Integrating effect of gender on an outcome

Hi All,

I am working on a project to study the influence of gender to predict cardiovascular outcome using baseline features. Let X= (NxM) matrix of M covariates with N subjects, Y=(Nx1) binary outcome and G = (Nx1) corresponds to gender (0=male, 1=female). Can someone suggest me a machine learning algorithm to implement so that i can derive a single number (or index) to study the influence of gender on the outcome?

[Walter Roberson]

If you are doing a medical study you really need to be more precise about what "gender" is. One of the few ways gender can be binary is if you define it in terms of detecting at least one SRY chromosome in multiple samples, and do not define the 0/1 as male or female but rather in terms of whether you detected any SRY or not. Is SRY even relevant? Probably not, but if you are forced to make a hypothesis and waste your time carrying out the experiments then at least it is objective. (And don't forget to read about mosaicing, especially after giving birth to a male child, as you are going to need some pretty good reasons to use with some of your female participants as to why you are classifying them as male.)

Answer 1

[MathWorks Bioinformatics Team]

rankfeatures can be used to rank your covariates with respect to your binary outcome & sex.

To Walter's comment, 'sex' is the word more commonly used in biological studies to define male/female-ness in humans and has a genetic basis, where most most subjects are strictly male or female. There are genetic variations where someone may not strictly satisfy the usual XX vs XY determination, but these instances are rare and may not apply to your subjects. 'Gender' is not a biological term, but refers more broadly to culturally masculine/feminine characteristics and is not considered a binary classifier.

There are many instances in which sex can be a determining factor in biological outcomes, and where treatment can differ between the sexes, so it should not be considered a 'waste of time'. Just be sure you are careful with your terminology.

[Walter Roberson]

The fraction that is not XX or XY is approximately the same as the fraction of the world that is red haired -- which is large enough that specific medical protocols are in place due to differences in reaction to anesthesia.

XX vs XY also does not tell you anything about Androgen Insensitivity Syndrome or the like. Or, as I mentioned earlier, about mosaicing.

[Sowmya MR]

@bioinformatics staff: I just read an article where one can use an approach called "uplift modelling" to model the influence bof a treatment or in this case, gwnder. Do you know if there is anything like uplift modelling in Matlab?