Menu

Multiomics integration [Discussion]

Reading 0:00
15s threshold

Sorry for the long rant - this is the reflections of a non-statistician on a very pressing problem in the biological sciences.

I am researcher often working with multivariate datasets derived from the same samples, but with very different nature. An example would be 100 stool samples analyzed by their microbiome (the microbial composition) as well as for metabolomics (the chemical composition). The purpose here would be to associate the microbes with the metabolites, principally to see which microbes produce what metabolites and ideally correlate this to the status of the host. A wonderful example would be to observe how Roseburia produces butyrate and this protects the colon.

I am, however, missing some statistical rigour in these types of analysis. Most often, each type of data is processed individually and integration is merely done by discussion. Occasionally, spurious and cherry-picked variables are regressed in order to produce some level of statistical evidence.…

Continue reading — create a free account

Join HashtagPLUS to read full articles, follow hashtags, vote, and join the conversation.