Neuroscientists are working to discover how behavior and cognitive processes, particularly those relevant to mental illness, are precisely represented in the human brain to ultimately inform drug discovery targets.
Using functional MRI, genetics, and machine learning to undercover the biological basis of cognition, emotion, and motivation in the human brain.
The discovery of improved treatments/cures for mental illnesses is challenged by a weak understanding of the biological basis of these disorders. Because mental illnesses are disorders of the human brain, a key to identifying the important genetic/molecular mechanisms (to inform improved treatments) is understanding the genetic influences on brain activity related to cognition, emotion, and motivation that is often dysfunctional in mental disorders. One strategy at the LIBD to achieve this goal starts at the clinical/systems level in living people.
First, using functional magnetic resonance imaging (fMRI) and sophisticated imaging paradigms, we identify precise human brain activity underlying behavior and cognitive processes of interest by virtue of being dysfunctional in mental illness. In addition to determining the brain regions that are involved, we also use techniques such as dynamic causal modeling (DCM) to interrogate how brain regions are connected, or communicating with each other, to produce particular functions. Identifying important regional brain activity and connectivity patterns, however, is not defining mechanisms or uncovering treatment targets. It is necessary to go a step further to the cellular and molecular mechanisms that underlie neural activity and speak to individual variations, which cannot be ascertained with fMRI alone.
“Imaging genetics” is a technique that involves correlating genetic variation with neuroimaging signals to uncover genetic influences on neural activation patterns, which provide information regarding cellular/molecular underlying biology and explain individual differences. Traditional use of imaging genetics, however, has tended to involve the effect of one, or just a few, genetic variations at a time, without accounting for genetic background and probable complex multi-genetic influences. We have overcome these challenges by innovative implementation of a data driven strategy called, machine learning, specifically Random Forest Multiple Regression, which has the potential to uncover the influence of interactions among multiple genes. Therefore, rather than being limited to identifying the effect of an isolated genetic variation on brain activity which is not true-to-life, we can determine the influence of a group of genetic variations together in a more realistic fashion. Using machine learning allows us to also determine genetic variations that are most predictive of given neural activity, providing information regarding the most important molecular/cellular players underlying particular neural activity. We can use this information, in collaboration with basic and drug discovery scientists at the LIBD, to focus future investigations on particular promising genetic variations/pathways to inform drug discovery.