"Domain gauges": A reference system for multivariate profiling of brain fMRI activation patterns induced by...

Pharmacological magnetic resonance imaging (phMRI) of the brain has become a widely used tool in both preclinical and clinical drug research. One of its challenges is to condense the observed complex drug-induced brain-activation patterns into semantically meaningful metrics that can then serve as a basis for informed decision making. To aid interpretation of spatially distributed activation patterns, we propose here a set of multivariate metrics termed "domain gauges", which have been calibrated based on different classes of marketed or validated reference drugs. Each class represents a particular "domain" of interest, i.e., a specific therapeutic indication or mode of action. The drug class is empirically characterized by the unique activation pattern it evokes in the brain-the "domain profile". A domain gauge provides, for any tested intervention, a "classifier" as a measure of response strength with respect to the domain in question, and a "differentiator" as a measure of deviation from the domain profile, both along with error ranges. Capitalizing on our in-house database with an unprecedented wealth of standardized perfusion-based phMRI data obtained from rats subjected to various validated treatments, we exemplarily focused on 3 domains based on therapeutic indications: an antipsychotic, an antidepressant and an anxiolytic domain. The domain profiles identified as part of the gauge definition process, as well as the outputs of the gauges when applied to both reference and validation data, were evaluated for their reconcilability with prior biological knowledge and for their performance in drug characterization. The domain profiles provided quantitative activation patterns with high biological plausibility. The antipsychotic profile, for instance, comprised key areas (e.g., cingulate cortex, nucleus accumbens, ventral tegmental area, substantia nigra) which are believed to be strongly involved in mediating an antipsychotic effect, and which are in line with network-level dysfunctions observed in schizophrenia. The domain gauges plausibly positioned the vast majority of the pharmacological and even non-pharmacological treatments. The results also suggest the segregation of sub-domains based on, e.g., the mode of action. Upon judicious selection of domains and careful calibration of the gauges, our approach represents a valuable analytical tool for biological interpretation and decision making in drug discovery.