Disease Heterogeneity Analysis

We developed a framework based on distribution matching, for the clustering of heterogeneous disease effects observed in a population (CHIMERA). More precisely, our method determines a set of disease effects which should be applied to a group of control subjects to cover a distribution of patients. Each patient is assigned to the combination of disease effects which seems to have affected his brain. This assignment can be seen as a soft clustering of the patients into disease subtypes.


[A] CHIMERA finds a set of transformations $T_1,\ldots,T_k$ allowing to cover the distribution of petients $Y$ by modifying the distribution of control subjects $X$. The $T_i$ are assumed to correspond to different disease effects. Clustering the subjects by disease effect reveals disease subtypes. [B] Voxel-based Morphometry analysis of two Alzheimer's Disease subgroups found by CHIMERA.

associated publication