Fig. 2

Shifting procedure in simulation scenario B1. From time T1 to T4, taxa A-C decrease and taxon D increases in relative abundance. Compared to a low-risk set, taxa A, D are more abundant and taxa B, C are less abundant in a high-risk set. Taxon A is less abundant (i.e., \(M^{-}\)) and taxon B is more abundant (i.e., \(M^{+}\)) in a high-risk subject compared to the matched low-risk subject, both being true biomarkers at each time point. Taxon C is a pseudo biomarker (randomly selected as \(M^{0}\) at time point T1) with relative abundance automatically changed due to sum-to-one constraint, while taxon D is unchanged at T1