Hongsheng Gui
Early Career Tenure-track Researcher, Henry Ford Health System
7 active projects
Genomic association and risk prediction of mental health conditions Phase3
Scientific Questions Being Studied
Genome-wide association studies of mental health conditions have identified variants that contribute to a small portion of the total disorder variance. While these data have been useful in identifying novel genetic associations, in large part these data have not been used predict risk of disease. Our hope is to leverage the All of Us data to estimate disease effect estimates for multiple mental health conditions in diverse populations. Specific questions include: (1) what are the most significant non-genetic and genetic factors in association with serious mental illnesses? (2) How can we combine both clinical risk factors and genomic risk scores to predict mental health conditions? We hypothesize that a risk model incorporating both non-genetic and genetic factors will have better power to explain and predict mental health conditions. This study will develop and evaluate end-to-end predictive models for mental health phenotype in diverse groups. Phase-3 will focus on HLA region.
Project Purpose(s)
- Disease Focused Research (Serious mental illness and suicide)
- Population Health
- Social / Behavioral
- Methods Development
- Ancestry
Scientific Approaches
We will cover all eligible participants with different background in All of Us. First, self-reported surveys (e.g., lifestyle and medical history) and electronic health records (EHR) will be used to determine serious mental illness cases and their matched controls. Exposure variables will be assessed and harmonized across resources; and pre-processed for missing value and outliers. Second, disease risk factors will then be interrogated by regularized regression (non-genetic) and genome-wide association test (genetic). Third, clinical risk score and genomic risk score will be constructed by weighted sum of significant risk factors retained, and then combined by logistic regression and random forest. Fourth, a set of candidate risk models will be validated and then optimized in prospective sample, via C statistics and other diagnostic metrics. The whole framework will also be repeated in stratified samples so as to tune the parameters for specific groups (e.g., age, sex, and race).
Anticipated Findings
Findings from this project would identify novel genetic biomarkers and establish a predictive model to classify individuals at different risk for developing serious mental illness. We expect to find: (1) significant clinical/social factors and biomarkers that are associated with focused diseases; (2) polygenic risk scores tailored for overall population and different subgroups; (3) risk model comprised of non-genetic risk factors and polygenic risk scores. We also anticipate some risk factors may present different effect sizes across populations. There is a need to revise risk model in general population to increase its performance in specific group. However, those difference observed across populations may be caused by social determinants, instead of biology; and some of them may be not well covered in All of Us. We will mitigate the potential stigmatization risk of our models in future publications and presentations by providing more discussion on its benefit and limitation.
Demographic Categories of Interest
This study will not center on underrepresented populations.
Data Set Used
Controlled TierResearch Team
Owner:
- Hongsheng Gui - Early Career Tenure-track Researcher, Henry Ford Health System
Collaborators:
- Ze Meng - Early Career Tenure-track Researcher, Henry Ford Health System
Genomic association and risk prediction of mental health conditions Phase2
Scientific Questions Being Studied
Genome-wide association studies of mental health conditions (e.g., schizophrenia and suicide) have identified variants that contribute to a small portion of the total disorder variance. While these data have been useful in identifying novel genetic associations, in large part these data have not been used predict risk of disease.. Our hope is to leverage the All of Us data to estimate disease effect estimates for multiple mental health conditions in diverse populations. Specific questions include: (1) what are the most significant non-genetic and genetic factors in association with serious mental illnesses? (2) How can we combine both clinical risk factors and genomic risk scores to predict mental health conditions? We hypothesize that a risk model incorporating both non-genetic and genetic factors will have better power to explain and predict mental health conditions. This study will develop and evaluate end-to-end predictive models for mental health phenotype in diverse groups.
Project Purpose(s)
- Disease Focused Research (Serious mental illness and suicide)
- Population Health
- Social / Behavioral
- Methods Development
- Ancestry
Scientific Approaches
We will cover all eligible participants with different background in All of Us. First, self-reported surveys (e.g., lifestyle and medical history) and electronic health records (EHR) will be used to determine serious mental illness cases and their matched controls. Exposure variables will be assessed and harmonized across resources; and pre-processed for missing value and outliers. Second, disease risk factors will then be interrogated by regularized regression (non-genetic) and genome-wide association test (genetic). Third, clinical risk score and genomic risk score will be constructed by weighted sum of significant risk factors retained, and then combined by logistic regression and random forest. Fourth, a set of candidate risk models will be validated and then optimized in prospective sample, via C statistics and other diagnostic metrics. The whole framework will also be repeated in stratified samples so as to tune the parameters for specific groups (e.g., age, sex, and race).
Anticipated Findings
Findings from this project would identify novel genetic biomarkers and establish a predictive model to classify individuals at different risk for developing serious mental illness. We expect to find: (1) significant clinical/social factors and biomarkers that are associated with focused diseases; (2) polygenic risk scores tailored for overall population and different subgroups; (3) risk model comprised of non-genetic risk factors and polygenic risk scores. We also anticipate some risk factors may present different effect sizes across populations. There is a need to revise risk model in general population to increase its performance in specific group. However, those difference observed across populations may be caused by social determinants, instead of biology; and some of them may be not well covered in All of Us. We will mitigate the potential stigmatization risk of our models in future publications and presentations by providing more discussion on its benefit and limitation.
Demographic Categories of Interest
This study will not center on underrepresented populations.
Data Set Used
Controlled TierResearch Team
Owner:
- Hongsheng Gui - Early Career Tenure-track Researcher, Henry Ford Health System
Collaborators:
- Ze Meng - Early Career Tenure-track Researcher, Henry Ford Health System
Investigation on Suicide in the COVID-19 pandemic Phase 2
Scientific Questions Being Studied
Outbreak of Coronavirus Disease 2019 (COVID-19) has caused a new psychological burden. Patient Health Questionnaire (PHQ-9) can be used to evaluate mood status, monitor changes in signs/symptoms of suicide, and assess suicidal ideation. Here our study aims to describe the basis statistics of PHQ-9 scores and its inferred depression or suicide risk for all participants in All of US COPE survey.
Project Purpose(s)
- Disease Focused Research (Suicidal behaviors/thoughts)
Scientific Approaches
PHQ-9 questions and answers will be retrieved for participants involved in six different time points. Response to each question will be converted to numeric scores (0, 1, 2, 3), and then summed up to derive the PHQ-9 total score. Participants missing any individual score were not included in this study. Binary status of suicidal ideation will be defined using item-9 answer (i.e., yes for >0). Distributions of PHQ-9 total score, suicidal ideation status at each time session will be reported by descriptive statistics stratified by age, sex, and ancestry. Their changes across different time sessions were tested by Kruskal-Wallis (KW) test, Friedman test, or chi-square test. Multivariable analyses are going to be conducted by generalized linear mixed models.
Anticipated Findings
We anticipate the descriptive statistics, pairwise correlations, and multivariable model fitting results will tell us the trajectories of suicidal thoughts and behaviors in the COVID-19 pandemic. They will not only help to verify the known relationship between suicide and gender or age, but also will provide new evidence of mood status changes along COVID-19 pandemic at both population and individual level.
Demographic Categories of Interest
This study will not center on underrepresented populations.
Data Set Used
Registered TierResearch Team
Owner:
- Hongsheng Gui - Early Career Tenure-track Researcher, Henry Ford Health System
Collaborators:
- Hsueh-Han Yeh - Research Associate, Henry Ford Health System
Genomic risk prediction of opioid use disorder in diverse populations
Scientific Questions Being Studied
Opioid epidemic is an on-going crisis in the United States, but better understanding of opioid use and use disorder (OUD) have to consider both biological and social factors. There is still a lack of actionable biomarkers or genomic risk scores for OUD screening or treatment. Specific questions we will ask: (1) what are the most significant risk factors that associate with OUD onset in All of Us retrospective cohort? (2) How can we combine clinical/social risk factors and genetic/genomic risk factors together in the risk prediction framework for OUD prospective development across different population groups? We hypothesize that a risk model incorporating both non-genetic and genetic factors will have better power to explain and predict OUD. This study will develop and evaluate end-to-end predictive models for OUD phenotype in diverse groups, will then suggest new strategy for future OUD screening and prevention in a precision way.
Project Purpose(s)
- Disease Focused Research (Opioid use disorder and related mental comorbidities)
- Population Health
- Social / Behavioral
- Methods Development
- Ancestry
Scientific Approaches
We will include all participants in All of Us project. First, self-reported surveys (e.g., lifestyle and medical history) and electronic health records (EHR) will be used to determine OUD cases, opioid-exposed controls, and general controls. Exposure variables will be assessed and harmonized across resources; and pre-processed for missing value and outliers. Second, OUD risk factors will then be interrogated by regularized regression (non-genetic) and genome-wide association test (genetic). Third, clinical risk score and genomic risk score will be constructed by weighted sum of significant risk factors retained, and then combined by logistic regression and random forest. Fourth, a set of candidate risk models will be validated and then optimized in prospective sample, via C statistics and other diagnostic metrics. The whole framework will also be repeated in stratified samples so as to tune the parameters for specific groups (e.g., age, sex, and race).
Anticipated Findings
Findings from this project would establish a predictive model to identify individuals at low, moderate, and high risk for developing OUD. We expect to find: (1) significant clinical/social factors and biomarkers that are associated with opioid use initiation and OUD status; (2) polygenic risk scores tailored for overall population and different subgroups; (3) risk model comprised of non-genetic risk factors and polygenic risk scores.
We also anticipate some risk factors may present different effect sizes across populations. There is also a need to revise risk model in general population to increase its performance in specific group. However, those difference observed across populations may be caused by social determinants, instead of biology; and some of them may be not well covered in All of Us project. We will mitigate the potential stigmatization risk of our models in future publications and presentations by providing more discussion on its benefit and limitation.
Demographic Categories of Interest
This study will not center on underrepresented populations.
Data Set Used
Controlled TierResearch Team
Owner:
- Hongsheng Gui - Early Career Tenure-track Researcher, Henry Ford Health System
Collaborators:
- Ze Meng - Early Career Tenure-track Researcher, Henry Ford Health System
Investigation on Suicide in the COVID-19 pandemic
Scientific Questions Being Studied
Outbreak of Coronavirus Disease 2019 (COVID-19) has caused a new psychological burden. Patient Health Questionnaire (PHQ-9) can be used to evaluate mood status, monitor changes in signs/symptoms of depression, and assess suicidal ideation. Here our study aims to describe the basis statistics of PHQ-9 scores and its inferred depression or suicide risk for all participants in All of US COPE survey.
Project Purpose(s)
- Disease Focused Research (Depression, Suicidal behaviors)
Scientific Approaches
PHQ-9 questions and answers will be retrieved for participants involved in six different time points. Response to each question will be converted to numeric scores (0, 1, 2, 3), and then summed up to derive the PHQ-9 total score. Participants missing any individual score were not included in this study. Depression levels will be categorized into 5 different ordinals according to their PHQ-9 total scores. Binary status of suicidal ideation will be defined using item-9 answer (i.e., yes for >0). Distributions of PHQ-9 total score, depression severity, and suicidal ideation status at each time session will be reported by descriptive statistics stratified by age, sex, and ancestry. Their changes across different time sessions were tested by Kruskal-Wallis (KW) test, Friedman test, or chi-square test. Multivariable analyses are going to be conducted by generalized linear mixed models.
Anticipated Findings
We anticipate the descriptive statistics, pairwise correlations, and multivariable model fitting results will tell us the trajectories of major depression disorder and suicidal behaviors in the COVID-19 pandemic. They will not only help to verify the known relationship between depression and gender or age, but also will provide new evidence of mood status changes along COVID-19 pandemic at both population and individual level.
Demographic Categories of Interest
- Race / Ethnicity
- Age
- Geography
- Education Level
Data Set Used
Registered TierRelationship between depression symptoms and COVID-19
Scientific Questions Being Studied
This will be an epidemiology study that aims to interrogate the relationship between major depressive disorder (MDD) and COVID-19 symptoms. Depression is the most common mental health condition and affects more than 300 million people around the world. It is also a leading cause of suicidal death. Coronavirus disease 2019 (COVID-19) has affected >200 million people and caused >4.5 million death since its outbreak in late 2019. It is known that people with mental health problems may have worse physical health, lower immunity, and higher susceptibility to infection. We hypothesize that individuals with MDD before the COVID-19 outbreak are more easily affected by COVID-19 than those without MDD; and vice versa, individuals affected with COVID-19 symptoms are more likely to develop or deteriorate depression symptoms than those not affected. A clear relationship between MDD and COVID-19 status will help to establish better screening and prevention strategy for both two health conditions.
Project Purpose(s)
- Disease Focused Research (major depressive disorder)
Scientific Approaches
Both electronic health records and survey data from All of Us program will be used to examine the relationship between MDD and COVID-19. The definition of MDD will come from electronic health record (ICD9/10) and PHQ-9 questionnaire from COPE survey. COVID-19 symptoms will also come from COPE survey. We will describe the MDD population statistics before and after COVID-19 pandemic outbreak. COPE surveys from 3 different time points (May, June, July/August 2020) will be leveraged to perform combined analysis and time-varying analysis. Environmental variables (e.g., demographic, socioeconomics status, and disease comorbidities) will be extracted as confounder or covariates in our statistical modeling. Univariable analysis will be carried out by chi-square test, log-rank test, and correlation. Multivariable analysis will be performed by logistic regression, cox regression, and generalized linear mixed models.
Anticipated Findings
Expected outcomes will include: 1) descriptive statistic of MDD and COVID-19 by different stratum in All of Us populations; 2) correlation and association pattern between MDD symptoms and COVID-19 status or monitoring strategies; 3) risk stratification model for COVID-19 onset and development that include MDD symptoms as predictors.
With all above expected outcomes, our findings will not only provide additional insights on how to better manage MDD by controlling COVID-19 affection, but also have better strategy of preventing COVID-19 affection by screening high risk MDD patients for earlier intervention and vaccination.
Demographic Categories of Interest
This study will not center on underrepresented populations.
Data Set Used
Registered TierGenomic risk prediction of substance dependence in diverse populations
Scientific Questions Being Studied
Genome-wide association studies of substance addiction and dependence have identified variants that contribute to a small portion of the total disorder variance. While these data have been useful in identifying novel genetic associations, in large part these data have not been used predict risk of disease. Recently, groups have begun leverage the contribution of variants that do not reach the level of genome-wide significance to improve the prediction of complex diseases. Unfortunately, these measures have not been developed or applied to non-European populations. Our hope is leverage the All of Us data to estimate disease effect estimates for substance use in diverse populations so as to be able to appropriately apply a ancestry-tailored risk component to polygenic risk scores involving admixed populations (i.e., populations admixed with multiple continental including European ancestry).
Project Purpose(s)
- Disease Focused Research (substance dependence)
- Population Health
- Ancestry
Scientific Approaches
We will cover both non-Hispanic white population and black populations in this study. Three types of phenotypes for substance use will be investigated: 1) alcohol use; 2) smoking; 3) Opioid use.
Aim 1: perform GWAS, create and validate PRS in All-of-US dataset
A standard GWAS will be conducted. PRS will be created by a few different variant sets and approaches. Approaches include BSLMM, LDPred, PRSice and Lasso penalized regression. These PRSs will be evaluated in the second half of All-of-Us cohort. The one with highest Area under ROC curve, or other discrimination metrics will be selected.
Aim 2: test and evaluate PRS in internal testing dataset from a health system
We will test PRS in our study populations consisting of European Americans (EA) and African Americans (AA) individuals. Testing risk scores in African Americans will involve applying the appropriate risk estimate for African and European ancestry to each individual based on their admixture and local ancestry.
Anticipated Findings
1. Ancestry-shared and ancestry-specific susceptibility loci for alcohol, smoking and opioid use and dependence.
2. Optimized PRS for each disease or phenotype in European population and African-descent populations.
3. Clinically actionable risk prediction model incorporating PRS, family history, and environmental factors will be built to help patient treatment and risk management.
4. This will greatly improve risk prediction in non-white, minority populations for substance use.
Demographic Categories of Interest
- Race / Ethnicity
- Age
Data Set Used
Registered TierRequest 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.
