Is there a way to perform clustering analysis by taking into account both gene fold change (e.g LFC) and their corresponding p-values?. My intuition is that while performing sample clustering, especially on data originating from different studies in complex diseases, there are some genes that contribute greater to sample variance than others. To take this into account, I am thinking of using both gene LFC and p-value to have a score representative of this fact and subsequently measure their euclidean distances. I have thought of multiplying LFC with -log10 p-value but I am wondering if there a reasonable justification (Statistical or otherwise) to do this?

The recently proposed Topconfect package is the closest to this approach using a different method of course, but is there a better justification, if any, to my proposed approach?