Skip to navigation Skip to main content Skip to footer

Approved Research

Identification of Genomic Risk Factors and Prediction of Problematic Alcohol Use Phenotypes, Outcomes, and Associated Cardiometabolic Phenotypes through Machine Learning

Principal Investigator: Dr Eric Zorrilla
Approved Research ID: 82297
Approval date: April 13th 2022

Lay summary

While risk factors for problematic alcohol use and alcohol use disorder (AUD) are becoming incorporated into risk prediction models, they have not yet reached utility to guide clinical management. A key barrier to clinical implementation of such risk models is building a reproducible risk model that may:

  1. Help prevent an alcohol-related hospitalization in someone who has never had an alcohol-related hospitalization.
  1. Estimate the risk that a patient hospitalized for problematic alcohol use may have subsequent alcohol-related hospitalization(s) or death.
  1. Estimate risk for adverse cardiometabolic outcomes associated with problematic alcohol use.

One kind of risk model that may help clinicians in the above points is a polygenic risk score (PRS). In short, a PRS would take genetic and non-genetic risk factors for problematic alcohol and output a numerical risk score for a specific patient. While PRSs can help clinicians identify high-risk individuals and guide treatment decisions, they have not been adopted specifically for problematic alcohol use/AUD. In addition, PRSs used in different fields are limited by two


  1. They include a small number of genetic variants .
  2. They do not capture variant-variant interactions or interactions between variants and environmental/clinical variables.

We will use a new technique to calculate an individual risk score, and our investigation of the genetic, environmental, and clinical features predictive of alcohol risk will not only provide insight into the biological basis of problematic alcohol use, but may even improve risk stratification as we build novel PRSs for alcohol-related hospitalizations and negative medical outcomes. Even further, we hope that our PRSs may help set the stage to help clinicians efficiently craft personalized treatment plans for their unique patients. The project will span at least three years, with novel risk factors identified within the first year and a preliminary personalized PRS anticipated within three years.