Workflow from Scientific Research

Schematic description of the methodology to compute the stability-association scores. The method was based on the computation of follow-up distances (described in the top panel). Follow-up distance was the distance between the microbiome from a given subject at a given time point and that collected from the same subject at the immediately subsequent time point using any distance measures (we used Aitchison and Bray-Curtis). To rank taxa for their association with microbiome stability, we performed 10 iterations. In each iteration, we selected a subset of 12 studies and performed random-effects model-based analysis computing the association of each taxon with the two follow-up distances. In each iteration, for a given distance measure, we computed stability-association scores for each taxon by multiplying the magnitude of the summarized estimate (multiplied by 1 as follow-up distance is an inverse indicator of stability), the significance (log10ofqvalue) and the consistency (seeSTAR Methods) of its abundance association with the Aitchison and Bray-Curtis follow-up distances across cohort. Separate scores for the two distance measures were then averaged to get a final iteration-specific stability-association score. This iterative boot-strapped-based study-selection approach enabled us to reduce the influence of individual studies with possibly atypical effects. The final stability-association score (SS) was then calculated as the overall mean of the iteration-specific stability-association scores.

DOI