The human brain is the prime example of “highly specialized and diverse functions arising from the same genetic program.” But that comes at steep price. Our most complex organ is, well…very complex. Not only do it’s epigenetic marks constantly change both spatially and temporally but it’s also mainly and almost equally composed of two very different cell types; Neurons and their underdog ‘helper’ pals, glia, which both come in a number of flavors. This heterogeneity is the stuff of nightmares and dreams.
Apparently after one nightmare too many, Dr. Margaret Taub and a research team at Johns Hopkins University decided it was time for a change. They set out to develop “a novel statistical epigenetics approach that takes advantage of the stability and cell-type specificity of DNA methylation…in order to deconvolve the two main cell components in the brain.” Here’s what they did:
- Created an algorithm that doesn’t require cell-type sorting, which is a major time and sample hog.
- This algorithm was based on the principle that “the methylation signal for any sample at a given genomic location, can be modeled as a linear combination of the methylation levels of neuronal and glial fractions in the brain region where the sample was obtained”
- Although developed on the CHARM platform, the authors demonstrated that it could also be effectively applied to the ever-popular 450k array.
- The authors point out that “this statistical framework also exposes the problem with existing naïve approaches to assess DNA methylation signatures in mixed samples.” Strikingly, they go on to show how other approaches give false positives due to “cell-fraction variation.”
Ultimately this method enables the calculation for the “first approximation [of] the difference in DNA methylation that is neuron- or glia- specific. Moreover, once sorted data is available for a given brain region, investigators can use such data to calculate cell proportions on any unsorted sample measured on the same methylation platform without the need to sort themselves.” The authors conclude, “this approach should have broad application to a range of problems in neurodevelopment and disease research.”
Check out this brain teaser over at Genome Biology, September 2013.