Translational Bioinformatics and Precision Therapeutics Group

The Translational Bioinformatics and Precision Therapeutics Group identifies therapeutic opportunities by developing and applying machine learning approaches for integrating genetic data with multi-omics datasets towards the development of novel therapeutic approaches and precision psychiatry.

Under the direction of Giorgios Voloudakis, MD, PhD, our group performs three functions within the Center for Disease Neurogenomics (CDN):

  1. Identification of actionable genetic variation associated with neuropsychiatric phenotypes
  2. Gene target prioritization for neuropsychiatric disorders
  3. Repurposing existing drugs for neuropsychiatric disorders and precision psychiatry

Identification of Actionable Genetic Variation Associated with Neuropsychiatric Phenotypes

As part of the discovery arm of the Center, we develop and apply algorithmic phenotyping approaches to uncover the genetic architecture of mental illness and relevant dimensions such as psychiatric symptoms and neurocognitive dimensions, which have resulted in the first genome-wide association study of binge-eating disorder). We carry out most of this work in genotyped cohorts with Electronic Health Record (EHR) information such as the Veteran Affairs Million Veteran Program (MVP) and Mount Sinai’s BioMe BioBank Program, and also develop approaches to better bridge biobank findings with traditional case-control studies.

With information on a population-level scale, we are able to categorize individuals according to risk and severity of disease in subgroups for each disorder. We are also able to see what medications were prescribed and the results. Based on this cumulative data, we can start to understand how we can treat people differently that have the same disorder when factoring in their genetic makeup.

Gene Target Prioritization for Neuropsychiatric Disorders

The discovery arm of the Center identifies variants associated with neuropsychiatric disease. However, with thousands of variants each contributing a small factor to the overall disorder, it is vital to translate these findings to actionable approaches. The first step is to integrate genetic information with multi-omics datasets generated at our Center to prioritize which genes to target.

As the translational arm of the Center, we are developing and applying novel machine learning approaches for the integration of biological and clinical data towards gene target prioritization for neuropsychiatric disorders. We build predictive models based on the molecular profiling efforts of the Center to infer how genetic variation drives changes in the epigenome, transcriptome, and proteome across different areas and cell types in the brain. We can then apply these predictive models to impute disease-specific gene expression dysregulation and then identify approaches to reverse it.

By leveraging available gene perturbation libraries, we are using prioritization strategies to determine the most effective genes to target for a specific disorder. We were able to validate this new method in the early days of the COVID-19 pandemic. We used this approach to identify the IL10RB gene as a key regulator of host susceptibility and severity for COVID-19. An infection is a simpler case than complex brain disorders, and it allowed us to test this approach with fewer variables. Today, we are in the process of applying it to neuropsychiatric disease and developing the ability to expand in vitro testing in specific cell types disproportionally affected by disease (e.g. microglia for Alzheimer’s disease and neurons for bipolar disorder and schizophrenia).

The Center is at the forefront of brain molecular profiling efforts. As a result, we constantly innovate and develop new methods which will leverage emerging genomic datasets to address existing limitations in our field. Once brain-specific annotation of the genome became available, we introduced a machine learning approach called EpiXcan. EpiXcan increases prediction accuracy in transcriptome imputation by integrating epigenetic data to model the prior probability that a SNP affects transcription. Our approach improved the identification of gene-trait associations with tissue-specificity, which is critical in research about brain disorders because gene expression varies by cell-type and plays an important role in pathogenesis. We are now developing the next generation framework that jointly considers the effect of genetic variation on epigenome, transcriptome, and proteome dysregulation with spatial, cell type, and temporal resolution.

Repurposing Existing Drugs for Neuropsychiatric Disorders and Precision Psychiatry

We are confident that leveraging genetically driven disease-specific epigenome, transcriptome, and proteome dysregulation will jumpstart future drug development by identifying new candidate gene targets and molecular pathways. However, this is a lengthy and costly process, usually taking years and billions of dollars until it starts improving patient lives. Thus, we have a parallel process where we identify which medications approved by the Food and Drug Administration can be repositioned for treatment of psychiatric disorders by antagonizing the genetically driven gene expression dysregulation of a given disorder. By leveraging gene expression signature libraries of medications, we are able to:

  • Identify medications that can be repositioned for the treatment of neuropsychiatric disorders. We have tested the former in vitro and it worked in principle for COVID-19, and we are now starting to apply it for neuropsychiatric disorders. The identified compounds may be a great starting point for accelerated preclinical and clinical studies or, alternatively, they may provide a starting point for further modification.
  • Identify genetically based optimal treatments for patients. The approach holds great promise for precision and personalized psychiatry, as it can suggest specific drugs based on each individual’s genetic profile.