Normalising fluxes between reactions

[manas-kohli]

Dear all,

I'm trying an idea a bit out there but I'm trying to use flux values for reactions in machine learning settings. Different reactions obviously exist on different flux scales as one can expect with metabolism but I was wondering if there were some way of normalising fluxes so they can exist on the same scale which would be important for feature contribution in machine learning. I've tried the usual methods, standardization to make it fit a normal distribution as much as possible or robust scaling i.e. (x'-median(x)/(IQR(x)). These methods work to some degree but any other input would be greatly appreciated.