Yajing Sun
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Yajing Sun has a total of 24 co-authors. Here are the top co-authors by number of shared publications:
- Huan Song
Sichuan University
6 shared publications - Huazhen Yang
West China Biomedical Big Data Center, West China Hospital, Sichuan University, Chengdu, China; Med-X Center for Informatics, Sichuan University, Chengdu, China
6 shared publications - Yuanyuan Qu
Mental Health Center and West China Biomedical Big Data Center, West China Hospital, Sichuan University, Chengdu, China
6 shared publications - Yu Zeng
Med-X Center for Informatics, Sichuan University, Chengdu, China
5 shared publications - Wenwen Chen
Mental Health Center and West China Biomedical Big Data Center, West China Hospital, Sichuan University, Chengdu, China
4 shared publications - Can Hou
West China Biomedical Big Data Center, West China Hospital, Sichuan University, Chengdu, China
3 shared publications
Latest Publications
- 12 August 2024
- Published by Springer Nature in Phenomics
- Vol. 4 (5) , 415-429
Disease Modules Associated with Unfavorable Sleep Patterns and Their Genetic Determinants: A Prospective Cohort Study of the UK Biobank
Despite the established associations between sleep-related traits and major diseases, comprehensive assessment on affected disease modules and their genetic determinants is lacking. Using multiple correspondence analysis and the k-means clustering algorithm, 235,826 eligible participants were clustered into distinct unfavorable sleep patterns [short sleep duration (n = 10,073), snoring (22,419), insomnia (102,771), insomnia and snoring (62,909)] and favorable sleep pattern groups (37,654). The associations of unfavorable sleep patterns with 134 diseases were estimated using Cox regression models; and comorbidity network analyses were applied for disease module identification. Genetic determinants associated with each disease module were identified by genome-wide association studies (GWAS). During an average follow-up of 10.80 years, unfavorable sleep patterns featured by ‘short sleep duration’, ‘snoring’, ‘insomnia’, and ‘insomnia and snoring’ were associated with increased risk of 0, 9, 10, and 19 diseases, respectively. Furthermore, comorbidity network analyses categorized these affected diseases into three disease modules, characterized by predominant diseases related to digestive system, circulatory and endocrine systems (snoring-related patterns only), and musculoskeletal system (insomnia-related patterns only). Using the number of affected diseases, as an index of a person's susceptibility to each disease module [i.e., susceptible score (SS)], GWAS analyses identified five, one, and three significant loci associated with the residual SS of these aforementioned disease modules, respectively, which mapped to several potential biological pathways, including those related to hormone regulation and catecholamine uptake. In conclusion, individuals with unfavorable sleep patterns, particularly snoring and insomnia, had increased risk of multiple diseases. The identification of three major disease modules with their relevant genetic determinants may facilitate strategy development for precision prevention of future health decline.
- 12 August 2024
- Published by Springer Nature in Phenomics
- Vol. 4 (5) , 415-429
Disease Modules Associated with Unfavorable Sleep Patterns and Their Genetic Determinants: A Prospective Cohort Study of the UK Biobank
Despite the established associations between sleep-related traits and major diseases, comprehensive assessment on affected disease modules and their genetic determinants is lacking. Using multiple correspondence analysis and the k-means clustering algorithm, 235,826 eligible participants were clustered into distinct unfavorable sleep patterns [short sleep duration (n = 10,073), snoring (22,419), insomnia (102,771), insomnia and snoring (62,909)] and favorable sleep pattern groups (37,654). The associations of unfavorable sleep patterns with 134 diseases were estimated using Cox regression models; and comorbidity network analyses were applied for disease module identification. Genetic determinants associated with each disease module were identified by genome-wide association studies (GWAS). During an average follow-up of 10.80 years, unfavorable sleep patterns featured by ‘short sleep duration’, ‘snoring’, ‘insomnia’, and ‘insomnia and snoring’ were associated with increased risk of 0, 9, 10, and 19 diseases, respectively. Furthermore, comorbidity network analyses categorized these affected diseases into three disease modules, characterized by predominant diseases related to digestive system, circulatory and endocrine systems (snoring-related patterns only), and musculoskeletal system (insomnia-related patterns only). Using the number of affected diseases, as an index of a person's susceptibility to each disease module [i.e., susceptible score (SS)], GWAS analyses identified five, one, and three significant loci associated with the residual SS of these aforementioned disease modules, respectively, which mapped to several potential biological pathways, including those related to hormone regulation and catecholamine uptake. In conclusion, individuals with unfavorable sleep patterns, particularly snoring and insomnia, had increased risk of multiple diseases. The identification of three major disease modules with their relevant genetic determinants may facilitate strategy development for precision prevention of future health decline.
- Open Access
- 6 August 2024
- Published by Wolters Kluwer Health in Journal of the American Heart Association
- Vol. 13 (15) , e031280
Cardiovascular Disease, Genetic Susceptibility, and Risk of Psychiatric Disorders and Suicide Attempt: A Community‐Based Matched Cohort Study Based on the UK Biobank
Background: The associations between cardiovascular disease (CVD) and multiple psychiatric disorders and suicide attempt, and whether different genetic susceptibilities affect such links, have not been investigated clearly. Methods and Results: Based on the UK Biobank, we conducted a matched cohort study involving 63 923 patients who were first hospitalized with a CVD diagnosis between 1997 and 2020, and their 127 845 matched unexposed individuals. Cox models were used to examine the subsequent risk of psychiatric disorders and suicide attempt (ie, anxiety, depression, stress‐related disorder, substance misuse, psychotic disorder, and suicide behaviors) following CVD. We further performed stratified analyses by polygenic risk score for each studied psychiatric condition to detect the possible effects of genetic susceptibility on the observed associations. We found an increased risk of any psychiatric disorders and suicide attempt among CVD patients, compared with matched unexposed individuals, particularly within 1 year following the CVD (fully adjusted hazard ratio [HR] within 1 year, 1.83 [95% CI, 1.58–2.12]; HR after 1 year, 1.24 [95% CI, 1.16–1.32]). By subtype, the risk elevations existed for any psychiatric disorders and suicide attempt following most categories of CVDs. Analyses stratified by polygenic risk score revealed little impact of genetic predisposition to studied psychiatric conditions on these observed links. Conclusions: Patients hospitalized for CVD were at increased subsequent risk of multiple types of psychiatric disorders and suicide attempt, especially in the first year after hospitalization, irrespective of their genetic susceptibilities to studied psychiatric conditions, and these findings underscore the necessity of developing timely psychological interventions for this vulnerable population.
- 6 August 2024
- Published by Wolters Kluwer Health in Journal of the American Heart Association
- Vol. 13 (15) , e031280
Open AccessCardiovascular Disease, Genetic Susceptibility, and Risk of Psychiatric Disorders and Suicide Attempt: A Community‐Based Matched Cohort Study Based on the UK Biobank
Background: The associations between cardiovascular disease (CVD) and multiple psychiatric disorders and suicide attempt, and whether different genetic susceptibilities affect such links, have not been investigated clearly. Methods and Results: Based on the UK Biobank, we conducted a matched cohort study involving 63 923 patients who were first hospitalized with a CVD diagnosis between 1997 and 2020, and their 127 845 matched unexposed individuals. Cox models were used to examine the subsequent risk of psychiatric disorders and suicide attempt (ie, anxiety, depression, stress‐related disorder, substance misuse, psychotic disorder, and suicide behaviors) following CVD. We further performed stratified analyses by polygenic risk score for each studied psychiatric condition to detect the possible effects of genetic susceptibility on the observed associations. We found an increased risk of any psychiatric disorders and suicide attempt among CVD patients, compared with matched unexposed individuals, particularly within 1 year following the CVD (fully adjusted hazard ratio [HR] within 1 year, 1.83 [95% CI, 1.58–2.12]; HR after 1 year, 1.24 [95% CI, 1.16–1.32]). By subtype, the risk elevations existed for any psychiatric disorders and suicide attempt following most categories of CVDs. Analyses stratified by polygenic risk score revealed little impact of genetic predisposition to studied psychiatric conditions on these observed links. Conclusions: Patients hospitalized for CVD were at increased subsequent risk of multiple types of psychiatric disorders and suicide attempt, especially in the first year after hospitalization, irrespective of their genetic susceptibilities to studied psychiatric conditions, and these findings underscore the necessity of developing timely psychological interventions for this vulnerable population.
- Open Access
- 8 July 2024
- Published by Springer Nature in Phenomics
- Vol. 4 (4) , 327-338
Risk of Cardiovascular Disease Hospitalization After Common Psychiatric Disorders: Analyses of Disease Susceptibility and Progression Trajectory in the UK Biobank
Whether associations between psychiatric disorders and hospitalization for cardiovascular diseases (CVDs) can be modified by disease susceptibility and the temporal pattern of these associated CVDs remain unknown. In our study, we conducted a matched cohort study of the UK Biobank including 44,430 patients with common psychiatric disorders (anxiety, depression, and stress-related disorders) between 1997 and 2019, together with 222,150 sex-, Townsend deprivation index-, and birth year- individually matched unexposed individuals. The hazard ratios (HRs) for CVD hospitalization associated with a prior psychiatric disorder were derived from Cox models, adjusted for multiple confounders. We then stratified the analyses by self-reported family history of CVD and CVD polygenic risk score (PRS) calculated based on summary statistics of independent genome-wide association studies. We further conducted disease trajectory analysis and visualized the temporal pattern of CVDs after common psychiatric disorders. During a mean follow-up of 12.28 years, we observed an elevated risk of CVD hospitalization among patients with psychiatric disorders, compared with matched unexposed individuals (hazard ratios [HRs] = 1.20, 95% confidence interval [CI]: 1.18–1.23), especially during the first six months of follow-up (1.72 [1.55–1.91]). The stratification analyses by family history of CVD and by CVD PRS obtained similar estimates between subgroups with different susceptibilities to CVD. We conducted trajectory analysis to visualize the temporal pattern of CVDs after common psychiatric disorders, identifying primary hypertension, acute myocardial infarction, and stroke as three main intermediate steps leading to further increased risk of other CVDs. In conclusion, the association between common psychiatric disorders and subsequent CVD hospitalization is not modified by predisposition to CVD. Hypertension, acute myocardial infarction, and stroke are three initial CVDs linking psychiatric disorders to other CVD sequelae, highlighting a need of timely intervention on these targets to prevent further CVD sequelae among all individuals with common psychiatric disorders.
- 8 July 2024
- Published by Springer Nature in Phenomics
- Vol. 4 (4) , 327-338
Open AccessRisk of Cardiovascular Disease Hospitalization After Common Psychiatric Disorders: Analyses of Disease Susceptibility and Progression Trajectory in the UK Biobank
Whether associations between psychiatric disorders and hospitalization for cardiovascular diseases (CVDs) can be modified by disease susceptibility and the temporal pattern of these associated CVDs remain unknown. In our study, we conducted a matched cohort study of the UK Biobank including 44,430 patients with common psychiatric disorders (anxiety, depression, and stress-related disorders) between 1997 and 2019, together with 222,150 sex-, Townsend deprivation index-, and birth year- individually matched unexposed individuals. The hazard ratios (HRs) for CVD hospitalization associated with a prior psychiatric disorder were derived from Cox models, adjusted for multiple confounders. We then stratified the analyses by self-reported family history of CVD and CVD polygenic risk score (PRS) calculated based on summary statistics of independent genome-wide association studies. We further conducted disease trajectory analysis and visualized the temporal pattern of CVDs after common psychiatric disorders. During a mean follow-up of 12.28 years, we observed an elevated risk of CVD hospitalization among patients with psychiatric disorders, compared with matched unexposed individuals (hazard ratios [HRs] = 1.20, 95% confidence interval [CI]: 1.18–1.23), especially during the first six months of follow-up (1.72 [1.55–1.91]). The stratification analyses by family history of CVD and by CVD PRS obtained similar estimates between subgroups with different susceptibilities to CVD. We conducted trajectory analysis to visualize the temporal pattern of CVDs after common psychiatric disorders, identifying primary hypertension, acute myocardial infarction, and stroke as three main intermediate steps leading to further increased risk of other CVDs. In conclusion, the association between common psychiatric disorders and subsequent CVD hospitalization is not modified by predisposition to CVD. Hypertension, acute myocardial infarction, and stroke are three initial CVDs linking psychiatric disorders to other CVD sequelae, highlighting a need of timely intervention on these targets to prevent further CVD sequelae among all individuals with common psychiatric disorders.
- Open Access
- 3 August 2023
- Published by Wiley in Depression and Anxiety
- Vol. 2023, 1-11
Genetic Determinants of the Association between Osteoarthritis and Psychiatric Disorders
Background. The associations between hip/knee osteoarthritis (OA) and various psychiatry disorders, as well as the underlying genetic determinants, remain unclear. Methods. Based on the community-based prospective data of UK Biobank, we constructed two matched cohorts to assess the bidirectional associations between OA and five common psychiatry disorders. Then, we further examined the existence of overall genetic association for those disease pairs with demonstrated phenotypic association through polygenic risk score (PRS) prediction using individual-level genotyping data of UK Biobank and LD score regression (LDSC) analysis utilizing publicly available GWAS summary statistics. Last, also based on GWAS summary statistics, we performed enrichment analyses to pinpoint specific genetic determinants that might contribute to the observed overall genetic shares. Results. The phenotypic analyses revealed an elevated risk of hip/knee OA among individuals with any psychiatric disorders, compared to their matched unexposed individuals (hazard , 95% confidence interval (CI): 1.57-1.68), and vice versa (HR =1.93, 95% CI: 1.83-2.04). We further observed positive associations of knee OA with depression and stress-related disorder in the PRS analyses, which corroborated with the results of LDSC analyses ( for (95% CI: 0.15-0.25) and for (95% CI: 0.19-0.40), respectively). Using GWAS summary statistics, we identified several shared genes and pathways, particularly the biological process related to HDAC histones, between knee OA and depression/stress-related disorder. Conclusions. Our study demonstrated a bidirectional association between OA and multiple psychiatric disorders, and the findings of shared genetic architectures between knee OA and depression/stress-related disorder provided possible targets for further mechanistic exploration and intervention development.
- 3 August 2023
- Published by Wiley in Depression and Anxiety
- Vol. 2023, 1-11
Open AccessGenetic Determinants of the Association between Osteoarthritis and Psychiatric Disorders
Background. The associations between hip/knee osteoarthritis (OA) and various psychiatry disorders, as well as the underlying genetic determinants, remain unclear. Methods. Based on the community-based prospective data of UK Biobank, we constructed two matched cohorts to assess the bidirectional associations between OA and five common psychiatry disorders. Then, we further examined the existence of overall genetic association for those disease pairs with demonstrated phenotypic association through polygenic risk score (PRS) prediction using individual-level genotyping data of UK Biobank and LD score regression (LDSC) analysis utilizing publicly available GWAS summary statistics. Last, also based on GWAS summary statistics, we performed enrichment analyses to pinpoint specific genetic determinants that might contribute to the observed overall genetic shares. Results. The phenotypic analyses revealed an elevated risk of hip/knee OA among individuals with any psychiatric disorders, compared to their matched unexposed individuals (hazard , 95% confidence interval (CI): 1.57-1.68), and vice versa (HR =1.93, 95% CI: 1.83-2.04). We further observed positive associations of knee OA with depression and stress-related disorder in the PRS analyses, which corroborated with the results of LDSC analyses ( for (95% CI: 0.15-0.25) and for (95% CI: 0.19-0.40), respectively). Using GWAS summary statistics, we identified several shared genes and pathways, particularly the biological process related to HDAC histones, between knee OA and depression/stress-related disorder. Conclusions. Our study demonstrated a bidirectional association between OA and multiple psychiatric disorders, and the findings of shared genetic architectures between knee OA and depression/stress-related disorder provided possible targets for further mechanistic exploration and intervention development.
- Open Access
- 20 February 2023
- Published by JMIR Publications Inc. in JMIR Public Health and Surveillance
- Vol. 9 (1) , e43419
Prediction of Suicidal Behaviors in the Middle-aged Population: Machine Learning Analyses of UK Biobank
Background: Suicidal behaviors, including suicide deaths and attempts, are major public health concerns. However, previous suicide models required a huge amount of input features, resulting in limited applicability in clinical practice. Objective: We aimed to construct applicable models (ie, with limited features) for short- and long-term suicidal behavior prediction. We further validated these models among individuals with different genetic risks of suicide. Methods: Based on the prospective cohort of UK Biobank, we included 223 (0.06%) eligible cases of suicide attempts or deaths, according to hospital inpatient or death register data within 1 year from baseline and randomly selected 4460 (1.18%) controls (1:20) without such records. We similarly identified 833 (0.22%) cases of suicidal behaviors 1 to 6 years from baseline and 16,660 (4.42%) corresponding controls. Based on 143 input features, mainly including sociodemographic, environmental, and psychosocial factors; medical history; and polygenic risk scores (PRS) for suicidality, we applied a bagged balanced light gradient-boosting machine (LightGBM) with stratified 10-fold cross-validation and grid-search to construct the full prediction models for suicide attempts or deaths within 1 year or between 1 and 6 years. The Shapley Additive Explanations (SHAP) approach was used to quantify the importance of input features, and the top 20 features with the highest SHAP values were selected to train the applicable models. The external validity of the established models was assessed among 50,310 individuals who participated in UK Biobank repeated assessments both overall and by the level of PRS for suicidality. Results: Individuals with suicidal behaviors were on average 56 years old, with equal sex distribution. The application of these full models in the external validation data set demonstrated good model performance, with the area under the receiver operating characteristic (AUROC) curves of 0.919 and 0.892 within 1 year and between 1 and 6 years, respectively. Importantly, the applicable models with the top 20 most important features showed comparable external-validated performance (AUROC curves of 0.901 and 0.885) as the full models, based on which we found that individuals in the top quintile of predicted risk accounted for 91.7% (n=11) and 80.7% (n=25) of all suicidality cases within 1 year and during 1 to 6 years, respectively. We further obtained comparable prediction accuracy when applying these models to subpopulations with different genetic susceptibilities to suicidality. For example, for the 1-year risk prediction, the AUROC curves were 0.907 and 0.885 for the high (>2nd tertile of PRS) and low (<1st) genetic susceptibilities groups, respectively. Conclusions: We established applicable machine learning–based models for predicting both the short- and long-term risk of suicidality with high accuracy across populations of varying genetic risk for suicide, highlighting a cost-effective method of identifying individuals with a high risk of suicidality.
- 20 February 2023
- Published by JMIR Publications Inc. in JMIR Public Health and Surveillance
- Vol. 9 (1) , e43419
Open AccessPrediction of Suicidal Behaviors in the Middle-aged Population: Machine Learning Analyses of UK Biobank
Background: Suicidal behaviors, including suicide deaths and attempts, are major public health concerns. However, previous suicide models required a huge amount of input features, resulting in limited applicability in clinical practice. Objective: We aimed to construct applicable models (ie, with limited features) for short- and long-term suicidal behavior prediction. We further validated these models among individuals with different genetic risks of suicide. Methods: Based on the prospective cohort of UK Biobank, we included 223 (0.06%) eligible cases of suicide attempts or deaths, according to hospital inpatient or death register data within 1 year from baseline and randomly selected 4460 (1.18%) controls (1:20) without such records. We similarly identified 833 (0.22%) cases of suicidal behaviors 1 to 6 years from baseline and 16,660 (4.42%) corresponding controls. Based on 143 input features, mainly including sociodemographic, environmental, and psychosocial factors; medical history; and polygenic risk scores (PRS) for suicidality, we applied a bagged balanced light gradient-boosting machine (LightGBM) with stratified 10-fold cross-validation and grid-search to construct the full prediction models for suicide attempts or deaths within 1 year or between 1 and 6 years. The Shapley Additive Explanations (SHAP) approach was used to quantify the importance of input features, and the top 20 features with the highest SHAP values were selected to train the applicable models. The external validity of the established models was assessed among 50,310 individuals who participated in UK Biobank repeated assessments both overall and by the level of PRS for suicidality. Results: Individuals with suicidal behaviors were on average 56 years old, with equal sex distribution. The application of these full models in the external validation data set demonstrated good model performance, with the area under the receiver operating characteristic (AUROC) curves of 0.919 and 0.892 within 1 year and between 1 and 6 years, respectively. Importantly, the applicable models with the top 20 most important features showed comparable external-validated performance (AUROC curves of 0.901 and 0.885) as the full models, based on which we found that individuals in the top quintile of predicted risk accounted for 91.7% (n=11) and 80.7% (n=25) of all suicidality cases within 1 year and during 1 to 6 years, respectively. We further obtained comparable prediction accuracy when applying these models to subpopulations with different genetic susceptibilities to suicidality. For example, for the 1-year risk prediction, the AUROC curves were 0.907 and 0.885 for the high (>2nd tertile of PRS) and low (<1st) genetic susceptibilities groups, respectively. Conclusions: We established applicable machine learning–based models for predicting both the short- and long-term risk of suicidality with high accuracy across populations of varying genetic risk for suicide, highlighting a cost-effective method of identifying individuals with a high risk of suicidality.
- Open Access
- 11 October 2022
- Published by JMIR Publications Inc. in JMIR Preprints
Prediction of Suicidal Behaviors in the Middle-aged Population: Machine Learning Analyses of UK Biobank (Preprint)
BACKGROUND Suicidal behaviors, including suicide deaths and attempts, are major public health concerns. However, previous suicide models required a huge amount of input features, resulting in limited applicability in clinical practice. OBJECTIVE We aimed to construct applicable models (ie, with limited features) for short- and long-term suicidal behavior prediction. We further validated these models among individuals with different genetic risks of suicide. METHODS Based on the prospective cohort of UK Biobank, we included 223 (0.06%) eligible cases of suicide attempts or deaths, according to hospital inpatient or death register data within 1 year from baseline and randomly selected 4460 (1.18%) controls (1:20) without such records. We similarly identified 833 (0.22%) cases of suicidal behaviors 1 to 6 years from baseline and 16,660 (4.42%) corresponding controls. Based on 143 input features, mainly including sociodemographic, environmental, and psychosocial factors; medical history; and polygenic risk scores (PRS) for suicidality, we applied a bagged balanced light gradient-boosting machine (LightGBM) with stratified 10-fold cross-validation and grid-search to construct the full prediction models for suicide attempts or deaths within 1 year or between 1 and 6 years. The Shapley Additive Explanations (SHAP) approach was used to quantify the importance of input features, and the top 20 features with the highest SHAP values were selected to train the applicable models. The external validity of the established models was assessed among 50,310 individuals who participated in UK Biobank repeated assessments both overall and by the level of PRS for suicidality. RESULTS Individuals with suicidal behaviors were on average 56 years old, with equal sex distribution. The application of these full models in the external validation data set demonstrated good model performance, with the area under the receiver operating characteristic (AUROC) curves of 0.919 and 0.892 within 1 year and between 1 and 6 years, respectively. Importantly, the applicable models with the top 20 most important features showed comparable external-validated performance (AUROC curves of 0.901 and 0.885) as the full models, based on which we found that individuals in the top quintile of predicted risk accounted for 91.7% (n=11) and 80.7% (n=25) of all suicidality cases within 1 year and during 1 to 6 years, respectively. We further obtained comparable prediction accuracy when applying these models to subpopulations with different genetic susceptibilities to suicidality. For example, for the 1-year risk prediction, the AUROC curves were 0.907 and 0.885 for the high (>2nd tertile of PRS) and low (<1st) genetic susceptibilities groups, respectively. CONCLUSIONS We established applicable machine learning–based models for predicting both the short- and long-term risk of suicidality with high accuracy across populations of varying genetic risk for suicide, highlighting a cost-effective method of identifying individuals with a high risk of suicidality.
- 11 October 2022
- Published by JMIR Publications Inc. in JMIR Preprints
Open AccessPrediction of Suicidal Behaviors in the Middle-aged Population: Machine Learning Analyses of UK Biobank (Preprint)
BACKGROUND Suicidal behaviors, including suicide deaths and attempts, are major public health concerns. However, previous suicide models required a huge amount of input features, resulting in limited applicability in clinical practice. OBJECTIVE We aimed to construct applicable models (ie, with limited features) for short- and long-term suicidal behavior prediction. We further validated these models among individuals with different genetic risks of suicide. METHODS Based on the prospective cohort of UK Biobank, we included 223 (0.06%) eligible cases of suicide attempts or deaths, according to hospital inpatient or death register data within 1 year from baseline and randomly selected 4460 (1.18%) controls (1:20) without such records. We similarly identified 833 (0.22%) cases of suicidal behaviors 1 to 6 years from baseline and 16,660 (4.42%) corresponding controls. Based on 143 input features, mainly including sociodemographic, environmental, and psychosocial factors; medical history; and polygenic risk scores (PRS) for suicidality, we applied a bagged balanced light gradient-boosting machine (LightGBM) with stratified 10-fold cross-validation and grid-search to construct the full prediction models for suicide attempts or deaths within 1 year or between 1 and 6 years. The Shapley Additive Explanations (SHAP) approach was used to quantify the importance of input features, and the top 20 features with the highest SHAP values were selected to train the applicable models. The external validity of the established models was assessed among 50,310 individuals who participated in UK Biobank repeated assessments both overall and by the level of PRS for suicidality. RESULTS Individuals with suicidal behaviors were on average 56 years old, with equal sex distribution. The application of these full models in the external validation data set demonstrated good model performance, with the area under the receiver operating characteristic (AUROC) curves of 0.919 and 0.892 within 1 year and between 1 and 6 years, respectively. Importantly, the applicable models with the top 20 most important features showed comparable external-validated performance (AUROC curves of 0.901 and 0.885) as the full models, based on which we found that individuals in the top quintile of predicted risk accounted for 91.7% (n=11) and 80.7% (n=25) of all suicidality cases within 1 year and during 1 to 6 years, respectively. We further obtained comparable prediction accuracy when applying these models to subpopulations with different genetic susceptibilities to suicidality. For example, for the 1-year risk prediction, the AUROC curves were 0.907 and 0.885 for the high (>2nd tertile of PRS) and low (<1st) genetic susceptibilities groups, respectively. CONCLUSIONS We established applicable machine learning–based models for predicting both the short- and long-term risk of suicidality with high accuracy across populations of varying genetic risk for suicide, highlighting a cost-effective method of identifying individuals with a high risk of suicidality.