Species-Shared and -Specific Structural Connections Revealed by Dirty Multi-task Regression

Comparative studies across species, such as primates in this work, reveal the shared structural or functional patterns that might be inherited from the common ancestors and the specific ones that might be related to individualized evolution strategies. Both the shared or specific patterns could help promote the understanding of mechanisms of brain structural and functional architectures and brain dynamics. Many previous studies can be found to report the comparative results on species pairs. However, very few studies were found to perform a comparison of large-scale connectomes across multiple species via a data-driven method. To this end, we construct brain connectomes for three primates, macaque, chimpanzee and human, by using Brodmann areas as graphic nodes and diffusion MRI derived white matter fibers to define edges and edge weights. On these connectomes, a novel dirty multi-task regression method is developed in the attempt to automatically identified the species-shared and -specific connections. The concordance of the findings via our method and previous reports demonstrate the effectiveness and the promise of this framework.