Public Library of Science
Browse
Figure_1.tif (620.17 kB)

Microarray based detection of allele-specific methylation.

Download (0 kB)
figure
posted on 2014-06-09, 02:58 authored by John N. Hutchinson, Towfique Raj, Jes Fagerness, Eli Stahl, Fernando T. Viloria, Alexander Gimelbrant, Johanna Seddon, Mark Daly, Andrew Chess, Robert Plenge

A) A simplified representation of the Methyl-Sensitive Restriction Enzyme (MSRE) based Allele-Specific Methylation (ASM) assay. DNA is MSRE treated (left panels) and MSRE sites with methylated CpGs protected from digestion (upper panels, Allele-A) while its homologous chromosomal region with unmethylated CpGs are not (lower panels, Allele-B). The DNA is digested with StyI and NspI to form 200–1200 bp fragments, linkers ligated and DNA amplified to create amplicons that are hybridized to the array. Only regions with protected MSRE sites (methylated CpG) are amplified and can hybridize to show signal on the array (final panel). B) Bioinformatic detection of Allele-Specific Methylation (ASM) from Affymetrix SNP 6.0 arrays signals after MSRE digestion. In the scatter plot on the left, 4 different expected states after MSRE digest at a heterozygous region are compared to the typical distribution of probe intensities observed within the HapMap samples for the same MPR (here portrayed by light grey squares): biallelic methylation (dark grey circles), monoallelic A methylation (blue circles), monoallelic B methylation (yellow circles) and finally biallelic lack of methylation (red circles). The primary calling method relies on feature extraction by way of conversion of 2-dimensional A and B probe intensity data (scatter plot) from heterozygotes to log2(A/B) values and is compared against the typical log2(A/B) distribution observed for this MPR within the HapMap samples (histogram, light grey). Simply put, MPRs diverging from this distribution after MSRE treatment are called ASM. Using this method, biallelic unmethylated states have the potential to result in false positive ASM calls as any log2(A/B) value would be based on background noise, so are filtered out by removing MPRS with low total intensities (highlighted here with a red quarter-circle, for further information on how this filter was devised, see Figure S8).

History