In our recent paper accepted by Cell Host & Microbe that examines intermicrobial interactions and subspecies transfer in FMT datasets, we find two superficially conflicting findings: i). SGBs with over 2-fold post-FMT abundance changes have greater numbers of associated subspecies taxa than those without (thus, SGB-SGB association may gradually alter intra-species SGB composition); ii). integrated data from datasets show that at the population level, the intra-species SGB composition is stable. Our latest findings may help understand this discrepancy: we divide multi-SGB bacteria into those with stable SGB composition and those with dynamic composition and investigate the effect size of significant SGB-SGB associations (FDR < 0.1) in different ranges of Spearman association rho values on the SGB composition of the two types of bacteria, and we find that stable composition bacteria and dynamic composition bacteria have indistinguishable effect sizes in all rho value ranges except rho > 0.99, in which dynamic SGB bacteria have a higher effect size, and the SGB-SGB with rho > 0.99, despite having a tiny fraction among all significant SGB-SGB association, display the highest impact on microbiota composition (see figure below); we also examine the mash-based genetic distances between these pairs of SGB-SGB association with rho > 0.99 and find the genetic distances are over 5%, just to make sure that these highly stiff associations do not represent a same subspecies somehow splitting into different metagenomic clustering. Thus, these highly stiff subspecies-level association, between entities of different genera, may alter conspecific lineage composition.
With this in mind, we want to expand our investigation into wide, non-FMT metagenomic datasets with at least three sampling timepoints per subject to contest our preliminary findings. Specifically for Lifeline dataset, as for other datasets, we will do SGB annotation (Pasolli et al., Cell, 2019), Spearman association analysis between SGBs, and intra-species SGB composition dynamics, and focus on around 70 bacteria with at least two detectable SGBs per species. We are also entertaining a thought of using 3% genetic distance (as opposed to 5% used by SGB clustering) to vigorously examine this question, namely subspecies-level interactions on intra-species composition.