Divergence in virus-host protein interactomes across species explains disparities in translational therapeutic success.

Infectious viral diseases remain a major and persistent threat to human health, and findings from animal models often translate poorly to clinical applications. Here, we constructed protein interactomes for seven species, including humans, revealing substantial differences in research biases. A random sampling strategy was employed to generate human interactomes, with sizes matched to those of the other six species to ensure cross-species comparability. We found that the clustering of virally targeted proteins within the interactomes (measured as largest connected component [LCC] proportion) and network fragmentation after their removal correlated with network density. Viral perturbations in species other than humans were heterogeneous, but generally showed reduced clustering and increased network fragments. The relative degree of network fragmentation (quantified by relative IC values) was correlated with local network conservation rather than sequence similarity, indicating that local network structure is selected, maintaining resilience during virus-host interactions. Furthermore, relative IC values and LCC proportion differences were positively associated with non-vaccine therapeutic success rates from phase I trials to approval. Our findings present a novel paradigm for the comparative analysis of disease phenotypes across species, thereby providing a new evaluative metric for selecting animal models in translational biomedical research.IMPORTANCECross-species network comparisons help uncover the molecular mechanisms of complex phenotypes. The viral module formation within interactomes and the network resilience during viral infection are of key impacts on hosts. The analysis of the interactomes of seven species shows that network density critically influences the network metrics performance. Under size-matched conditions, non-human species exhibited more complex viral perturbations than humans, but with smaller viral module size and more network fragmentation. The difference in fragmentation had a positive correlation with the conservation of the local network structure of virally targeted proteins, implying that viral perturbations act as evolutionary signals at the network level. Additionally, differences in fragmentation and viral module size were positively associated with the rates of successful progression from phase I trials to approval. This work offers a new framework for cross-species disease analysis and guides model organism selection for viral infection research.