‘Never give in, never give in, never, never, never.’

– Winston Churchill

One person dies by suicide every 40 s worldwide, despite suicide being a potentially preventable tragedy. It is clear that better assessment, tracking, and prediction of suicidality (ideation, actions) are needed. We endeavored to develop better clinical data tools and objective blood biomarkers for such purposes. The two approaches are complementary, synergistic, and can inform each other.

First, in terms of clinical data, we developed a simple polyphenic risk score, based on phenes (phenotypic items) that are known risk factors for suicide. Similar to polygenic scoring, these were scored in a binary fashion as 0 or 1 (absent or present). This questionnaire/app, known as Convergent Functional Information for Suicide (CFI-S), has 22 easy-to-answer items, related to life issues, mental health, physical health, environmental stress, addictions, and cultural factors. It purposefully does not ask about suicidal ideation, as that is a loaded question in most settings, and because people who truly want to kill themselves may not share that information for fear of being stopped. CFI-S can be self-scored or clinician scored, obtained from patient interview, or from information from next of kin or medical records. CFI-S has shown good predictive ability for suicidal ideation and for future hospitalizations for suicidality, in men (Niculescu et al, 2015) and in women (Levey et al, 2016). It can have a therapy-guiding component, in that over half of the risk factors it identifies are correctable. In addition, we have developed and used a simple visual analog scale type of questionnaire/app, known as Simple Affective State Scale (SASS), that provides a score for anxiety and one for mood at a particular moment in time (Niculescu et al, 2015; Levey et al, 2016). High anxiety scores and low mood scores were predictive for suicidal ideation and future hospitalizations by themselves, and even more so when combined with the CFI-S score, which provides broader contextual information.

Second, as the brain (the target organ) cannot be biopsied in live individuals, we endeavored to develop a liquid biopsy, ie, to find blood biomarkers for suicidality that would have translational and practical clinical utility. The possibility of finding blood biomarkers for psychiatric phenotypes had been substantiated by earlier pilot studies by our group focused on mood (Le-Niculescu et al, 2009), psychosis (Kurian et al, 2011), and suicide (Le-Niculescu et al, 2013). How and why peripheral molecular changes reflect what is happening in the brain is an area of ongoing research. Brain cells and blood cells share some genetic polymorphisms, as well as exposure to the internal milieu and whole-body responses such as stress. There are also more direct neuro-immune connections. However, it is likely that only a small proportion of peripheral changes in gene expression are relevant to and/or concordant with brain changes related to suicide. Finding the true signal among all the noise is key to success. It is predicated on having, for example, a powerful longitudinal within-subject design to conduct the discovery, prioritization of findings using convergent functional genomics, and validation of the biomarkers in relevant independent cohorts (Niculescu et al, 2015). Using such approaches, we were successful in finding blood biomarkers that were predictive of suicidality, at least in the psychiatric patient populations that we have studied so far (Niculescu et al, 2015; Levey et al, 2016; Le-Niculescu et al, 2013). Some of them have prior brain gene expression evidence (Niculescu et al, 2015; Levey et al, 2016), which is part of our convergent approach to prioritize them. While the biomarkers were reasonably predictive by themselves, they were in general less predictive than the clinical information scores we described above. The biomarkers may serve a useful purpose when clinical information is not available and/or to (1) provide a window into the biology of suicidality, (2) help stratify patients, and (3) monitor disease course and response to treatment in an objective fashion. Importantly, when combined, the biomarker scores and clinical information scores were synergistic in terms of improved predictive ability (Niculescu et al, 2015; Levey et al, 2016).

It has not escaped our attention that the general strategy outlined here can be used to try to understand, predict, and improve other behavioral phenotypes.


This work was supported by an NIH Directors’ New Innovator Award (1DP2OD007363) and a VA Merit Award (2I01CX000139) to ABN. ABN is listed as inventor on a patent application being filed by Indiana University, and is a co-founder of MindX Sciences. HL-N declares no conflict of interest.