Eric Zorrilla
Early Career Tenure-track Researcher, Scripps Research
2 active projects
Problematic Alcohol Use and Related Psychiatric and Cardiometabolic Conditions
Scientific Questions Being Studied
The aim of this proposal is to identify new genetic and non-genetic factors that are leveraged through a machine-learning based approach to predict individual 1) risk for problematic alcohol use, 2) outcome from problematic alcohol use (e.g., recurrence of alcohol-related hospitalization or death), and 3) genetic and non-genetic alcohol risk factors that are shared with other psychiatric diseases and obesity-related cardiometabolic phenotypes. Prediction will utilize a combination of genetic, clinical, and lifestyle risk factors. Ultimately, we aim to identify not only individual predictors, but build a novel risk prediction model that improves on currently developed polygenic risk scores, which show clinical promise, but have methodological shortcomings that limit their accuracy.
Project Purpose(s)
- Disease Focused Research (alcohol misuse and associated diseases)
- Social / Behavioral
- Ancestry
Scientific Approaches
We are integrating data from the UK Biobank and the NIH All of Us Research Program. We will conduct genome-wide studies (GWAS) with REGENIE or Tractor to identify new variants associated with our alcohol use phenotypes. Potential clinical predictors will be identified from literature and machine learning (e.g., gradient boosting) analyses of separate lifestyle domains. For individual risk prediction, we will use neural networks to reduce the dimensionality of genetic data into interpretable independent latent factors, demonstrate that the latent factors recapitulate the alcohol-related GWAS associations that we identified, before finally using the latent factors in a multi-layer perceptron model to predict broad alcohol use phenotypes. We will also use neural networks to predict different sub-components of problematic alcohol use and negative alcohol use outcomes. Similar stepwise approaches will be performed for Mondrian cross-conformal prediction and gradient boosting classification.
Anticipated Findings
There is growing interest in using genetic risk scores in clinical practice, especially for decision making surrounding early intervention in high-risk individuals as well as triaging alcohol- and cardiometabolic-risk after an alcohol-related medical event for increased follow-up. Current genetic risk models can stratify individuals into large buckets of risk, but many of these methods discard useful information in favor of simple models or fail to generalize to real-world settings. A further limitation broadly plaguing the field of genomics research is the drastic overrepresentation of individuals of white European ancestry. By leveraging the diversity of the All of Us Research Program, we seek to build reproducible and equitable risk prediction models that may benefit a diverse patient population.
Demographic Categories of Interest
- Race / Ethnicity
- Gender Identity
- Geography
- Education Level
- Income Level
Data Set Used
Controlled TierResearch Team
Owner:
- Gordon Ye - Undergraduate Student, University of California, San Diego
- Eric Zorrilla - Early Career Tenure-track Researcher, Scripps Research
Collaborators:
- Sandra Sanchez-Roige - Early Career Tenure-track Researcher, University of California, San Diego
- Emily Zhu - Other, Scripps Research
- John Huber - Other, Scripps Research
- Jennifer Zhang - Project Personnel, All of Us Program Operational Use
- Eli Browne - Other, Scripps Research
- Bhadra Rupesh - Other, Scripps Research
- Paulina Ai - Other, Scripps Research
Problematic Alcohol Use and Related Psychiatric and Cardiometabolic Conditions 7
Scientific Questions Being Studied
The aim of this proposal is to identify new genetic and non-genetic factors that are leveraged through a machine-learning based approach to predict individual 1) risk for problematic alcohol use, 2) outcome from problematic alcohol use (e.g., recurrence of alcohol-related hospitalization or death), and 3) genetic and non-genetic alcohol risk factors that are shared with other psychiatric diseases and obesity-related cardiometabolic phenotypes. Prediction will utilize a combination of genetic, clinical, and lifestyle risk factors. Ultimately, we aim to identify not only individual predictors, but build a novel risk prediction model that improves on currently developed polygenic risk scores, which show clinical promise, but have methodological shortcomings that limit their accuracy.
Project Purpose(s)
- Disease Focused Research (alcohol misuse and associated diseases)
- Social / Behavioral
- Ancestry
Scientific Approaches
We are integrating data from the UK Biobank and the NIH All of Us Research Program. We will conduct genome-wide studies (GWAS) with REGENIE or Tractor to identify new variants associated with our alcohol use phenotypes. Potential clinical predictors will be identified from literature and machine learning (e.g., gradient boosting) analyses of separate lifestyle domains. For individual risk prediction, we will use neural networks to reduce the dimensionality of genetic data into interpretable independent latent factors, demonstrate that the latent factors recapitulate the alcohol-related GWAS associations that we identified, before finally using the latent factors in a multi-layer perceptron model to predict broad alcohol use phenotypes. We will also use neural networks to predict different sub-components of problematic alcohol use and negative alcohol use outcomes. Similar stepwise approaches will be performed for Mondrian cross-conformal prediction and gradient boosting classification.
Anticipated Findings
There is growing interest in using genetic risk scores in clinical practice, especially for decision making surrounding early intervention in high-risk individuals as well as triaging alcohol- and cardiometabolic-risk after an alcohol-related medical event for increased follow-up. Current genetic risk models can stratify individuals into large buckets of risk, but many of these methods discard useful information in favor of simple models or fail to generalize to real-world settings. A further limitation broadly plaguing the field of genomics research is the drastic overrepresentation of individuals of white European ancestry. By leveraging the diversity of the All of Us Research Program, we seek to build reproducible and equitable risk prediction models that may benefit a diverse patient population.
Demographic Categories of Interest
- Race / Ethnicity
- Gender Identity
- Geography
- Education Level
- Income Level
Data Set Used
Controlled TierResearch Team
Owner:
- Gordon Ye - Undergraduate Student, University of California, San Diego
- Eric Zorrilla - Early Career Tenure-track Researcher, Scripps Research
Collaborators:
- Sandra Sanchez-Roige - Early Career Tenure-track Researcher, University of California, San Diego
- Emily Zhu - Other, Scripps Research
- John Huber - Other, Scripps Research
- Jennifer Zhang - Project Personnel, All of Us Program Operational Use
- Eli Browne - Other, Scripps Research
- Bhadra Rupesh - Other, Scripps Research
- Paulina Ai - Other, Scripps Research
Request a Review of this Research Project
You can request that the All of Us Resource Access Board (RAB) review a research purpose description if you have concerns that this research project may stigmatize All of Us participants or violate the Data User Code of Conduct in some other way. To request a review, you must fill in a form, which you can access by selecting ‘request a review’ below.
