Learning about MACAU

Today I oriented myself to MACAU (Mixed model Association for Count data via data AUgmentation). MACAU is a program that assess the influence of a continuous predictor variable, e.g. age, on methylation while controlling for relatedness. To do so, it models raw read counts from bisulfie sequencing using a binomial mixed model. The software and manuals are available on the Zhou lab’s website

To learn about the principles underlying this program, and the problems this program addresses, I read this paper:
“A Flexible, Efficient Binomial Mixed Model for Identifying Differential DNA Methylation in Bisulfite Sequencing Data” by Amanda J. Lea, Jenny Tung, Xiang Zhou. https://doi.org/10.1371/journal.pgen.1005650.

The MACAU user manual is helpful to understand which input files I will need, which include:

1) Methylated read counts - Matrix containing read counts for methylated sites 2) Total read counts - Matrix containing read counts for all sites 3) Relatedness matrix - genetic relationship matrix for all individuals 4) Predictor - a vector of a continuous variable

Here are some helpful snippets from the paper and the manual:

MACAU analyzes raw read counts, not %, thus discerning between noise and DML. In practice, I’ll need to have 2 count files - 1) total read counts, and 2) methylated read counts. I don’t know wheter I have those files ready to go from Steven’s processing - that’s something to figure out.

MACAU includes a relatedness term in the model, thus controlling for genetic inheritance of methylation patterns. In practice, I will need a relatedness matrix file. Katherine generated this for me from SNPs. To start, I will use a relatedness files she generated using SNPs from only the Hood Canal and South Sound oysters, located in the repo’s 2bRAD directory.

MACAU tests the null hypothesis that the predictor of interest has no effect on DNA methylation levels:H0: β = 0. The predictor could be a variety of things. In our case, the predictor is shell size, wet weight. More than one variable can be inputed as a covariate matrix.

MACAU operates on a Linux computer. Sam installed the binaries on a few computers (see this github issue #722 (note: not the R package). Once I have all the input files I should be able to run MACAU fairly easily, since the command is simple, we’ll see!

./macau -g [filename] -t [filename] -p [filename] -k [filename] -bmm -o [prefix]

Written on August 7, 2019