Probabilistic mouse-human brain correspondence by multimodal optimal transport

The mouse has long been the go-to model for understanding brain mechanisms and diseases, despite the fact that certain findings often fail to translate well due to disparities in the relative size of homologous regions and the absence of clear mouse equivalents for some human territories. Researchers have now developed OTTER, a novel tool capable of learning mouse-human brain correspondences as a probabilistic, parcel-resolution coupling.

This innovative approach utilizes multimodal fused Gromov-Wasserstein optimal transport to integrate functional and structural connectivity, spatial position, and curated homologies.

OTTER has proven effective in recovering established homologues when tested against a transcriptomic benchmark, and it maintains a broad cross-species organization along the cortical areal hierarchy. When applied to human functional connectivity, OTTER uncovers a graded decline in the reconstruction of mouse-based networks across evolutionarily expanded association cortex, with the lowest values observed in the lateral prefrontal territory.

Furthermore, the bidirectional map generated by OTTER enables the generation of testable human predictions derived from mouse experiments, as well as the ranking of mouse circuits corresponding to human clinical targets. This powerful new technique offers a promising avenue for bridging the gap between mouse and human brain research, ultimately paving the way for more accurate and effective therapeutic interventions.