question about age coefficient (coef (age.logFC))

[PolinaShichkova]

Dear all,

Thank you for sharing the great data!

I have probably a naive question, as I'm not very familiar with MAST and the details of your methods, and I definitely don't want to mess up something by reinventing your methods.

In your great paper https://www.nature.com/articles/s41586-020-2496-1 it's written multiple times "age coefficient threshold of 0.005 (corresponding to an approximately 10%-fold change)".

In the file facs.Brain_Non-Myeloid.neuron.gz from https://figshare.com/articles/dataset/tms_gene_data_rv1/12827615?file=27856758 I see "coef (age.logFC)", which is the "age coefficient" from the paper as far as I understand.

My questions: 1) Which transformation is applied to go from the age coefficient 0.005 to 10% fold change? 2) Does 10% fold change in this case mean 100%(old - young)/young or 100%(old/young)? 3) Which base of the log is implied throughout the paper when it's not specified and given just as log?

In other words, what I need is to get rid of the log scale in age coefficient, and since none of the common base choices produce 0.1 from 0.005, I thought I may be missing something about the methods.

Thank you very much for the clarifications!

Best regards, Polina

[aopisco]

@martinjzhang can you help @PolinaShichkova

[martinjzhang]

Hi @PolinaShichkova

Thanks for your interest in our work.

We used a linear model of log expr ~ age + covariates, where age is in the unit of month, and the regression coefficient for age is in the unit of log expr per month.

Which transformation is applied to go from the age coefficient 0.005 to 10% fold change?

Old mice are around 24 months and the young mice are around 3 months. Then if the age coefficient is 0.005, the log fold change between 24m mice and 3m mice is

log expr(24m) - log expr(3m) = (24-3) * 0.005 which is roughly 0.1

Does 10% fold change in this case mean 100%(old - young)/young or 100%(old/young)?

It means log (old expr / young expr) is around 0.1.

Which base of the log is implied throughout the paper when it's not specified and given just as log?

We used the natural base as in Scanpy log1p transformation https://scanpy.readthedocs.io/en/stable/generated/scanpy.pp.log1p.html#scanpy.pp.log1p

Please see the comparison aging gene paper for more details. https://elifesciences.org/articles/62293

[PolinaShichkova]

Thank you very much! Everything is absolutely clear now! I just didn't expect that (24-3) multiplication factor.

Thank you!