How does MAST caculate log fc

0

Entering edit mode

Chenghao ▴ 10

@06f8a532

Last seen 27 days ago

United States

Hi everyone,

I have read the paper of MAST and I also looked through the issues on github, but I still don't understand how MAST calculates log fold-change.

According to this documentation https://rdrr.io/bioc/MAST/man/logFC.html, the log fold change from contrast0 to contrast1 is defined as u(contrast1)v(contrast1)-u(contrast0)v(contrast0).

But if the model adjusts for some covariates, such as CDR, while I want to compare the gene expressions between two groups of cells, then how is the contrast specified? The CDR's are not the same across the cells in either group.

Thank you

MAST • 1.2k views

ADD COMMENT • link 2.9 years ago • updated 4 weeks ago Chenghao ▴ 10

0

Entering edit mode

I also figure that what logFC calculates isn't really the log-fold change. logfc calculates $p_1 * log(y_1) - p_0 * log(y_0)$, while log-fold change should be $log(p_1) + log(y_1) - log(p_0) - log(y_0)$

Edit: A potential issue of $log(p) + log(y)$ is that it may give non-positive values when p is small. Maybe we can interpret $p*log(y)$ as scaling mean log normalized counts by p.

ADD REPLY • link 5 weeks ago • updated 4 weeks ago Chenghao ▴ 10

Login before adding your answer.