How can one use MSstats to handle confounding factor such as gender?

[shimin-chen]

Dear MSstats team,

Thank you all for the great tool!

I have a question about using MSstats to handle confounding factors. If the samples are collected from patients with different characteristics such as gender, race, etc., how can one use MSstats to handle the potential differences contributed by those characteristics beyond just the disease status? Is it possible to have a model like ~ Gender + Race + Disease.Status ? The tutorial I found only discuss simple grouping like control vs treatment. The sample annotation file only allows one grouping factor at a time.

Thank you in advance.

[tonywu1999]

Hi,

Yes, you can have a model that handles all of these characteristics.

In the Condition column of your annotation file, you would define your conditions as something like [gender]_[race]_[disease]. Then, in the groupComparison function, you can create a contrast matrix where you can do more complex comparisons (you can see examples below various contrasts):

This approach will also reduce the estimated standard errors and increase the sensitivity. But keep in mind that if the number of replicates in each combination of gender/race/disease is relatively small, this will introduce the risk of overfitting (and false positives).

[shimin-chen]

Hi,

Thank you for your response! I gave that a try and it works well.

A follow-up question- can I include a continuous variable such as age in MSstats statistical modelling?

Thank you.

[tonywu1999]

At the moment, I don't think this is supported. Our models only support using discrete variables (e.g. old vs young) for defining conditions and covariates.

[shimin-chen]

Thank you for your response!