The utility of hierarchical models of psychopathology in genetics and biomarker research

In their incisive paper, Lahey et al ¹ discuss empirically‐derived, hierarchical taxonomies of psychopathology that in recent years have gained prominence in psychiatric research. This is a timely opportunity to reflect on some of the implications and future directions for leveraging redefined psychiatric constructs in research efforts.

The authors highlight genetics as a leading edge in the paradigm shift in psychiatric nosology. Indeed, behavioral (family and twin) and molecular genetic studies provide some of the strongest evidence to date that biological vulnerability transcends diagnostic boundaries between disorders, as well as boundaries between psychopathology and normality ² .

First, there is ample evidence that genetic liability to mental illness is continuously distributed, on a dimension from heathy traits in the general population (e.g., personality trait neuroticism) to corresponding clinical diagnoses (e.g., major depressive disorder) ³ .

Second, the genetic architecture of psychopathology appears to consist of sets of genetic influences operating at different levels of specificity, across a multi‐tiered hierarchy. For example, models using genetic loci identified in genome‐wide association studies (GWAS) found that a significant proportion of genomic influences is common to numerous psychiatric disorders (e.g., schizophrenia, depression, attention‐deficit/hyperactivity disorder)4, 5. The remaining risk, which is considerable in size, is disorder‐specific, indicating that genetic factors unique to narrow constructs also play a role in the etiology of psychopathology. In sum, the genetic architecture of psychopathology is dimensional and hierarchical. Its correspondence to one classification system – the Hierarchical Taxonomy of Psychopathology (HiTOP) – has been detailed previously ² .

The hierarchical structure of genetic risk has major implications for future research, in that it provides empirically‐validated targets for genetic inquiry. In particular, it promises to advance GWAS, which continue to be a leading approach to discovering genetic variations associated with psychiatric conditions. Specifically, GWAS calibrated to reliable, empirical constructs at different levels of specificity – e.g., general factor or internalizing spectrum – could identify more genetic risk loci than traditional case‐control studies. For example, more genetic variants were found when a higher‐order, dimensional “fear” factor was used as a GWAS target, compared to a case‐control anxiety disorder status ⁶ . However, additional empirical studies are needed to interrogate this hypothesis comprehensively.

Consequently, the hierarchical approach to GWAS might help explicate which genet­ic effects are transdiagnostic vs. specific. Knowing the specificity of identified genetic loci is crucial for follow‐up characterization of downstream biological processes, as well as for translation of GWAS results into research tools and clinical instruments.

One such instrument is polygenic risk score, which captures part of an individual’s genetic susceptibility to a disease. An increasing number of research studies demonstrate associations between polygenic risk scores and psychiatric conditions, although the clinical utility of these scores has not yet been established. The quality of psychiatric assessment in the GWAS is a major determinant of the power and precision of the resulting polygenic risk score. Currently, genetic risk scores developed for one disorder (e.g., schizophrenia) have been found to predict many other conditions and outcomes (e.g., post‐traumatic stress disorder, substance use, cognitive performance), with little specificity, and thus limited potential for research and clinical utility ⁷ . Future GWAS on hierarchical and dimensional constructs could help create more robust polygenic risk scores.

This approach can be extended to longitudinal lifespan research which aims to investigate how genetic factors shape the course of psychopathology over time. Lahey et al ¹ describe key evidence that higher‐order dimensions, in particular the general factor, are highly stable across development. However, age differences and developmental trajectories of the hierarchically‐organized genetic influences have been investigated in only a handful of prospective longitudinal twin studies. Overall, evidence suggests that general, transdiagnostic genetic influences contribute to the continuity and co‐occurrence of psychopathology over time ⁸ . In other words, the developmental stability of the general factor of psychopathology appears to be driven predominantly by transdiagnostic genetic vulnerability.

One implication of this finding is that a polygenic risk score created explicitly to capture genetic risk to the general factor might in the future help predict individual’s vulnerability to a broad range of co‐occurring and chronic psychiatric illnesses, or help identify a subgroup of individuals at the highest genetic risk for recurrent, cross‐disorder psychiatric illness course. Such individuals at a very high genetic risk to enduring general psychopathology could be identified early and prioritized for prevention programs. Genetic influence (transdiagnostic, stable, or otherwise) does not preclude the possibility of effective prevention and treatment.

Beyond the interface with psychiatric genetics, Lahey et al ¹ highlight how hierarchical models have driven novel discoveries in neurobiology, such as by delineating patterns of gray matter volume alterations associated with the general factor. The same principles could be applied to other psychiatric biomarker research. To date, this literature largely consists of disparate studies of single diagnostic categories, obscuring transdiagnostic processes. When cross‐disorder research has been attempted, however, commonalities are observed. For example, there is a very high correlation between transcriptome profiles for bipolar disorder and schizophrenia, with both disorders also showing significant, albeit smaller, transcriptome profile overlap with depression ⁹ . This pattern of overlap suggests that gene expression in the brain could be mapped onto the general and specific factors of psychopathology.

Similarly, many epigenetic, inflammatory, hormonal and metabolic biomarkers are implicated across studies of different psychiatric disorders. The hierarchical approach provides a more powerful and systematic way for these fields to probe which biological correlates are general vs. disorder‐specific, allowing for the derivation of biomarker signatures at different levels of specificity. Importantly, significant genetic and environmental influences at the lower levels of the hierarchy suggest that symptom‐specific downstream biomarkers can be identified alongside transdiagnostic biomarkers. Consequently, screening and interventions could be developed to target biological processes that all dimensions of psychopathology appear to have in common, or target processes unique to one or a subset of dimensions.

To achieve these research goals, studies need to assess a wide range of psychopathology across the full spectrum of severity, ranging from personality traits to severe clinical problems. While a comprehensive dimensional measure of psychopathology is currently under construction by researchers affiliated with the HiTOP model, existing instruments can be combined to assess general and lower‐order dimensions ² . Many of these measures have been validated in short versions and can be administered remotely for feasibility.

Overall, as Lahey et al¹ point out, the hierarchical conceptualization of psychopathology will benefit clinical practice. This improvement will in part come from this model’s unique utility for advancing basic and translational psychiatric research.