From what I've skimmed genes and gene expression seem to play a huge role (e.g. [1] [2]), but it shows the signs of being a multidimensional space where researchers are far from finished even figuring out what the dimensions actually are. Some of the popular press has mutated this into the notion of generic "exercise super-responders" and "exercise non-responders", but as usual they seem to be dumbing it down to the point of saying something almost the opposite of what the research actually supports.

[1] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2886694/

[2] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2682667/