Fig. 1

Prediction of microbiome composition based on its species assemblage. An illustration of the framework that is used in order to predict the abundance profile (open red circle), based on its known species assemblage, given a training set of other abundance profiles for which both the species assemblages and their abundances are known (blue spheres). We consider the species assemblages of the training samples, illustrated by ‘projecting’ them onto the assemblage subspace, and the abundance of the sample of interest is predicted based on the samples with the most similar species assemblages, using the k-nearest neighbors (kNN) method