From trees to graphs: rethinking phylogeny in microbiome prediction

Microbiome prediction models overlook phylogeny or fail to preserve evolutionary structure. In a recent Applied and Environmental Microbiology article (B. Dong, B. Wang, J. Chen, X. Xu, and Z. Z. Xu, Appl Environ Microbiol 92:e00788-26, 2026, https://doi.org/10.1128/aem.00788-26), Dong et al. present PhyloGCNE, a graph-convolutional framework that addresses these limitations by preserving evolutionary topology and learning adaptive edge-aware signal propagation. This commentary evaluates whether and when phylogenetic information improves microbiome-based prediction and considers how biological relevance and transferability can be established across contexts.