An explainable graph neural framework to identify cancer-associated intratumoral microbial communities

Abstract

Microbes are extensively present among various cancer tissues and play a vital role in cancer prevention and treatment responses. However, the underlying relationships between intratumoral microbes and tumors are still not well understood. Here, we developed a MIcrobial Cancer-association Analysis using a Heterogeneous graph transformer (MICAH) to identify intratumoral cancer-associated microbial communities. MICAH integrates metabolic and phylogenetic relationships among microbes into a heterogeneous graph representation. It uses a graph attention transformer to holistically capture the relationships between intratumoral microbes and cancer tissues, which improves the explainability of the association between identified microbial communities and cancer. We applied MICAH to intratumoral microbiome data across five cancer types and demonstrated its good generalizability and reproducibility. We believe this graph neural network framework can provide novel insights into cancer pathogenesis associated with the intratumoral microbiome.