Use of multiple polygenic risk scores for distinguishing schizophrenia-spectrum disorder and affective psychosis categories in a first-episode sample; the EU-GEI study

Machine learning models for predicting blood pressure phenotypes by combining multiple polygenic risk scores

We construct non-linear machine learning (ML) prediction models for systolic and diastolic blood pressure (SBP, DBP) using demographic and clinical variables and polygenic risk scores (PRSs). We developed a two-model ensemble, consisting of a baseline model, where prediction is based on demographic and clinical variables only, and a genetic model, where we also include PRSs. We evaluate the use of a linear versus a non-linear model at both the baseline and the genetic model levels and assess the improvement in performance when incorporating multiple PRSs. We report the ensemble model's performance as percentage variance explained (PVE) on a held-out test dataset. A non-linear baseline model improved the PVEs from 28.1 to 30.1% (SBP) and 14.3% to 17.4% (DBP) compared with a linear baseline model. Including seven PRSs in the genetic model computed based on the largest available GWAS of SBP/DBP improved the genetic model PVE from 4.8 to 5.1% (SBP) and 4.7 to 5% (DBP) compared to using a single PRS. Adding additional 14 PRSs computed based on two independent GWASs further increased the genetic model PVE to 6.3% (SBP) and 5.7% (DBP). PVE differed across self-reported race/ethnicity groups, with primarily all non-White groups benefitting from the inclusion of additional PRSs. In summary, non-linear ML models improves BP prediction in models incorporating diverse populations.

Machine learning models for blood pressure phenotypes combining multiple polygenic risk scores

We construct non-linear machine learning (ML) prediction models for systolic and diastolic blood pressure (SBP, DBP) using demographic and clinical variables and polygenic risk scores (PRSs). We developed a two-model ensemble, consisting of a baseline model, where prediction is based on demographic and clinical variables only, and a genetic model, where we also include PRSs. We evaluate the use of a linear versus a non-linear model at both the baseline and the genetic model levels and assess the improvement in performance when incorporating multiple PRSs. We report the ensemble model's performance as percentage variance explained (PVE) on a held-out test dataset. A non-linear baseline model improved the PVEs from 28.1% to 30.1% (SBP) and 14.3% to 17.4% (DBP) compared with a linear baseline model. Including seven PRSs in the genetic model computed based on the largest available GWAS of SBP/DBP improved the genetic model PVE from 4.8% to 5.1% (SBP) and 4.7% to 5% (DBP) compared to using a single PRS. Adding additional 14 PRSs computed based on two independent GWASs further increased the genetic model PVE to 6.3% (SBP) and 5.7% (DBP). PVE differed across self-reported race/ethnicity groups, with primarily all non-White groups benefitting from the inclusion of additional PRSs.

Polygenic risk scores and the prediction of onset of mood and psychotic disorders in adolescents and young adults

AIM

To examine whether polygenic risk scores (PRS) for neuroticism, depression, bipolar disorder and schizophrenia are higher in individuals manifesting trans-diagnostic risk factors for the development of major mental disorders and whether PRS enhance prediction of early onset full-threshold disorders.

METHODS

Using data from the Brisbane Longitudinal Twin Study, we examined individual PRS for neuroticism, depression, bipolar disorder and schizophrenia, recorded evidence of subthreshold syndromes and family history of mood and/or psychotic disorders and noted progression to trans-diagnostic clinical caseness (onset of major mental disorders) at follow-up. We undertook multivariate, receiver operating curve and logistic regression analyses that were adjusted for known variables of influence (age, twin status, and so on).

RESULTS

Of 1473 eligible participants (female = 866, 59%; mean age 26.3 years), 28% (n = 409) met caseness criteria for a mood and/or psychotic disorder. All PRS were higher in cases versus non-cases but associations with different levels of risk were inconsistent. The prediction of caseness (reported as area under the curve with 95% confidence intervals [CI]) improved from 0.68 (95% CI: 0.65, 0.71) when estimated using clinical risk factors alone up to 0.71 (95% CI: 0.69, 0.73) when PRS were added to the model. Logistic regression identified five variables that optimally classified individuals according to caseness: age, sex, individual risk characteristics, PRS for depression and mental health case status of cotwins or siblings.

CONCLUSIONS

The findings need replication. However, this exploratory study suggests that combining PRS with other risk factors has the potential to improve outcome prediction in youth.

AI-based multi-PRS models outperform classical single-PRS models

Polygenic risk scores (PRS) calculate the risk for a specific disease based on the weighted sum of associated alleles from different genetic loci in the germline estimated by regression models. Recent advances in genetics made it possible to create polygenic predictors of complex human traits, including risks for many important complex diseases, such as cancer, diabetes, or cardiovascular diseases, typically influenced by many genetic variants, each of which has a negligible effect on overall risk. In the current study, we analyzed whether adding additional PRS from other diseases to the prediction models and replacing the regressions with machine learning models can improve overall predictive performance. Results showed that multi-PRS models outperform single-PRS models significantly on different diseases. Moreover, replacing regression models with machine learning models, i.e., deep learning, can also improve overall accuracy.

Effect of polygenic risk score, family load of schizophrenia and exposome risk score, and their interactions, on the long-term outcome of first-episode psychosis

BACKGROUND

Consistent evidence supports the involvement of genetic and environmental factors, and their interactions, in the etiology of psychosis. First-episode psychosis (FEP) comprises a group of disorders that show great clinical and long-term outcome heterogeneity, and the extent to which genetic, familial and environmental factors account for predicting the long-term outcome in FEP patients remains scarcely known.

METHODS

The SEGPEPs is an inception cohort study of 243 first-admission patients with FEP who were followed-up for a mean of 20.9 years. FEP patients were thoroughly evaluated by standardized instruments, with 164 patients providing DNA. Aggregate scores estimated in large populations for polygenic risk score (PRS-Sz), exposome risk score (ERS-Sz) and familial load score for schizophrenia (FLS-Sz) were ascertained. Long-term functioning was assessed by means of the Social and Occupational Functioning Assessment Scale (SOFAS). The relative excess risk due to interaction (RERI) was used as a standard method to estimate the effect of interaction of risk factors.

RESULTS

Our results showed that a high FLS-Sz gave greater explanatory capacity for long-term outcome, followed by the ERS-Sz and then the PRS-Sz. The PRS-Sz did not discriminate significantly between recovered and non-recovered FEP patients in the long term. No significant interaction between the PRS-Sz, ERS-Sz or FLS-Sz regarding the long-term functioning of FEP patients was found.

CONCLUSIONS

Our results support an additive model of familial antecedents of schizophrenia, environmental risk factors and polygenic risk factors as contributors to a poor long-term functional outcome for FEP patients.

Polygenic scores in biomedical research

Public health strategies aimed at disease prevention or early detection and intervention have the potential to advance human health worldwide. However, their success depends on the identification of risk factors that underlie disease burden in the general population. Genome-wide association studies (GWAS) have implicated thousands of single-nucleotide polymorphisms (SNPs) in common complex diseases or traits. By calculating a weighted sum of the number of trait-associated alleles harboured by an individual, a polygenic score (PGS), also called a polygenic risk score (PRS), can be constructed that reflects an individual’s estimated genetic predisposition for a given phenotype. Here, we ask six experts to give their opinions on the utility of these probabilistic tools, their strengths and limitations, and the remaining barriers that need to be overcome for their equitable use.

First-Episode Psychosis Patients Who Deteriorated in the Premorbid Period Do Not Have Higher Polygenic Risk Scores Than Others: A Cluster Analysis of EU-GEI Data

Cluster studies identified a subgroup of patients with psychosis whose premorbid adjustment deteriorates before the onset, which may reflect variation in genetic influence. However, other studies reported a complex relationship between distinctive patterns of cannabis use and cognitive and premorbid impairment that is worthy of consideration. We examined whether: (1) premorbid social functioning (PSF) and premorbid academic functioning (PAF) in childhood and adolescence and current intellectual quotient (IQ) define different clusters in 802 first-episode of psychosis (FEP) patients; resulting clusters vary in (2) polygenic risk scores (PRSs) for schizophrenia (SCZ_PRS), bipolar disorder (BD_PRS), major depression (MD_PRS), and IQ (IQ_PRS), and (3) patterns of cannabis use, compared to 1,263 population-based controls. Four transdiagnostic clusters emerged (BIC = 2268.5): (1) high-cognitive-functioning (n = 205), with the highest IQ (Mean = 106.1, 95% CI: 104.3, 107.9) and PAF, but low PSF. (2) Low-cognitive-functioning (n = 223), with the lowest IQ (Mean = 73.9, 95% CI: 72.2, 75.7) and PAF, but normal PSF. (3) Intermediate (n = 224) (Mean_IQ = 80.8, 95% CI: 79.1, 82.5) with low-improving PAF and PSF. 4) Deteriorating (n = 150) (Mean_IQ = 80.6, 95% CI: 78.5, 82.7), with normal-deteriorating PAF and PSF. The PRSs explained 7.9% of between-group membership. FEP had higher SCZ_PRS than controls [F(4,1319) = 20.4, P < .001]. Among the clusters, the deteriorating group had lower SCZ_PRS and was likelier to have used high-potency cannabis daily. Patients with FEP clustered according to their premorbid and cognitive abilities. Pronounced premorbid deterioration was not typical of most FEP, including those more strongly predisposed to schizophrenia, but appeared in a cluster with a history of high-potency cannabis use.