By Monica Heger
A recent comparison of four frequently used sequencing-based DNA methylation methods found that that they all provided comparable methylation calls, but differed in a number of features, including CpG coverage, resolution, quantitative accuracy, efficiency, and cost.
In a study published this week in Nature Biotechnology, researchers from several institutions described their evaluation of two methods based on bisulfite conversion: methylC-seq and reduced representation bisulfite sequencing; and two methods that use enrichment of methylated DNA: MeDIP-seq and MBD-seq. The team also developed a technique that combined MeDIP-seq to detect methylated CpGs with MRE-seq to detect unmethylated CpGs.
"There's a growing interest in epigenetic research, and a growing interest in methylation data," said Alan Harris, an assistant professor of bioinformatics at Baylor College of Medicine and lead author of the paper. "In order to be useful, we need to know how various techniques compare."
Each of the techniques was examined for accuracy, cost, concordance, and how CpG density and genomic content influenced the results. The different techniques have different advantages and drawbacks, Harris said. For instance, methylC-seq is the most expensive because it requires the most sequencing, but it also offers the highest resolution, he said.
The researchers tested three of the methods on biological replicates of a human embryonic stem cell line using the Illumina Genome Analyzer. For the methylC-seq method, they used previously generated data.
MethylC-seq and reduced representation bisulfite sequencing are similar in that both methods involve converting unmethylated cytosines to uracils, and then sequencing. But, while methylC-seq requires shotgun sequencing of the entire genome, the reduced representation method limits the portion of the genome that is sequenced through restriction enzyme digestion, which targets the portions of the genome containing CpG islands.
Both of these methods yield single-base resolution, said Harris. The difference is the amount of sequencing. "Reduced representation is a good way to go if you're interested in CpG islands," he said, because it offers single-base resolution, but is less costly than methylC-seq. MethylC-seq, on the other hand, would be most appropriate for researchers who require very high resolution coverage of the entire genome, and who also want to look for other mutations such as SNPs, which are detectable using the methylC-seq method.
The enrichment-based approaches — MeDIP-seq and MBD-seq — do not offer single-base resolution. The MeDIP-seq method first uses an antibody to pull down the methylated cytosines, similar to an exon capture, and then amplifies the methylated cytosines. The MBD-seq technique is the same, except it uses a protein for the pull down.
In both enrichment methods, Harris said that the main drawback is the lack of resolution, which is limited to the size of the fragments, in this case about 150 base pairs.
One advantage of the MBD-seq method is that by changing the salt concentration in the capture step, researchers can customize the regions that they pull down, opting to target regions that are more less CpG dense, Harris said.
Both enrichment methods are also less expensive than methylC-seq and target genome-wide CpGs better than the reduced representation method, so these methods could be useful if cost is an issue and researchers want to target more than just CpG islands, Harris said. Although, he added, costs of sequencing are declining that the methylC-seq technique may soon be affordable even for labs with modest budgets.
Comparing CpG coverage between the four methods found that the methylC-seq provided the highest coverage at 95 percent, while the reduced representation method covered the fewest CpGs genome-wide. However, reduced representation did provide substantial coverage of CpGs in CpG islands. MeDIP-seq and MBD-seq yielded 67 percent and 61 percent coverage of CpGs, respectively.
Additionally, the team analyzed a fifth method, which combines MeDIP-seq with MRE-seq — an enrichment method that enriches for the unmethylated portion of the genome. MRE-seq also helps improve resolution because the method yields single-base resolution, Harris said. Combining those two methods allowed the researchers to identify partial methylation, where only one allele is methylated.
"You see the signals separately," said Harris. "So, if you see both signals, it's a sign of intermediate methylation, or imprinting."
He said the next steps are to continue doing methylation profiles of normal genomes. As part of the National Institutes of Health's Epigenomic Roadmap Initiative, his team at Baylor is creating methylation profiles of breast cells using the MeDIP-seq and MRE-seq combination.
Eventually, researchers will be able to do methylation profiles of disease states and compare methylation in diseases such as cancer and Alzheimer's, to normal states.
Before researchers can determine the epigenetic modifications that are relevant to disease, however, they will need to know the normal states, he said. "Our role is to get some idea of what the normal epigenetic modifications are in humans."