Towards understanding and predicting suicidality in women

biomarkers and clinical risk assessment

D. F. Levey, E. M. Niculescu, Helen Le-Niculescu, H. L. Dainton, P. L. Phalen, T. B. Ladd, H. Weber, E. Belanger, D. L. Graham, F. N. Khan, N. P. Vanipenta, E. C. Stage, A. Ballew, M. Yard, T. Gelbart, Anantha Shekhar, N. J. Schork, S. M. Kurian, George Sandusky, D. R. Salomon & 1 others Alexander Niculescu

Research output: Contribution to journalArticle

24 Citations (Scopus)

Abstract

Women are under-represented in research on suicidality to date. Although women have a lower rate of suicide completion than men, due in part to the less-violent methods used, they have a higher rate of suicide attempts. Our group has previously identified genomic (blood gene expression biomarkers) and clinical information (apps) predictors for suicidality in men. We now describe pilot studies in women. We used a powerful within-participant discovery approach to identify genes that change in expression between no suicidal ideation (no SI) and high suicidal ideation (high SI) states (n=12 participants out of a cohort of 51 women psychiatric participants followed longitudinally, with diagnoses of bipolar disorder, depression, schizoaffective disorder and schizophrenia). We then used a Convergent Functional Genomics (CFG) approach to prioritize the candidate biomarkers identified in the discovery step by using all the prior evidence in the field. Next, we validated for suicidal behavior the top-ranked biomarkers for SI, in a demographically matched cohort of women suicide completers from the coroner’s office (n=6), by assessing which markers were stepwise changed from no SI to high SI to suicide completers. We then tested the 50 biomarkers that survived Bonferroni correction in the validation step, as well as top increased and decreased biomarkers from the discovery and prioritization steps, in a completely independent test cohort of women psychiatric disorder participants for prediction of SI (n=33) and in a future follow-up cohort of psychiatric disorder participants for prediction of psychiatric hospitalizations due to suicidality (n=24). Additionally, we examined how two clinical instruments in the form of apps, Convergent Functional Information for Suicidality (CFI-S) and Simplified Affective State Scale (SASS), previously tested in men, perform in women. The top CFI-S item distinguishing high SI from no SI states was the chronic stress of social isolation. We then showed how the clinical information apps combined with the 50 validated biomarkers into a broad predictor (UP-Suicide), our apriori primary end point, predicts suicidality in women. UP-Suicide had a receiver-operating characteristic (ROC) area under the curve (AUC) of 82% for predicting SI and an AUC of 78% for predicting future hospitalizations for suicidality. Some of the individual components of the UP-Suicide showed even better results. SASS had an AUC of 81% for predicting SI, CFI-S had an AUC of 84% and the combination of the two apps had an AUC of 87%. The top biomarker from our sequential discovery, prioritization and validation steps, BCL2, predicted future hospitalizations due to suicidality with an AUC of 89%, and the panel of 50 validated biomarkers (BioM-50) predicted future hospitalizations due to suicidality with an AUC of 94%. The best overall single blood biomarker for predictions was PIK3C3 with an AUC of 65% for SI and an AUC of 90% for future hospitalizations. Finally, we sought to understand the biology of the biomarkers. BCL2 and GSK3B, the top CFG scoring validated biomarkers, as well as PIK3C3, have anti-apoptotic and neurotrophic effects, are decreased in expression in suicidality and are known targets of the anti-suicidal mood stabilizer drug lithium, which increases their expression and/or activity. Circadian clock genes were overrepresented among the top markers. Notably, PER1, increased in expression in suicidality, had an AUC of 84% for predicting future hospitalizations, and CSNK1A1, decreased in expression, had an AUC of 96% for predicting future hospitalizations. Circadian clock abnormalities are related to mood disorder, and sleep abnormalities have been implicated in suicide. Docosahexaenoic acid signaling was one of the top biological pathways overrepresented in validated biomarkers, which is of interest given the potential therapeutic and prophylactic benefits of omega-3 fatty acids. Some of the top biomarkers from the current work in women showed co-directionality of change in expression with our previous work in men, whereas others had changes in opposite directions, underlying the issue of biological context and differences in suicidality between the two genders. With this study, we begin to shed much needed light in the area of female suicidality, identify useful objective predictors and help understand gender commonalities and differences. During the conduct of the study, one participant committed suicide. In retrospect, when the analyses were completed, her UP-Suicide risk prediction score was at the 100 percentile of all participants tested.Molecular Psychiatry advance online publication, 5 April 2016; doi:10.1038/mp.2016.31.

Original languageEnglish (US)
JournalMolecular Psychiatry
DOIs
StateAccepted/In press - Apr 5 2016

Fingerprint

Area Under Curve
Biomarkers
Suicide
Suicidal Ideation
Hospitalization
Psychiatry
Circadian Clocks
Genomics
Bipolar Disorder
Coroners and Medical Examiners
Social Isolation
Docosahexaenoic Acids
Omega-3 Fatty Acids
Mood Disorders
Lithium
ROC Curve
Psychotic Disorders
Genes
Publications
Schizophrenia

ASJC Scopus subject areas

  • Molecular Biology
  • Psychiatry and Mental health
  • Cellular and Molecular Neuroscience

Cite this

Towards understanding and predicting suicidality in women : biomarkers and clinical risk assessment. / Levey, D. F.; Niculescu, E. M.; Le-Niculescu, Helen; Dainton, H. L.; Phalen, P. L.; Ladd, T. B.; Weber, H.; Belanger, E.; Graham, D. L.; Khan, F. N.; Vanipenta, N. P.; Stage, E. C.; Ballew, A.; Yard, M.; Gelbart, T.; Shekhar, Anantha; Schork, N. J.; Kurian, S. M.; Sandusky, George; Salomon, D. R.; Niculescu, Alexander.

In: Molecular Psychiatry, 05.04.2016.

Research output: Contribution to journalArticle

Levey, DF, Niculescu, EM, Le-Niculescu, H, Dainton, HL, Phalen, PL, Ladd, TB, Weber, H, Belanger, E, Graham, DL, Khan, FN, Vanipenta, NP, Stage, EC, Ballew, A, Yard, M, Gelbart, T, Shekhar, A, Schork, NJ, Kurian, SM, Sandusky, G, Salomon, DR & Niculescu, A 2016, 'Towards understanding and predicting suicidality in women: biomarkers and clinical risk assessment', Molecular Psychiatry. https://doi.org/10.1038/mp.2016.31
Levey, D. F. ; Niculescu, E. M. ; Le-Niculescu, Helen ; Dainton, H. L. ; Phalen, P. L. ; Ladd, T. B. ; Weber, H. ; Belanger, E. ; Graham, D. L. ; Khan, F. N. ; Vanipenta, N. P. ; Stage, E. C. ; Ballew, A. ; Yard, M. ; Gelbart, T. ; Shekhar, Anantha ; Schork, N. J. ; Kurian, S. M. ; Sandusky, George ; Salomon, D. R. ; Niculescu, Alexander. / Towards understanding and predicting suicidality in women : biomarkers and clinical risk assessment. In: Molecular Psychiatry. 2016.
@article{ee8ff9f919064239965249ce237b95fa,
title = "Towards understanding and predicting suicidality in women: biomarkers and clinical risk assessment",
abstract = "Women are under-represented in research on suicidality to date. Although women have a lower rate of suicide completion than men, due in part to the less-violent methods used, they have a higher rate of suicide attempts. Our group has previously identified genomic (blood gene expression biomarkers) and clinical information (apps) predictors for suicidality in men. We now describe pilot studies in women. We used a powerful within-participant discovery approach to identify genes that change in expression between no suicidal ideation (no SI) and high suicidal ideation (high SI) states (n=12 participants out of a cohort of 51 women psychiatric participants followed longitudinally, with diagnoses of bipolar disorder, depression, schizoaffective disorder and schizophrenia). We then used a Convergent Functional Genomics (CFG) approach to prioritize the candidate biomarkers identified in the discovery step by using all the prior evidence in the field. Next, we validated for suicidal behavior the top-ranked biomarkers for SI, in a demographically matched cohort of women suicide completers from the coroner’s office (n=6), by assessing which markers were stepwise changed from no SI to high SI to suicide completers. We then tested the 50 biomarkers that survived Bonferroni correction in the validation step, as well as top increased and decreased biomarkers from the discovery and prioritization steps, in a completely independent test cohort of women psychiatric disorder participants for prediction of SI (n=33) and in a future follow-up cohort of psychiatric disorder participants for prediction of psychiatric hospitalizations due to suicidality (n=24). Additionally, we examined how two clinical instruments in the form of apps, Convergent Functional Information for Suicidality (CFI-S) and Simplified Affective State Scale (SASS), previously tested in men, perform in women. The top CFI-S item distinguishing high SI from no SI states was the chronic stress of social isolation. We then showed how the clinical information apps combined with the 50 validated biomarkers into a broad predictor (UP-Suicide), our apriori primary end point, predicts suicidality in women. UP-Suicide had a receiver-operating characteristic (ROC) area under the curve (AUC) of 82{\%} for predicting SI and an AUC of 78{\%} for predicting future hospitalizations for suicidality. Some of the individual components of the UP-Suicide showed even better results. SASS had an AUC of 81{\%} for predicting SI, CFI-S had an AUC of 84{\%} and the combination of the two apps had an AUC of 87{\%}. The top biomarker from our sequential discovery, prioritization and validation steps, BCL2, predicted future hospitalizations due to suicidality with an AUC of 89{\%}, and the panel of 50 validated biomarkers (BioM-50) predicted future hospitalizations due to suicidality with an AUC of 94{\%}. The best overall single blood biomarker for predictions was PIK3C3 with an AUC of 65{\%} for SI and an AUC of 90{\%} for future hospitalizations. Finally, we sought to understand the biology of the biomarkers. BCL2 and GSK3B, the top CFG scoring validated biomarkers, as well as PIK3C3, have anti-apoptotic and neurotrophic effects, are decreased in expression in suicidality and are known targets of the anti-suicidal mood stabilizer drug lithium, which increases their expression and/or activity. Circadian clock genes were overrepresented among the top markers. Notably, PER1, increased in expression in suicidality, had an AUC of 84{\%} for predicting future hospitalizations, and CSNK1A1, decreased in expression, had an AUC of 96{\%} for predicting future hospitalizations. Circadian clock abnormalities are related to mood disorder, and sleep abnormalities have been implicated in suicide. Docosahexaenoic acid signaling was one of the top biological pathways overrepresented in validated biomarkers, which is of interest given the potential therapeutic and prophylactic benefits of omega-3 fatty acids. Some of the top biomarkers from the current work in women showed co-directionality of change in expression with our previous work in men, whereas others had changes in opposite directions, underlying the issue of biological context and differences in suicidality between the two genders. With this study, we begin to shed much needed light in the area of female suicidality, identify useful objective predictors and help understand gender commonalities and differences. During the conduct of the study, one participant committed suicide. In retrospect, when the analyses were completed, her UP-Suicide risk prediction score was at the 100 percentile of all participants tested.Molecular Psychiatry advance online publication, 5 April 2016; doi:10.1038/mp.2016.31.",
author = "Levey, {D. F.} and Niculescu, {E. M.} and Helen Le-Niculescu and Dainton, {H. L.} and Phalen, {P. L.} and Ladd, {T. B.} and H. Weber and E. Belanger and Graham, {D. L.} and Khan, {F. N.} and Vanipenta, {N. P.} and Stage, {E. C.} and A. Ballew and M. Yard and T. Gelbart and Anantha Shekhar and Schork, {N. J.} and Kurian, {S. M.} and George Sandusky and Salomon, {D. R.} and Alexander Niculescu",
year = "2016",
month = "4",
day = "5",
doi = "10.1038/mp.2016.31",
language = "English (US)",
journal = "Molecular Psychiatry",
issn = "1359-4184",
publisher = "Nature Publishing Group",

}

TY - JOUR

T1 - Towards understanding and predicting suicidality in women

T2 - biomarkers and clinical risk assessment

AU - Levey, D. F.

AU - Niculescu, E. M.

AU - Le-Niculescu, Helen

AU - Dainton, H. L.

AU - Phalen, P. L.

AU - Ladd, T. B.

AU - Weber, H.

AU - Belanger, E.

AU - Graham, D. L.

AU - Khan, F. N.

AU - Vanipenta, N. P.

AU - Stage, E. C.

AU - Ballew, A.

AU - Yard, M.

AU - Gelbart, T.

AU - Shekhar, Anantha

AU - Schork, N. J.

AU - Kurian, S. M.

AU - Sandusky, George

AU - Salomon, D. R.

AU - Niculescu, Alexander

PY - 2016/4/5

Y1 - 2016/4/5

N2 - Women are under-represented in research on suicidality to date. Although women have a lower rate of suicide completion than men, due in part to the less-violent methods used, they have a higher rate of suicide attempts. Our group has previously identified genomic (blood gene expression biomarkers) and clinical information (apps) predictors for suicidality in men. We now describe pilot studies in women. We used a powerful within-participant discovery approach to identify genes that change in expression between no suicidal ideation (no SI) and high suicidal ideation (high SI) states (n=12 participants out of a cohort of 51 women psychiatric participants followed longitudinally, with diagnoses of bipolar disorder, depression, schizoaffective disorder and schizophrenia). We then used a Convergent Functional Genomics (CFG) approach to prioritize the candidate biomarkers identified in the discovery step by using all the prior evidence in the field. Next, we validated for suicidal behavior the top-ranked biomarkers for SI, in a demographically matched cohort of women suicide completers from the coroner’s office (n=6), by assessing which markers were stepwise changed from no SI to high SI to suicide completers. We then tested the 50 biomarkers that survived Bonferroni correction in the validation step, as well as top increased and decreased biomarkers from the discovery and prioritization steps, in a completely independent test cohort of women psychiatric disorder participants for prediction of SI (n=33) and in a future follow-up cohort of psychiatric disorder participants for prediction of psychiatric hospitalizations due to suicidality (n=24). Additionally, we examined how two clinical instruments in the form of apps, Convergent Functional Information for Suicidality (CFI-S) and Simplified Affective State Scale (SASS), previously tested in men, perform in women. The top CFI-S item distinguishing high SI from no SI states was the chronic stress of social isolation. We then showed how the clinical information apps combined with the 50 validated biomarkers into a broad predictor (UP-Suicide), our apriori primary end point, predicts suicidality in women. UP-Suicide had a receiver-operating characteristic (ROC) area under the curve (AUC) of 82% for predicting SI and an AUC of 78% for predicting future hospitalizations for suicidality. Some of the individual components of the UP-Suicide showed even better results. SASS had an AUC of 81% for predicting SI, CFI-S had an AUC of 84% and the combination of the two apps had an AUC of 87%. The top biomarker from our sequential discovery, prioritization and validation steps, BCL2, predicted future hospitalizations due to suicidality with an AUC of 89%, and the panel of 50 validated biomarkers (BioM-50) predicted future hospitalizations due to suicidality with an AUC of 94%. The best overall single blood biomarker for predictions was PIK3C3 with an AUC of 65% for SI and an AUC of 90% for future hospitalizations. Finally, we sought to understand the biology of the biomarkers. BCL2 and GSK3B, the top CFG scoring validated biomarkers, as well as PIK3C3, have anti-apoptotic and neurotrophic effects, are decreased in expression in suicidality and are known targets of the anti-suicidal mood stabilizer drug lithium, which increases their expression and/or activity. Circadian clock genes were overrepresented among the top markers. Notably, PER1, increased in expression in suicidality, had an AUC of 84% for predicting future hospitalizations, and CSNK1A1, decreased in expression, had an AUC of 96% for predicting future hospitalizations. Circadian clock abnormalities are related to mood disorder, and sleep abnormalities have been implicated in suicide. Docosahexaenoic acid signaling was one of the top biological pathways overrepresented in validated biomarkers, which is of interest given the potential therapeutic and prophylactic benefits of omega-3 fatty acids. Some of the top biomarkers from the current work in women showed co-directionality of change in expression with our previous work in men, whereas others had changes in opposite directions, underlying the issue of biological context and differences in suicidality between the two genders. With this study, we begin to shed much needed light in the area of female suicidality, identify useful objective predictors and help understand gender commonalities and differences. During the conduct of the study, one participant committed suicide. In retrospect, when the analyses were completed, her UP-Suicide risk prediction score was at the 100 percentile of all participants tested.Molecular Psychiatry advance online publication, 5 April 2016; doi:10.1038/mp.2016.31.

AB - Women are under-represented in research on suicidality to date. Although women have a lower rate of suicide completion than men, due in part to the less-violent methods used, they have a higher rate of suicide attempts. Our group has previously identified genomic (blood gene expression biomarkers) and clinical information (apps) predictors for suicidality in men. We now describe pilot studies in women. We used a powerful within-participant discovery approach to identify genes that change in expression between no suicidal ideation (no SI) and high suicidal ideation (high SI) states (n=12 participants out of a cohort of 51 women psychiatric participants followed longitudinally, with diagnoses of bipolar disorder, depression, schizoaffective disorder and schizophrenia). We then used a Convergent Functional Genomics (CFG) approach to prioritize the candidate biomarkers identified in the discovery step by using all the prior evidence in the field. Next, we validated for suicidal behavior the top-ranked biomarkers for SI, in a demographically matched cohort of women suicide completers from the coroner’s office (n=6), by assessing which markers were stepwise changed from no SI to high SI to suicide completers. We then tested the 50 biomarkers that survived Bonferroni correction in the validation step, as well as top increased and decreased biomarkers from the discovery and prioritization steps, in a completely independent test cohort of women psychiatric disorder participants for prediction of SI (n=33) and in a future follow-up cohort of psychiatric disorder participants for prediction of psychiatric hospitalizations due to suicidality (n=24). Additionally, we examined how two clinical instruments in the form of apps, Convergent Functional Information for Suicidality (CFI-S) and Simplified Affective State Scale (SASS), previously tested in men, perform in women. The top CFI-S item distinguishing high SI from no SI states was the chronic stress of social isolation. We then showed how the clinical information apps combined with the 50 validated biomarkers into a broad predictor (UP-Suicide), our apriori primary end point, predicts suicidality in women. UP-Suicide had a receiver-operating characteristic (ROC) area under the curve (AUC) of 82% for predicting SI and an AUC of 78% for predicting future hospitalizations for suicidality. Some of the individual components of the UP-Suicide showed even better results. SASS had an AUC of 81% for predicting SI, CFI-S had an AUC of 84% and the combination of the two apps had an AUC of 87%. The top biomarker from our sequential discovery, prioritization and validation steps, BCL2, predicted future hospitalizations due to suicidality with an AUC of 89%, and the panel of 50 validated biomarkers (BioM-50) predicted future hospitalizations due to suicidality with an AUC of 94%. The best overall single blood biomarker for predictions was PIK3C3 with an AUC of 65% for SI and an AUC of 90% for future hospitalizations. Finally, we sought to understand the biology of the biomarkers. BCL2 and GSK3B, the top CFG scoring validated biomarkers, as well as PIK3C3, have anti-apoptotic and neurotrophic effects, are decreased in expression in suicidality and are known targets of the anti-suicidal mood stabilizer drug lithium, which increases their expression and/or activity. Circadian clock genes were overrepresented among the top markers. Notably, PER1, increased in expression in suicidality, had an AUC of 84% for predicting future hospitalizations, and CSNK1A1, decreased in expression, had an AUC of 96% for predicting future hospitalizations. Circadian clock abnormalities are related to mood disorder, and sleep abnormalities have been implicated in suicide. Docosahexaenoic acid signaling was one of the top biological pathways overrepresented in validated biomarkers, which is of interest given the potential therapeutic and prophylactic benefits of omega-3 fatty acids. Some of the top biomarkers from the current work in women showed co-directionality of change in expression with our previous work in men, whereas others had changes in opposite directions, underlying the issue of biological context and differences in suicidality between the two genders. With this study, we begin to shed much needed light in the area of female suicidality, identify useful objective predictors and help understand gender commonalities and differences. During the conduct of the study, one participant committed suicide. In retrospect, when the analyses were completed, her UP-Suicide risk prediction score was at the 100 percentile of all participants tested.Molecular Psychiatry advance online publication, 5 April 2016; doi:10.1038/mp.2016.31.

UR - http://www.scopus.com/inward/record.url?scp=84962122616&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84962122616&partnerID=8YFLogxK

U2 - 10.1038/mp.2016.31

DO - 10.1038/mp.2016.31

M3 - Article

JO - Molecular Psychiatry

JF - Molecular Psychiatry

SN - 1359-4184

ER -