The genomic physics of tumor–microenvironment crosstalk

The recent years have witnessed the explosive application of sequencing technologies to study tumor–microenvironment interactions and their role in shaping intratumoral heterogeneity, neoplastic progression and tumor resistance to anticancer drugs. Statistical modeling is an essential tool to decipher the function of cellular interactions from massive amounts of transcriptomic data. However, most available approaches can only capture the existence of cell interconnections, failing to reveal how cells communicate with each other in (bi)directional, signed, and weighted manners. Widely used ligand–receptor signaling analysis can discern pairwise or dyadic cell–cell interactions, but it has little power to characterize the rock–paper–scissors cycle of interdependence among a large number of interacting cells. Here, we introduce an emerging statistical physics theory, derived from the interdisciplinary cross-pollination of ecosystem theory, allometric scaling law, evolutionary game theory, predator–prey theory, and graph theory. This new theory, coined quasi-dynamic game-graph theory (qdGGT), is formulated as generalized Lotka–Volterra predator–prey equations, allowing cell–cell crosstalk networks across any level of organizational space to be inferred from any type of genomic data with any dimension. qdGGT can visualize and interrogate how genes reciprocally telegraph signals among cells from different biogeographical locations and how this process orchestrates tumor processes. We demonstrate the application of qdGGT to identify genes that drive intercellular cooperation or competition and chart mechanistic cell–cell interaction networks that mediate the tumor–microenvironment crosstalk.