseq and five relative methylation assays based on three alternative technologies such as MethyLight. The 32 reference samples encompassed DNA from primary colon tumor and normal colon samples; DNA from cell lines before and after drug treatment; a titration series of partially methylated DNA spiked into unmethylated DNA; and more.
Most assays, Bock and his colleagues reported, gave high accuracy and were robust, though they noted some variations across assay types and labs. Assays measuring relative methylation were, in general, less accurate and exhibited less concordance with each other than did absolute assays, they noted.
Still, they reported that the relative assays could distinguish between methylated and unmethylated regions as well as between tumor and normal samples.
Overall, they concluded that absolute DNA methylation assays "are the method of choice when validating DNA methylation differences in large cohorts."
UCL researchers led by Stephan Beck, meanwhile, conducted a saturation analysis for whole-genome bisulfite sequencing data to determine what depth of coverage is needed to capture most methylation signals, which they also described in Nature Biotechnology.
They analyzed more than a dozen whole-genome bisulfite sequencing-generated methylomes and conducted a down-sampling analysis to examine what information is lost at lower coverage levels. They correlated this coverage to the ability to pick up informative CpG sites, differentially methylated positions (DMP), differentially methylated regions, blocks of comethylation (COMETs), and differentially methylated COMETs (DMCs).
At 30X coverage — the current reference methylome coverage — they noted that some 50 percent differentially methylated regions are lost.
To extend that report, the UCL researchers presented in Nature Communications two algorithms to recover, in part, some of that lost information.
Dubbed COMETgazer and COMETvintage, the algorithms use the stochastic oscillations of DNA methylation to segment methylomes into COMETs and then call DMCs. Through this approach, Beck and his colleagues reported that they could recover about 30 percent of lost DMP information, even at low 5X coverage. This, they noted, is about double what could be retrieved through DMR analysis.
The approach, though, has its own limitations, they added, as it requires two methylome replicates.
Finally, CeMM's Bock and his colleagues showed in a separate Nature Communications paper that a large-scale chromatin analysis of chronic lymphocytic leukemia could uncover shared as well as subtype-specific chromatic variation.
In particular, they used the ATAC-seq assay to map chromatin accessibility in 88 primary CLL samples from 55 patients. For 10 of those samples, they also used ChIPmentation to develop histone mark profiles and RNA-sequencing to yield transcriptome profiles.
From this, Bock and his colleagues found that while there was a shared core of CLL regulatory regions, there was also widespread heterogeneity.
The chromatin profiles they generated could predict samples' IGHV mutation status — CLL with unmutated IGHV is typically more aggressive — as well as tease out two intermediary types of CLL falling between IGHV-mutated CLL and IGHV-unmutated CLL.
They further reported differences in gene regulatory networks involved in the two major disease subtypes, and these, they added, could be starting points for finding therapeutics
This all suggests, the Blueprint consortium said, that epigenetic profiling might soon be of use for clinical diagnoses and personalizing medicine.