Exemplar Scoring Identifies Genetically Separable Phenotypes of Lithium Responsive Bipolar Disorder

Abstract

Predicting lithium response (LiR) in bipolar disorder (BD) could expedite effective pharmacotherapy, but phenotypic heterogeneity of bipolar disorder has complicated the search for genomic markers. We thus sought to determine whether patients with “exemplary phenotypes”—those whose clinical features are reliably predictive of LiR and non-response (LiNR)—are more genetically separable than those with less exemplary phenotypes. We applied machine learning methods to clinical data collected from people with BD (n=1266 across 7 international centres; 34.7% responders) to compute an “exemplar score,” which identified a subset of subjects whose clinical phenotypes were most robustly predictive of LiR/LiNR. For subjects whose genotypes were available (n=321), we evaluated whether responders/non-responders with exemplary phenotypes could be more accurately classified based on genetic data than those with non-exemplary phenotypes. We showed that the best LiR exemplars had later illness onset, completely episodic clinical course, absence of rapid cycling and psychosis, and few psychiatric comorbidities. The best exemplars of LiR and LiNR were genetically separable with an area under the receiver operating characteristic curve of 0.88 (IQR [0.83, 0.98]), compared to 0.66 [0.61, 0.80] (p=0.0032) among the poor exemplars. Variants in the Alzheimer’s amyloid secretase pathway, along with G-protein coupled receptor, muscarinic acetylcholine, and histamine H1R signaling pathways were particularly informative predictors. In sum, the most reliably predictive clinical features of LiR and LiNR patients correspond to previously well-characterized phenotypic spectra whose genomic profiles are relatively distinct. Future work must enlarge the sample for genomic classification and include prediction of response to other mood stabilizers.