Fig. 1: Schematic of the analysis steps and modelling framework.

A Resting-state functional connectivity (FC) of the frontostriatal system was computed from subjects with OCD and matched healthy controls. B Frontostriatal resting-state neural dynamics were simulated using our computational model. The model parameters were optimised to fit the empirical FC of the two groups. C The optimisations produced group-specific distributions of the model parameters (denoted by \({{\boldsymbol{\theta }}}\)) and simulated FC. Parameter differences correspond to core neural mechanisms underpinning OCD frontostriatal pathology. Virtual interventions were performed by permuting sets of model parameters from OCD subjects (orange) to distributions observed in healthy controls (blue). D Subjects with OCD (real data) were paired to their digital twins (simulated data) via a distance measure in FC space (Methods). E Predictions of model parameters that contribute the most to normalise FC were derived from the virtual interventions. These predictions were tested by assessing the relationship between changes in the isolated neural parameters, FC, and fluctuations of obsessive and compulsive symptoms over time.