NEW YORK (GenomeWeb) – Researchers from the University of Southampton have developed and published a method that uses Illumina HumanMethylation450 BeadChip arrays to conduct genome-wide methylation analysis in single samples without requiring large case-control cohorts.
The team shared its results in a study published last month in Clinical Epigenetics, showing that the method reliably picked up previously known and also novel methylated sites in patients with ultra-rare imprinting disorders.
Faisal Rezwan, the study's first author, told GenomeWeb in an email that he and his team hoped to develop an array-based method that could provide a cornerstone for better genetic diagnosis of these extremely rare disorders.
"Our main goal was to develop this method as a diagnostic tool, which will produce consistent results for any patients and also would be cheaper" than approaches like next-generation sequencing," he wrote.
In genomic imprinting, the expression of certain genes is altered due to parent-specific epigenetic changes. In other words, when an allele inherited from one parent is imprinted it results in only the allele from the other parent being expressed.
According to the Southamptom researchers, disruption of normal methylation — at a total of eight loci so far — has been associated with specific imprinting disorders including Prader-Willi syndrome, Angelman syndrome, and pseudohypoparathyroidism.
Using genome-wide methylation arrays could help expand the number of known imprinted loci, and also to help to establish better methods for clinical diagnosis, Rezwan and his coauthors wrote. However, cases of these disorders are extremely rare, meaning that standard array analysis techniques that look for small changes in methylation among large cohorts wouldn't work.
Previously, Rezwan and his colleagues created an informatic strategy capable of extracting statistically significant results from methylation array data using only five cases, which allowed them to detect methylation changes at known imprinted loci, and also identify 25 novel candidate imprinted regions.
In their most recent publication, Rezwan and his colleagues pushed this further to allow the detection of altered methylation sites in array data from single samples and without the need for large numbers of controls.
Whereas their approach for analyzing data from groups of five or more patients used linear regression as its statistical method, the researchers' single-sample strategy relies on something called a t-test, specifically the Crawford-Howell t-test.
In their study, the investigators tested this new analysis approach on patients with ultra-rare imprinting disorders who had known points of aberrant DNA methylation at multiple locations previously detected by targeted methods.
"For these disorders, we already knew some loci with hypomethylation. We took those loci as our positive examples and tried to identify any additional affected loci/regions, which can be associated with the disorder," Rezwan explained.
The team reported that its single-sample array analysis outperformed targeted testing in several ways: it detected novel methylated loci not normally covered by current targeted tests and it detected methylation changes too subtle for targeted testing technologies.
For example, in all the analyzed patients, the group was able to identify cardinal disease loci such as PLAGL1 and KCNQ1OT1 in the appropriate individual disorders.
Apart from these cardinal loci, the team's method also detected other significantly hypomethylated loci, both well-known and more recently identified. The strategy also identified other sites with altered methylation that had not been seen in prior targeted testing, including SVOPL and MAFG, the authors wrote.
The Crawford-Howell t-test method has the advantage of reporting not only the probability of significant methylation changes at a particular location, but also reports the "magnitude of the change by its effect size point estimate and confidence intervals," the researchers noted.
This allowed the investigators to determine the optimal number of control samples needed in the method. They found that fewer than 10 controls did not produce reliable enough effect sizes, whereas control sizes between 10 and 20 resulted in significant improvement, with only modest increases in effect sizes if the control group was inflated further.
They therefore set 20 controls as a suggested number for an optimal application of their t-test-based analysis.
According to the study authors, the data overall provide evidence that their informatics approach opens up the use of HumanMethylation450 BeadChip arrays for both diagnosis of known imprinting disorders and the detection of novel patterns of methylation.
Hopefully, this can help bring about "substantial improvements in the diagnostic rate and translational research for imprinting disorders, in the same way that genome-wide array analysis has advanced the clinical genetics of common diseases over the last fifteen years," the group wrote.