Development and validation of genomic biotypes for schizophrenia susceptibility from multiple polygenic scores

Development and validation of polygenic risk profiles of schizophrenia | medRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-P4HH5NV'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search Development and validation of polygenic risk profiles of schizophrenia View ORCID Profile Ersoy Kocak , View ORCID Profile Joonas Naamanka , View ORCID Profile Tobias Gradinger , View ORCID Profile Fiona Klaassen , View ORCID Profile Johannes Nitsche , View ORCID Profile Philipp Grotehusmann , View ORCID Profile Kristina Adorjan , View ORCID Profile Linda A. Antonucci , View ORCID Profile Giuseppe Blasi , View ORCID Profile Monika Budde , View ORCID Profile Piergiuseppe Di Palo , View ORCID Profile Maria Heilbronner , View ORCID Profile Gianluca C. Kikidis , View ORCID Profile Alba Navarro-Flores , View ORCID Profile Mojtaba Oraki Kohshour , View ORCID Profile Sergi Papiol , Alessandra Raio , View ORCID Profile Antonio Rampino , View ORCID Profile Daniela Reich-Erkelenz , View ORCID Profile Eva C. Schulte , View ORCID Profile Fanny Senner , View ORCID Profile Leonardo Sportelli , View ORCID Profile Alessandro Bertolino , View ORCID Profile Peter Falkai , View ORCID Profile Urs Heilbronner , View ORCID Profile Giulio Pergola , View ORCID Profile Thomas G. Schulze , View ORCID Profile Andreas Meyer-Lindenberg , View ORCID Profile Fabian Streit , View ORCID Profile Emanuel Schwarz , FinnGen doi: https://doi.org/10.1101/2025.04.29.25326656 Ersoy Kocak 1 Department of Psychiatry and Psychotherapy, Central Institute of Mental Health, Medical Faculty Mannheim, Heidelberg University , Mannheim, Germany 2 Hector Institute for Artificial Intelligence in Psychiatry, Central Institute of Mental Health, Medical Faculty Mannheim, Heidelberg University , Mannheim, Germany MSc Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Ersoy Kocak Joonas Naamanka 1 Department of Psychiatry and Psychotherapy, Central Institute of Mental Health, Medical Faculty Mannheim, Heidelberg University , Mannheim, Germany 2 Hector Institute for Artificial Intelligence in Psychiatry, Central Institute of Mental Health, Medical Faculty Mannheim, Heidelberg University , Mannheim, Germany 3 SleepWell Research Program, Faculty of Medicine, University of Helsinki , Helsinki, Finland MSc Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Joonas Naamanka Tobias Gradinger 1 Department of Psychiatry and Psychotherapy, Central Institute of Mental Health, Medical Faculty Mannheim, Heidelberg University , Mannheim, Germany 2 Hector Institute for Artificial Intelligence in Psychiatry, Central Institute of Mental Health, Medical Faculty Mannheim, Heidelberg University , Mannheim, Germany MD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Tobias Gradinger Fiona Klaassen 1 Department of Psychiatry and Psychotherapy, Central Institute of Mental Health, Medical Faculty Mannheim, Heidelberg University , Mannheim, Germany 2 Hector Institute for Artificial Intelligence in Psychiatry, Central Institute of Mental Health, Medical Faculty Mannheim, Heidelberg University , Mannheim, Germany MSc Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Fiona Klaassen Johannes Nitsche 1 Department of Psychiatry and Psychotherapy, Central Institute of Mental Health, Medical Faculty Mannheim, Heidelberg University , Mannheim, Germany 2 Hector Institute for Artificial Intelligence in Psychiatry, Central Institute of Mental Health, Medical Faculty Mannheim, Heidelberg University , Mannheim, Germany MSc Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Johannes Nitsche Philipp Grotehusmann 1 Department of Psychiatry and Psychotherapy, Central Institute of Mental Health, Medical Faculty Mannheim, Heidelberg University , Mannheim, Germany 2 Hector Institute for Artificial Intelligence in Psychiatry, Central Institute of Mental Health, Medical Faculty Mannheim, Heidelberg University , Mannheim, Germany MSc Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Philipp Grotehusmann Kristina Adorjan 4 Institute of Psychiatric Phenomics and Genomics (IPPG), LMU University Hospital , Munich, Germany 5 University Hospital of Psychiatry and Psychotherapy, University of Bern , Bern, Switzerland MD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Kristina Adorjan Linda A. Antonucci 6 Department of Translational Biomedicine and Neuroscience, University Of Bari Aldo Moro , Bari, Italy PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Linda A. Antonucci Giuseppe Blasi 6 Department of Translational Biomedicine and Neuroscience, University Of Bari Aldo Moro , Bari, Italy 7 U.O.C. Psichiatria Universitaria, Azienda Ospedaliero-Universitaria Consorziale Policlinico , Bari, Italy MD, PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Giuseppe Blasi Monika Budde 4 Institute of Psychiatric Phenomics and Genomics (IPPG), LMU University Hospital , Munich, Germany PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Monika Budde Piergiuseppe Di Palo 6 Department of Translational Biomedicine and Neuroscience, University Of Bari Aldo Moro , Bari, Italy 7 U.O.C. Psichiatria Universitaria, Azienda Ospedaliero-Universitaria Consorziale Policlinico , Bari, Italy MD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Piergiuseppe Di Palo Maria Heilbronner 4 Institute of Psychiatric Phenomics and Genomics (IPPG), LMU University Hospital , Munich, Germany MSc Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Maria Heilbronner Gianluca C. Kikidis 6 Department of Translational Biomedicine and Neuroscience, University Of Bari Aldo Moro , Bari, Italy 8 Lieber Institute for Brain Development, Johns Hopkins Medical Campus , Baltimore, MD, USA MSc Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Gianluca C. Kikidis Alba Navarro-Flores 4 Institute of Psychiatric Phenomics and Genomics (IPPG), LMU University Hospital , Munich, Germany 9 International Max Planck Research School for Translational Psychiatry (IMPRS-TP), Max-Planck-Institute of Psychiatry , Munich, Germany MD, MSc Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Alba Navarro-Flores Mojtaba Oraki Kohshour 4 Institute of Psychiatric Phenomics and Genomics (IPPG), LMU University Hospital , Munich, Germany 10 Department of Immunology, Faculty of Medicine , Ahvaz, Iran PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Mojtaba Oraki Kohshour Sergi Papiol 4 Institute of Psychiatric Phenomics and Genomics (IPPG), LMU University Hospital , Munich, Germany 11 Department of Psychiatry and Psychotherapy, LMU University Hospital , Munich, Germany 12 Department Clinical Translation, Max Planck Institute of Psychiatry , Munich, Germany , Munich, Germany PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Sergi Papiol Alessandra Raio 6 Department of Translational Biomedicine and Neuroscience, University Of Bari Aldo Moro , Bari, Italy PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site Antonio Rampino 6 Department of Translational Biomedicine and Neuroscience, University Of Bari Aldo Moro , Bari, Italy 7 U.O.C. Psichiatria Universitaria, Azienda Ospedaliero-Universitaria Consorziale Policlinico , Bari, Italy MD, PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Antonio Rampino Daniela Reich-Erkelenz 4 Institute of Psychiatric Phenomics and Genomics (IPPG), LMU University Hospital , Munich, Germany 11 Department of Psychiatry and Psychotherapy, LMU University Hospital , Munich, Germany MSc Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Daniela Reich-Erkelenz Eva C. Schulte 4 Institute of Psychiatric Phenomics and Genomics (IPPG), LMU University Hospital , Munich, Germany 11 Department of Psychiatry and Psychotherapy, LMU University Hospital , Munich, Germany 13 Department of Psychiatry and Psychotherapy, Faculty of Medicine and University Hospital Bonn , Bonn, Germany 14 Institute of Human Genetics, Faculty of Medicine and University Hospital Bonn , Bonn, Germany 15 German Center for Mental Health (DZPG), partner site Munich/Augsburg , Germany MD, PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Eva C. Schulte Fanny Senner 4 Institute of Psychiatric Phenomics and Genomics (IPPG), LMU University Hospital , Munich, Germany 11 Department of Psychiatry and Psychotherapy, LMU University Hospital , Munich, Germany 16 Centres for Psychiatry Suedwuerttemberg, Centres for Psychiatry Suedwuerttemberg , Ravensburg, Germany MD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Fanny Senner Leonardo Sportelli 8 Lieber Institute for Brain Development, Johns Hopkins Medical Campus , Baltimore, MD, USA PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Leonardo Sportelli Alessandro Bertolino 6 Department of Translational Biomedicine and Neuroscience, University Of Bari Aldo Moro , Bari, Italy 7 U.O.C. Psichiatria Universitaria, Azienda Ospedaliero-Universitaria Consorziale Policlinico , Bari, Italy MD, PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Alessandro Bertolino Peter Falkai 11 Department of Psychiatry and Psychotherapy, LMU University Hospital , Munich, Germany 12 Department Clinical Translation, Max Planck Institute of Psychiatry , Munich, Germany , Munich, Germany 15 German Center for Mental Health (DZPG), partner site Munich/Augsburg , Germany MD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Peter Falkai Urs Heilbronner 5 University Hospital of Psychiatry and Psychotherapy, University of Bern , Bern, Switzerland PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Urs Heilbronner Giulio Pergola 8 Lieber Institute for Brain Development, Johns Hopkins Medical Campus , Baltimore, MD, USA 17 Department of Psychiatry and Behavioral Sciences, Johns Hopkins University School of Medicine , Baltimore, MD, USA 18 Department of Genetic Medicine, Johns Hopkins University School of Medicine , Baltimore, MD, USA PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Giulio Pergola Thomas G. Schulze 4 Institute of Psychiatric Phenomics and Genomics (IPPG), LMU University Hospital , Munich, Germany 15 German Center for Mental Health (DZPG), partner site Munich/Augsburg , Germany 17 Department of Psychiatry and Behavioral Sciences, Johns Hopkins University School of Medicine , Baltimore, MD, USA 19 Department of Psychiatry and Behavioral Sciences, SUNY Upstate Medical University , NY, USA MD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Thomas G. Schulze Andreas Meyer-Lindenberg 1 Department of Psychiatry and Psychotherapy, Central Institute of Mental Health, Medical Faculty Mannheim, Heidelberg University , Mannheim, Germany MD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Andreas Meyer-Lindenberg Fabian Streit 1 Department of Psychiatry and Psychotherapy, Central Institute of Mental Health, Medical Faculty Mannheim, Heidelberg University , Mannheim, Germany 2 Hector Institute for Artificial Intelligence in Psychiatry, Central Institute of Mental Health, Medical Faculty Mannheim, Heidelberg University , Mannheim, Germany 20 German Center for Mental Health (DZPG), Partner Site Mannheim - Heidelberg - Ulm, Germany PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Fabian Streit Emanuel Schwarz 1 Department of Psychiatry and Psychotherapy, Central Institute of Mental Health, Medical Faculty Mannheim, Heidelberg University , Mannheim, Germany 2 Hector Institute for Artificial Intelligence in Psychiatry, Central Institute of Mental Health, Medical Faculty Mannheim, Heidelberg University , Mannheim, Germany 20 German Center for Mental Health (DZPG), Partner Site Mannheim - Heidelberg - Ulm, Germany PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Emanuel Schwarz For correspondence: emanuel.schwarz{at}zi-mannheim.de Abstract Full Text Info/History Metrics Supplementary material Data/Code Preview PDF Abstract Importance Schizophrenia is clinically and biologically heterogeneous, with marked variability in course and treatment response. Stratification of patients may advance individualized therapy and clarify underlying mechanisms. Objective To identify, validate, and replicate genetic risk factors distinct to each other in schizophrenia patients using polygenic risk scores of large numbers of psychiatry-relevant phenotypes. Design We analyzed genetic and phenotypic data from FinnGen (discovery dataset) and two independent cohorts for validation (PsyCourse, Bari). Using PRScope, a framework for standardizing polygenic score calculation and for patient stratification, we calculated 413 psychiatry-related polygenic scores for individuals with schizophrenia and controls. The resulting multi-PGS matrix was used to stratify patients through a data-driven approach. The findings were validated via cross-validation in FinnGen, and replicated in PsyCourse and Bari. Setting We used available large-scale datasets containing genetic and phenotypic information and publicly available GWAS summary statistics. Participants Data from FinnGen (7,486 schizophrenia cases; 27,288 controls), PsyCourse (419 cases; 299 controls), and Bari (531 cases; 738 controls) were included in this study. Main outcome and measures We examined the validity of the genetic differences among patients, predictive accuracy across datasets, contributing genetic domains, and phenotypic differences, including clinical severity proxies. Results Two data-driven clusters of patients with differing genetic risk profiles were identified. Both showed comparable genetic liability for schizophrenia, but diverged strongly in genetic risk for multiple psychiatry-related traits. The profile of the first cluster was characterized by a higher risk of depression, neuroticism, and low cognitive performance, whereas the second cluster showed a profile similar to that of healthy controls in these dimensions. Despite equal genetic liability for schizophrenia, the first profile showed higher predictability for schizophrenia in a case–control prediction model and was associated with significant differences in clinical severity indicators, such as higher clozapine use. Conclusions and Relevance A data-driven, polygenic risk–based approach revealed two biologically and clinically distinct genetic risk profiles. Genetic liability for depression traits, neuroticism, and low cognition beyond schizophrenia liability itself, appears to shape disease penetrance and severity of schizophrenia. These findings highlight the potential of multivariate polygenic risk stratification for refining schizophrenia nosology and tailoring interventions. Question Can polygenic scores derived from many psychiatry-relevant GWAS be used to stratify schizophrenia patients in a biologically and clinically meaningful way? Findings We identified two clusters of schizophrenia patients and characterized them as distinct genetic risk profiles based on their polygenic risk contributions. While both profiles showed comparable risk for schizophrenia, one profile was associated with higher liability for neuroticism and depression, reduced cognitive performance, and more complex clinical manifestations compared with patients in the second profile showing the opposite characteristics. Meaning Our results provide replicable insights into the genetic architecture of schizophrenia and have the potential to inform future personalized treatments. Introduction Schizophrenia (SCZ) is a severe, usually chronic mental disorder with a lifetime prevalence estimate of up to 1% and an incidence peak in early adulthood 1 . Its severity is underscored by a life expectancy approximately 15-years shorter than the general population 2 and a mere 10-20% competitive employment rate 1 , 3 , 4 . SCZ is a phenotypically complex disease, and its individually variable manifestations span multiple domains like perception, affect, cognition, behavior, speech, motor skills, and autonomic regulation. Furthermore, the complexity and heterogeneity of SCZ extend to treatment response with approximately one-third of patients responding poorly 5 , 6 . Flattening of affect, avolition and cognitive decline (collectively negative symptoms), are especially persistent 7 , 8 . Formulations of sub-classification of SCZ have been abandoned in the latest editions of the commonly used diagnostic manuals (ICD-11 and DSM-V) due to a lack of reliability and clinical utility 9 . Evidence suggests that SCZ encompasses a multitude of overlapping neurobiological processes 10 , 11 . Identifying more homogenous subgroups in terms of these processes would advance the etiological understanding of the disorder and personalized medicine approaches 12 . In family and twin studies, heritability for SCZ has been estimated at 60%-80% overall 13 – 15 , among the highest for psychiatric disorders, and the phenotypic and biological heterogeneity of SCZ has been hypothesized to depend in part on genetic heterogeneity 16 – 18 . Therefore, the investigation of genetic variation is a promising basis for understanding the source of that heterogeneity through stratification 1 , 19 , 20 . Genome-wide association studies (GWAS) of SCZ indicate that it is a highly polygenic disorder with thousands of common variants (i.e., single-nucleotide polymorphisms (SNPs)) contributing to SCZ liability, each exerting a small effect on overall disorder risk 21 , 22 . It is estimated that around 25% of liability to SCZ is attributable to the total contribution of common genetic variation (i.e., SNP-heritability). To estimate the genetic predisposition of an individual to a disease or a trait, the effects conferred by individual SNPs can be summed up into polygenic scores (PGS) (also called polygenic risk scores (PRS)) for diseases and a wide variety of traits. These indices of polygenic predisposition show robust predictive performance in independent target samples, but they currently fall short of explaining disease liability to the degree suggested by SNP-heritability estimates 19 , 23 . Therefore, considered together with moderate-to-high genetic correlations with other psychiatric disorders (e.g., r g =0.7 with bipolar disorder (BIP) and r g =0.35 with depression) 15 , 22 and related traits (e.g., r g =0.21 with Neuroticism) 23 , 24 , an approach based on a broad spectrum of PGS seems promising. Furthermore, SCZ PGS are primarily designed to distinguish cases from controls, a limitation potentially remedied by a multi-PGS strategy in case-case discrimination. While the PGS of SCZ combined with additional psychiatric PGS with known associations with SCZ have been used for subtyping 25 , and a data-driven multi-PGS approach has demonstrated superior performance in predicting SCZ case-control status over single-PGS models 26 , the latter strategy has not been applied to characterize patients based on distinct genetic risk profiles (GRPs) before. Building on this recent research, we aimed to: (i) identify distinct GRPs in SCZ by integrating large numbers of PGS of psychiatry-relevant phenotypes using robust, reproducible machine learning (ML) methods; (ii) validate these distinct GRPs in independent datasets; (iii) characterize the SCZ patients showing these GRPs in terms of clinical manifestations. To achieve this, we developed PRS cope, a framework that generates PGS from a broad set of GWAS and then uses the resulting multi-PGS matrix to identify patient clusters through a data-driven approach and characterise them as GRPs based on their PGS compositions. Our analysis included data from a total of 8,436 patients with SCZ, and 28,325 healthy controls (HC) from three independent datasets, to identify and validate GRPs. In addition, we used data from 7,555 patients with BIP, and from 27,187 patients with Major Depressive Disorder (MDD), and from 37,539 HCs for a transdiagnostic analysis. Our method offers a novel approach to deconstructing the genetic heterogeneity of complex human disorders and may contribute to the personalization of medical care in SCZ and other disorders. Methods Overview The methods used can be separated into the following steps: I) Automated GWAS search, II) Quality control (QC) of the GWAS, III) QC of the genotype data in the target datasets, IV) Calculation of the corresponding PGS in the target datasets V) Clustering of patients based on multi-PGS similarities in the training dataset, VI) Predicting the cluster assignments in the validation datasets, VII) Phenotypic profiling, VIII) Genetic profiling. Steps I) - IV) were done using PRScope, and steps V) - VIII) were done by the ML framework implemented for phenotype or disease stratification (for a schematic overview of the Methods, see Figure 1 ; for more details, see eMethods in the Supplement). Download figure Open in new tab Figure 1. Overview of the PRScope framework and analysis pipeline. a) Psychiatry-related GWAS, annotated with relevant ontologies and fulfilling our filtering criteria, were downloaded from the GWAS Catalog and pre-processed for the subsequent PGS calculation. b) Genotype data from the three cohorts were pre-processed. PGS calculations for 413 pre-processed GWAS summary statistics were performed for each GWAS summary statistic using Bayesian-based PGS method LDPred2. c) ML was applied to identify two data-driven clusters with differing PGS profiles in large-scale training dataset. These GRPs were then validated in two independent datasets and through cross-validation in the discovery dataset. Definition of Cluster : Predicted cluster memberships derived from cluster-specific models include both HC and SCZ subjects. Definition of GRP : SCZ patients uniquely correctly predicted by these models. (i) Ranking of PGS with regard to their absolute correlation with SCZ. The top 45 PGS that showed the strongest correlation with SCZ were included in ML analysis (from 1 to 45, stepwise). To prevent data leakage during cross-validation in the training dataset, feature selection was performed separately for each cross-validation iteration. For validation, features were extracted using the entire training dataset. (ii) Clustering of SCZ patients based on a multi-PGS profile into two clusters via k-means clustering method using the selected features. The PGS of SCZ patients for a given GWAS were adjusted by subtracting the median of the HC and dividing by the standard deviation of the HC. (iii) Random forest ML was applied to train models for each cluster separately. (iv) Cluster memberships were predicted by the model in independent data and cross-validation test folds. d) The characterization of the multi-PGS signature of each GRP. (i) Associations between the predicted cluster memberships and diagnoses were quantified as odds ratios and significance was determined using two-sided Fisher’s exact test. (ii) The beta coefficients obtained from logistic regression represent the associations between each of the 45 most SCZ-relevant PGS and GRP relative to HC. The table shows the PGS categories identified based on genetic correlation among GWAS. SCZ category included only SCZ PGS. Bipolar disorder, personality trait and alcohol related traits were not included in GRP characterization. (iii) PCs were computed separately for each GRP using all PGS within a PGS category, and the mean values of the first PCs were plotted for each GRP and feature size. e) The k-means clustering solution for a given feature size was compared with the clustering results of other feature sizes and visualized as a heatmap. f) To support the clinical validity of the identified GRPs, differences in clinical features were assessed across the investigated datasets. PRScope PRScope extracts GWAS from the GWAS Catalog 27 based on a predefined selection of psychiatry-relevant categories specified by a list of Experimental Factor Ontology 28 (EFO) ids, which were provided as input to the pipeline. The harmonized GWAS fulfilling the filtering criteria and annotated with the provided EFOs were downloaded from GWAS Catalog and quality controlled, resulting in 348 GWAS. Furthermore, 65 GWAS, not listed in the GWAS Catalog but accessible from Psychiatric Genomics Consortium (PGC), Complex Trait Genetics (CTG), and other public sources, were identified and included in the analysis, using the same harmonization pipeline 29 as applied by the GWAS Catalog. The corresponding PGS were then calculated for each of the subjects in the target genotype datasets using the Bayesian-based PGS method LDPred-2 30 . The application of PRScope resulted in quality-controlled PGS for 413 psychiatry-relevant phenotypes that were computed in three cohorts (Train FinnGen , Val PsyCourse , Val Bari ) ( Figure 1a , 1b ). Machine learning framework ML was applied to identify and validate subgroups of patients with SCZ characterized by distinct multi-PGS profiles. The analytical pipeline involved: (a) ranking of PGS with regard to their absolute correlation with SCZ; (b) clustering of SCZ patients based on multi-PGS profile into two clusters via k-means clustering using the most relevant PGS; (c) training a separate random forest 31 model for each cluster to distinguish SCZ cases from HC; and (d) predicting cluster membership across cross-validation folds on Train FinnGen and on validation datasets (models trained on Train FinnGen , and validated on Val PsyCourse and Val Bari ), resulting in one prediction from Cluster 1 model and one from Cluster 2 model for each dataset ( Figure 1c ). These predictions were then used to evaluate the predictive accuracy of each model. SCZ patients predicted as patients by exactly one of the models were assigned to the corresponding group (hereafter referred to as GRP) and the GRPs were characterized based on their polygenic risk contributions, as well as psychopathological measures (Val Bari ), comorbid disorders and medication information (Train FinnGen ), and selected phenotypes (Val PsyCourse ). The whole procedure was applied for each selected feature size separately, starting with the highest ranked PGS, adding PGS based on their ranking, resulting in a gradually increasing feature size ( Figure 1d ). For comparison, an alternative scheme where we group patients according to their predicted probabilities of having schizophrenia, as derived from prediction models trained on the same PGS as in our primary approach, was applied, hereafter referred to as composite PGS . Genetic correlations among GWAS were used to categorize the 45 most SCZ-relevant PGS into meaningful PGS categories. Finally, this analysis was complemented by conjunctional False Discovery Rate analysis 32 , 33 to investigate whether genetic susceptibility for such phenotypes was mediated by shared genetics between PGS categories. Cohorts Data from FinnGen release 12 (Train FinnGen ) (7,486/27,288 cases/controls for SCZ, 7,555/11,198 for BIP, 27,187/26,993 for MDD. Controls for BIP and MDD are partly shared, but not with SCZ controls) were used for the identification of multi-PGS profiles, and to test associations. Data from the PsyCourse Study 34 (Val PsyCourse ) (419/299 cases/controls for SCZ) and Bari cohort 21 (Val Bari ) (531/738 cases/controls for SCZ) were used for GRP validation. All cohorts were independent of the original GWAS the PGS were derived from. Results The PRScope procedure was applied to data from the Train FinnGe 35 to identify two data-driven clusters of SCZ. Separate ML models were then trained to characterize each of these clusters as GRPs and enable their prediction in independent test data. The predictive value of these cluster models, and their dependence on the number of included PGS, was first tested in Train FinnGen using a cross-validation approach ( Figure 2a , for uncorrected p-values see eTable 1). Subsequently, these models were applied to the independent test data, illustrating that GRPs could be successfully validated at the selected feature sizes ( Figure 2a ). Similarly to Train FinnGen cross-validation, in both validation datasets (Val PsyCourse and Val Bari ), we observed that both GRP 1 and 2 were significantly positively associated with SCZ from feature size 5 onwards (p < rejection threshold of the Benjamini-Hochberg (BH) procedure, see for untransformed p-values and predicted GRP proportions eTable 2-6), although for Val PsyCourse the p-value exceeded 0.05 for the GRP 2 model at feature size 24 ( Figure 2a ; shown as a dip due to the negative log transformation). The GRPs did not differ significantly in terms of age or sex in any of the three datasets. Download figure Open in new tab Figure 2. Results of GRP validation. a) Associations between the predicted cluster memberships and diagnoses were quantified as odds ratios and significance were determined using two-sided Fisher’s exact test where the input was a cross table of either GRP 1 or GRP 2 model and SCZ. The raw p-values were -log10 transformed to increase the readability of the plot. Feature sizes from 1 to 45 (x-axis) were plotted to demonstrate the formation with increasing feature size. An OR larger than 1 (dashed line) means that a prediction of having SCZ by a given model increases the odds of truly having SCZ. The dashed lines on the p-value plots represent the rejection threshold of the BH procedure (number of tests = 26). The point symbols on the p-value plots indicate the significance of the odds ratios based on a BH p-value threshold. b) The PGS compositions of the GRPs were illustrated for the three PGS categories (neuroticism, depression, and cognition-related PGS categories) and for SCZ PGS. The mean of the first PC (to ensure comparability across units of analysis, all measures were scaled) of each PGS category was plotted across the feature sizes from 1 to 45 (x-axis) for GRP 1, GRP 2. The mean values were adjusted such that HC were centered at zero. With increasing feature size, the composition of the GRPs changed ( Figure 2b ), and we observed a substantial reallocation of patients, which started leveling off with approximately feature size 12 ( Figure 3 ). Thus, in order to test clinical associations with the GRPs, we decided to predict the GRPs at feature sizes (1-6, 12, 24, 33, 45) that represent the points of a high rate of reallocation (see eMethods) into two validation datasets (Cohorts’ demographics; eTable 7-9). Download figure Open in new tab Figure 3. PGS composition of the GRPs and stability across the feature sizes. The polar plot was generated using the FinnGen cohort’s association results for the features size 45. To enable comparison across PGS categories, the PGS categories were plotted sequentially. The legend (in a form of a table) of the polar plot shows the ranked order of the 45 most SCZ-relevant PGS. The beta coefficients obtained from logistic regression represent the associations between each PGS and GRP relative to HC. Adjusted rand index (ARI) comparisons were generated using the FinnGen cohort’s k-means clustering results and plotted as heatmap. ARI was calculated directly using the whole sample prediction results. To enable comparison across features, 0.7 is selected as the threshold for high similarity. Detailed information about the 45 GWAS listed in the table can be found in eTable 14. The values used to create the polar plot can be found in eTable 15. The ARI values used to generate the heatmap can be found in eTable 16. To assess the merits of our clustering approach, we additionally trained models simply predicting SCZ in the FinnGen data with the same PGS features we use in PRScope and used the median split of the predicted probabilities in CV test folds as groups that correspond to the GRP grouping. This alternative approach will be referred to as the composite PGS (see eMethods). The contribution of neuroticism, depression and cognitive performance-related predisposition to SCZ susceptibility is mediated by shared genetic variation The top 45 selected features were grouped using pairwise genetic correlation analysis of the corresponding GWAS, resulting in 4 categories of PGS which can be characterized as relating to SCZ, cognition, depression and neuroticism (see eFigure 1,2, and eMethods Formation of the PGS categories ). Figure 2 illustrates that the two GRPs were (at higher feature sizes) defined by a divergence in these PGS categories while their overall SCZ risk was very similar. Specifically, GRP 1 showed markedly elevated neuroticism and depression related PGS and relatively low cognition related PGS, whereas GRP 2 displayed an opposite pattern for neuroticism and depression related PGS and intermediate levels (between GRP 1 and HC) for the cognition related PGS. The composite PGS approach shows a somewhat similar pattern of genetic risk, with the predicted high risk group showing higher levels of SCZ, depression and neuroticism, and lower cognition PGS. However, the differences are markedly smaller except for SCZ PGS, for which it is much larger (eFigure 3). Since SCZ has well-established genetic correlations with neuroticism, depression, and cognition 21 , 36 , we investigated whether these aspects were collectively reflected as disproportional risk contributions of pleiotropic SNPs. For this, we constructed a subset PGS based on SNPs that were associated with SCZ and each of the PGS categories, using conjunctional FDR 32 , 33 (conjFDR). All SNP weights were taken from the SCZ summary statistic. We could show that all three of these subset PGS predicted SCZ better than PGS constructed of p-value matched (as indicated in the SCZ GWAS) SNPs outside their respective pleiotropy SNP group ( Figure 4 ). eTable 10-12 show the pleiotropic variants as identified by conjFDR. Download figure Open in new tab Figure 4. Predictive performance of the conjFDR-based PGS in Train FinnGen . The diamonds indicate the odds ratios for a PGS constructed with conjunctional FDR with a given PGS category (see Methods) in a logistic regression where true SCZ status was the dependent variable. The boxplots represent estimates from similar models with different PGSs serving as predictors. Each PGS was constructed on a p-value-matched random sample of SCZ risk SNPs. The p-values are derived from Cochran’s Q test of heterogeneity of effects, where the maximum estimate and the corresponding standard error from the 100 respective comparative PGSs are contrasted with the conjFDR PGS. Genetic risk profiles have transdiagnostic relevance and differ in clinical measures To clinically characterize the identified GRPs, differences in clinical features were assessed across the investigated datasets. All phenotypic results are shown for feature size 24 only for simplicity of illustration (for other feature sizes see eFigure 4-19). In Train FinnGen , GRP 1 was associated with a more severe and complex clinical manifestation in patients with SCZ as indicated by higher rates of antidepressant prescriptions and comorbid psychiatric diagnoses; including depression, neurotic (F40-F48 in ICD10), personality, and eating disorders ( Figure 5 ). Importantly, the composite PGS comparison showed no significant differences in comorbidity rates (eFigure 20-30, and eTable 13 for group sizes). Download figure Open in new tab Figure 5. Phenotypic profiling of GRP 1 and GRP 2. Feature size = 24. Baseline: GRP 2. All comparisons are adjusted for the SCZ PGS, sex, and age. In Val PsyCourse and Val Bari additionally adjusted for center/study; all results with the 95% CI (displayed) not overlapping the null effect are also significant with an FDR of 0.05 applied within each category (for comorbid disorders only t e significant ones are shown); analyzed case numbers of GRP 1 and GRP 2 are listed in brackets (n GRP 2 / n GRP 1 ): Clozapine (105/88), Antidepressants (105/88), PANSS = Positive and Negative Syndrome Scale, positive, negative, general and total score (Val PsyCourse 108/108, Val Bari 89/95). CGI = Clinical global impression (109/108), GAF = Global Assessment of Functioning (108/108), Neuroticism = Dimension from Personality (87/76), IDS-C 30 = Inventory of depressive symptomatology (109/103), BDI-II = Beck Depression Inventory-II (98/93), DST = Digit-Symbol-Test (105/103), Digit s. fwd = Verbal digit span forward (108/105), Digit s. bwd = Verbal digit span backward (108/105), TMT B-A = Trail making test part B - part A (105/197), MWT-B = Multiple-Choice Vocabulary Intelligence Test (74/74). For all FinnGen analyses, the corresponding group sizes are (2696/3077). Both in Val PsyCourse and Val Bari the statistical power of the phenotypic analysis was substantially lower than in Train FinnGen due to a far smaller sample size. Both in Val PsyCourse and Val Bari the Positive and Negative Syndrome Scale (PANSS) showed no significant differences in any of the scales ( Figure 5 ). In Val PsyCourse we further investigated a collection of phenotypic variables representing the major risk components of the GRPs as represented by the PGS categories defining them. We found a significantly (FDR corrected p-values < 0.05) higher score for GRP 1 in the Inventory of depressive symptomatology (IDS-C 30 ) ( Figure 5 ). In Train FinnGen , GRP 1 model predictions were also associated with BIP (n cases = 7,555; n controls = 11,198) and MDD (n cases = 27,187; n controls = 26,993) in a gradient-dependent manner, with the highest OR observed for SCZ, intermediate for BIP, and the lowest for MDD. The corresponding associations for GRP 2 model predictions followed a slightly different pattern, where the associations with both BIP and DEP were negative, with no consistent difference in magnitude across feature sizes (eFigure 15). Discussion Here, we aimed to capture distinct GRPs within the SCZ patients through the simultaneous investigation of patients’ susceptibility to psychiatry-relevant phenotypes using 413 publicly available GWAS, leveraging a massive training dataset (Train FinnGen ). We asked whether finding two genetically distinctive clusters of SCZ patients might help to identify phenotypically distinct and clinically relevant subgroups and lead to a better understanding of the genetic risk architecture of SCZ. This resulted in two distinct GRPs with similar SCZ risk; one GRP was characterized by high PGS of phenotypes relating to neuroticism and depression and low PGS for cognitive traits (GRP 1), whereas the other GRP (GRP 2) had a profile similar to HC in those dimensions. While the GRPs exhibited approximately the same elevated levels of genetic liability for SCZ, a random forest ML model based on GRP 1 clearly outperformed a model based on GRP 2 in predicting SCZ in Train FinnGen and in independent validation datasets (Val PsyCourse and Val Bari ). Furthermore, patients with a GRP 1 tended to have clinical manifestations that indicate higher rates of disease complexity (psychiatric comorbidity), a clinical profile which does not show up when grouping the patients along their SCZ PGS alone. Interestingly, the two groups did not differ in regards to phenotypic measures which are commonly used to stratify SCZ patients (e.g., PANSS score / subscores), indicating their potential value beyond a phenotypical or clinical assessment of patients. While previous studies have demonstrated that disease prediction can be improved with a multi-PGS approach 26 , this is the first time it has been applied to data-driven clustering of patients. The observation that the model based on GRP 1 is a considerably better predictor of SCZ than the one based on GRP 2, despite the similar SCZ risk, has several potential explanations that are not mutually exclusive. First, the genetic risk contributions for neuroticism, depression-related traits, and low cognitive performance may reflect biological mechanisms facilitating the actualization of disease risk. Second, the corresponding genetic risk may be only partially captured by SCZ PGS. Third, SCZ genetic risk estimation may also be more accurate in GRP 1 than in GRP 2 patients. In this scenario, the higher levels of polygenic risk of non-SCZ traits/disorders in GRP 1 would suggest a higher prevalence of alleles that confer risk to both SCZ and other related traits. These pleiotropic SNPs could be more often truly causal for SCZ than the ones only associated with SCZ risk, as the additional associations can be considered as an increase in the strength of statistical evidence. Our finding that the PGS based on conjFDR SNPs are slightly better predictors than PGS based on comparable subsets of SCZ risk SNPs lends support for this. Finally, phenotypic mediation, such as neuroticism being associated with higher stress levels, which in turn increase the risk of developing SCZ, could play a role 37 , 38 . The current study has several limitations. First, as one of the main goals of this study was reproducibility, and thus considering the sample size limitations of the validation datasets, we restricted the analysis to classify patients into two GRPs. This will likely not capture the entire complexity of SCZ risk mediated by common variants. It should also be noted that the observed divergence in GRPs can arise from the correlation structure (a result of the univariate feature selection) of the input variables to the clustering algorithm, even without the existence of truly separate clusters. Second, our method only considered genetic predisposition to psychiatry-related traits. It is possible that the including of genetic risk in somatic domains would improve the classification and provide more insight into underlying biological mechanisms 39 – 41 . Third, the heterogeneity between the datasets we used in terms of the availability of phenotypic data hampered us from conducting more harmonized analyses in that regard, drawing less definite conclusions about generalizability. On the other hand, the fact that our procedure works across such different datasets can be seen as an indication of good portability. Finally, as only samples of European descent and GWAS from that ancestry were used, the findings presented in this paper may not generalize to individuals with non-European ancestry 42 , 43 . In summary, our data-driven ML scheme successfully leveraged a large number of psychiatry-related PGS, for the first time, to derive two GRPs of SCZ patients. A prediction model based on GRP 1, characterized by high neuroticism and depression related traits and low cognitive functioning, is a promising step toward identifying SCZ patients with a poor prognosis at an early stage. Especially due to the GRPs not being linked to common clinical measures (e.g., clinical questionnaires like the PANSS), they may also prove useful in personalizing treatments, a goal which a stratification based on clinical measures so far failed to accomplish despite multiple attempts 44 , 45 . A prospect that could be evaluated in a randomized and controlled clinical trial. Taken together, this work establishes a framework for integrating multi-phenotype PGS with ML approaches for the genetic stratification of complex disorders. While demonstrated here in SCZ, the framework can be applied to other psychiatric and somatic conditions. Data Availability For FinnGen, researchers can apply for health data from the Finnish Data Authority Findata (https://findata.fi/en/permits/) and individual-level genotype data available through the Fingenious portal (https://site.fingenious.fi/en/). These resources are hosted by the Finnish Biobank Cooperative FINBB (https://finbb.fi/en/). Access can only be provided for research projects within the scope of the Finnish Biobank Act, which includes health promotion, understanding disease mechanisms or developing medical products or treatment practices. PsyCourse data is available to bona fide reseachers upon a submission and approval of a research proposal. Individual genotype and clinical data from the University of Bari Aldo Moro with demographic and behavioral characteristics cannot be shared at individual level in raw format because of ethic restrictions. http://www.psycourse.de/openscience-de.html https://docs.finngen.fi/ Funding This work was supported by the Hector foundation II, the German Federal Ministry of Education and Research (BEST project, grant 01EK2101B), the German Research Foundation (DFG) (TRR 265/2 TP A06), and was endorsed by German Center for Mental Health (DZPG). U.H. was supported by European Union’s Horizon 2020 Research and Innovation Programme (PSY-PGx, grant agreement No 945151) and the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation, project number 514201724). T.G.S. was supported by the European Union Horizon 2020 Research and Innovation Program (PSY-PGx, grant agreement No 945151), and also by the Deutsche Forschungsgemeinschaft within the framework of the projects www.kfo241.de and www.PsyCourse.de [SCHU 1603/4–1, 5–1, 7–1]. T.G.S. was further supported by the Dr Lisa Oehler Foundation (Kassel, Germany), the Bundesministerium für Bildung und Forschung (BMBF, Federal Ministry of Education and Research; projects: IntegraMent [01ZX1614K], BipoLife [01EE1404H], e:Med Program [01ZX1614K]) and European Union’s Horizon 2020 Research and Innovation Programme (ERA-NET Neuron Projects GEPI-BIOPSY [BMBF No 01EW2005] and MulioBio [BMBF No 01EW2009]). G.C.K.’s PhD scholarship is financially supported by Exprivia S.p.A. under the Italian ministerial decree D.M. 351. Furthermore, this study was supported by FAIR - Future AI Research (PE00000013), spoke 6 – Symbiotic AI, under the NRRP MUR program funded by the NextGenerationEU, awarded to L.A.A. and G.P.; Project CUP : H97G22000210007, and the Apulian regional government for the project: Early Identification of Psychosis Risk, awarded to A.B. Conflicts of Interest G.B. has received lecture and consultant fees by Lundbeck. A.R. has received lecture fees by Janssen. A.B. received consulting fees from Biogen and lecture fees from Otsuka, Janssen, and Lundbeck. G.P. and L.A.A have received lecture fees by Lundbeck. A.M.-L. has received consultant fees from: Agence Nationale de la Recherche, Brain Mind Institute, Brainsway, CISSN (Catania International Summer School of Neuroscience), Daimler und Benz Stiftung, Fondation FondaMental, Hector Stiftung II, Janssen-Cilag GmbH, Lundbeck A/S, Lundbeckfonden, Lundbeck Int. Neuroscience Foundation, MedinCell, Sage Therapeutics, Techspert.io, The LOOP Zürich, University Medical Center Utrecht, von Behring Röntgen Stiftung. A.M.-L.has received speaker fees from: Ärztekammer Nordrhein, BAG Psychiatrie Oberbayern, Biotest ÄG, Forum Werkstatt Karlsruhe, International Society of Psychiatric Genetics, Brentwood, Klinik für Psychiatrie und Psychotherapie Ingolstadt, Lundbeck SAS France, med Update GmbH, Merz-Stiftung, Siemens Healthineers, Society of Biological Psychiatry. A.M.-L. has received editorial fees from: American Association for the Advancement of Science, Elsevier, Thieme Verlag. E.S. received speaker fees from bfd buchholz-fachinformationsdienst GmbH and Lundbeckfonden as well as editorial fees from Lundbeckfonden and the Wellcome Trust. The other authors have declared that there are no conflicts of interest in relation to the subject of this study. Acknowledgments We thank all research participants and all researchers and clinicians who collected, generated, or processed the data used in this study. Footnotes ↵ # Joint senior authors Revisions were made to the paragraph structure of the abstract. References 1. ↵ Marder SR , Cannon TD. Schizophrenia . N Engl J Med. 2019 ; 381 ( 18 ): 1753 – 1761 . doi: 10.1056/NEJMra1808803 OpenUrl CrossRef PubMed 2. ↵ Chang CK , Hayes RD , Perera G , et al. Life expectancy at birth for people with serious mental illness and other major disorders from a secondary mental health care case register in London . PLoS One . 2011 ; 6 ( 5 ): e19590 . doi: 10.1371/journal.pone.0019590 OpenUrl CrossRef PubMed 3. ↵ Fleischhacker WW , Arango C , Arteel P , et al. Schizophrenia--time to commit to policy change . Schizophr Bull . 2014 ; 40 Suppl 3 (Suppl 3): S165 – S194 . doi: 10.1093/schbul/sbu006 OpenUrl CrossRef PubMed 4. ↵ Kahn RS , Sommer IE , Murray RM , et al. Schizophrenia . Nat Rev Dis Primers . 2015 ; 1 : 15067 . doi: 10.1038/nrdp.2015.67 OpenUrl CrossRef PubMed 5. ↵ Fusar-Poli P , Papanastasiou E , Stahl D , et al. Treatments of negative symptoms in schizophrenia: Meta-analysis of 168 randomized placebo-controlled trials . Schizophr Bull . 2015 ; 41 ( 4 ): 892 – 899 . doi: 10.1093/schbul/sbu170 OpenUrl CrossRef PubMed 6. ↵ Potkin SG , Kane JM , Correll CU , et al. The neurobiology of treatment-resistant schizophrenia: paths to antipsychotic resistance and a roadmap for future research . NPJ Schizophr . 2020 ; 6 ( 1 ): 1 . doi: 10.1038/s41537-019-0090-z OpenUrl CrossRef PubMed 7. ↵ Lincoln TM , Dollfus S , Lyne J . Current developments and challenges in the assessment of negative symptoms . Schizophr Res . 2017 ; 186 : 8 – 18 . doi: 10.1016/j.schres.2016.02.035 OpenUrl CrossRef PubMed 8. ↵ Dollfus S , Lyne J . Negative symptoms: History of the concept and their position in diagnosis of schizophrenia . Schizophr Res . 2017 ; 186 : 3 – 7 . doi: 10.1016/j.schres.2016.06.024 OpenUrl CrossRef PubMed 9. ↵ Braff DL , Ryan J , Rissling AJ , Carpenter WT . Lack of use in the literature from the last 20 years supports dropping traditional schizophrenia subtypes from DSM-5 and ICD-11 . Schizophr Bull . 2013 ; 39 ( 4 ): 751 – 753 . doi: 10.1093/schbul/sbt068 OpenUrl CrossRef PubMed Web of Science 10. ↵ Owen MJ . New approaches to psychiatric diagnostic classification . Neuron . 2014 ; 84 ( 3 ): 564 – 571 . doi: 10.1016/j.neuron.2014.10.028 OpenUrl CrossRef PubMed 11. ↵ Clementz BA , Sweeney JA , Hamm JP , et al. Identification of distinct psychosis biotypes using brain-based biomarkers . Am J Psychiatry . 2016 ; 173 ( 4 ): 373 – 384 . doi: 10.1176/appi.ajp.2015.14091200 OpenUrl CrossRef PubMed 12. ↵ Millan MJ , Andrieux A , Bartzokis G , et al. Altering the course of schizophrenia: progress and perspectives . Nat Rev Drug Discov . 2016 ; 15 ( 7 ): 485 – 515 . doi: 10.1038/nrd.2016.28 OpenUrl CrossRef PubMed 13. ↵ Hilker R , Helenius D , Fagerlund B , et al. Heritability of schizophrenia and schizophrenia spectrum based on the nationwide Danish Twin Register . Biol Psychiatry . 2018 ; 83 ( 6 ): 492 – 498 . doi: 10.1016/j.biopsych.2017.08.017 OpenUrl CrossRef PubMed 14. Cardno AG , Marshall EJ , Coid B , et al. Heritability estimates for psychotic disorders: the Maudsley twin psychosis series . Arch Gen Psychiatry . 1999 ; 56 ( 2 ): 162 – 168 . doi: 10.1001/archpsyc.56.2.162 OpenUrl CrossRef PubMed Web of Science 15. ↵ Lichtenstein P , Yip BH , Björk C , et al. Common genetic determinants of schizophrenia and bipolar disorder in Swedish families: a population-based study . Lancet . 2009 ; 373 ( 9659 ): 234 – 239 . doi: 10.1016/S0140-6736(09)60072-6 OpenUrl CrossRef PubMed Web of Science 16. ↵ Kendler K . What psychiatric genetics has taught us about the nature of psychiatric illness and what is left to learn . Mol Psychiatry . 2013 ; 18 : 1058 – 1066 . doi: 10.1038/mp.2013.50 OpenUrl CrossRef PubMed Web of Science 17. Alnæs D , Kaufmann T , van der Meer D , et al. Brain heterogeneity in schizophrenia and its association with polygenic risk . JAMA Psychiatry . 2019 ; 76 ( 7 ): 739 – 748 . doi: 10.1001/jamapsychiatry.2019.0257 OpenUrl CrossRef 18. ↵ Pergola G , Penzel N , Sportelli L , Bertolino A . Lessons learned from parsing genetic risk for schizophrenia into biological pathways . Biol Psychiatry . 2023 ; 94 ( 2 ): 121 – 130 . doi: 10.1016/j.biopsych.2022.10.009 OpenUrl CrossRef PubMed 19. ↵ Owen MJ , Legge SE , Rees E , Walters JTR , O’Donovan MC . Genomic findings in schizophrenia and their implications . Mol Psychiatry . 2023 ; 28 ( 9 ): 3638 – 3647 . doi: 10.1038/s41380-023-02293-8 OpenUrl CrossRef 20. ↵ Jablensky A . Subtyping schizophrenia: implications for genetic research . Mol Psychiatry . 2006 ; 11 ( 9 ): 815 – 836 . doi: 10.1038/sj.mp.4001857 OpenUrl CrossRef PubMed Web of Science 21. ↵ Trubetskoy V , Pardiñas AF , Qi T , et al. Mapping genomic loci implicates genes and synaptic biology in schizophrenia . Nature . 2022 ; 604 ( 7906 ): 502 – 508 . doi: 10.1038/s41586-022-04434-5 OpenUrl CrossRef PubMed 22. ↵ Legge SE , Santoro ML , Periyasamy S , Okewole A , Arsalan A , Kowalec K . Genetic architecture of schizophrenia: a review of major advancements . Psychol Med . 2021 ; 51 ( 13 ): 2168 – 2177 . doi: 10.1017/S0033291720005334 OpenUrl CrossRef 23. ↵ O’Donovan MC , Owen MJ . The implications of the shared genetics of psychiatric disorders . Nat Med . 2016 ; 22 ( 11 ): 1214 – 1219 . doi: 10.1038/nm.4196 OpenUrl CrossRef PubMed 24. ↵ Gupta P , Galimberti M , Liu Y , et al. A genome-wide investigation into the underlying genetic architecture of personality traits and overlap with psychopathology . Nat Hum Behav . 2024 ; 8 ( 11 ): 2235 – 2249 . doi: 10.1038/s41562-024-01951-3 OpenUrl CrossRef 25. ↵ Chen J , Mize T , Wu JS , et al. Polygenic risk scores for subtyping of schizophrenia . Schizophr Res Treatment . 2020 ; 2020 ( 1 ): 1638403 . doi: 10.1155/2020/1638403 OpenUrl CrossRef PubMed 26. ↵ Albiñana C , Zhu Z , Schork AJ , et al. Multi-PGS enhances polygenic prediction by combining 937 polygenic scores . Nat Commun . 2023 ; 14 ( 1 ): 4702 . doi: 10.1038/s41467-023-40330-w OpenUrl CrossRef PubMed 27. ↵ Sollis E , Mosaku A , Abid A , et al. The NHGRI-EBI GWAS Catalog: knowledgebase and deposition resource . Nucleic Acids Res . 2023 ; 51 ( D1 ): D977 – D985 . doi: 10.1093/nar/gkac1010 OpenUrl CrossRef 28. ↵ Malone J , Holloway E , Adamusiak T , et al. Modeling sample variables with an Experimental Factor Ontology . Bioinformatics . 2010 ; 26 ( 8 ): 1112 – 1118 . doi: 10.1093/bioinformatics/btq099 OpenUrl CrossRef PubMed Web of Science 29. ↵ Buniello A , MacArthur JAL , Cerezo M , et al. The NHGRI-EBI GWAS Catalog of published genome-wide association studies, targeted arrays and summary statistics 2019 . Nucleic Acids Res . 2019 ; 47 ( D1 ): D1005 – D1012 . doi: 10.1093/nar/gky1120 OpenUrl CrossRef PubMed 30. ↵ Privé F , Arbel J , Vilhjálmsson BJ . LDpred2: better, faster, stronger . Bioinformatics . 2021 ; 36 ( 22-23 ): 5424 – 5431 . doi: 10.1093/bioinformatics/btaa1029 OpenUrl CrossRef PubMed 31. ↵ Breiman L. Random Forests . Mach Learn . 2001 ; 45 ( 1 ): 5 – 32 . doi: 10.1023/a:1010933404324 OpenUrl CrossRef PubMed 32. ↵ Andreassen OA , Thompson WK , Schork AJ , et al. Improved detection of common variants associated with schizophrenia and bipolar disorder using pleiotropy-informed conditional false discovery rate . PLoS Genet . 2013 ; 9 ( 4 ): e1003455 . doi: 10.1371/journal.pgen.1003455 OpenUrl CrossRef PubMed 33. ↵ Smeland OB , Frei O , Shadrin A , et al. Discovery of shared genomic loci using the conditional false discovery rate approach . Hum Genet . 2020 ; 139 ( 1 ): 85 – 94 . doi: 10.1007/s00439-019-02060-2 OpenUrl CrossRef PubMed 34. ↵ Budde M , Anderson-Schmidt H , Gade K , et al. A longitudinal approach to biological psychiatric research: The PsyCourse study . Am J Med Genet B Neuropsychiatr Genet . 2019 ; 180 ( 2 ): 89 – 102 . doi: 10.1002/ajmg.b.32639 OpenUrl CrossRef 35. ↵ Kurki MI , Karjalainen J , Palta P , et al. FinnGen provides genetic insights from a well-phenotyped isolated population . Nature . 2023 ; 613 ( 7944 ): 508 – 518 . doi: 10.1038/s41586-022-05473-8 OpenUrl CrossRef PubMed 36. ↵ Wootton O , Shadrin AA , Bjella T , et al. Genomic insights into the shared and distinct genetic architecture of cognitive function and schizophrenia . Sci Rep . 2024 ; 14 ( 1 ): 15356 . doi: 10.1038/s41598-024-66085-y OpenUrl CrossRef PubMed 37. ↵ Schneider TR . The role of neuroticism on psychological and physiological stress responses . J Exp Soc Psychol . 2004 ; 40 ( 6 ): 795 – 804 . doi: 10.1016/j.jesp.2004.04.005 OpenUrl CrossRef Web of Science 38. ↵ Almuqrin A , Georgiades A , Mouhitzadeh K , Rubinic P , Mechelli A , Tognin S . The association between psychosocial stress, interpersonal sensitivity, social withdrawal and psychosis relapse: a systematic review . Schizophrenia (Heidelb) . 2023 ; 9 ( 1 ): 22 . doi: 10.1038/s41537-023-00349-w OpenUrl CrossRef PubMed 39. ↵ Uellendahl-Werth F , Maj C , Borisov O , et al. Cross-tissue transcriptome-wide association studies identify susceptibility genes shared between schizophrenia and inflammatory bowel disease . Commun Biol . 2022 ; 5 ( 1 ): 80 . doi: 10.1038/s42003-022-03031-6 OpenUrl CrossRef PubMed 40. Gong W , Guo P , Li Y , et al. Role of the gut-brain axis in the shared genetic etiology between gastrointestinal tract diseases and psychiatric disorders: A genome-wide pleiotropic analysis . JAMA Psychiatry . 2023 ; 80 ( 4 ): 360 – 370 . doi: 10.1001/jamapsychiatry.2022.4974 OpenUrl CrossRef PubMed 41. ↵ Tylee DS , Lee YK , Wendt FR , et al. An atlas of genetic correlations and genetically informed associations linking psychiatric and immune-related phenotypes . JAMA Psychiatry . 2022 ; 79 ( 7 ): 667 – 676 . doi: 10.1001/jamapsychiatry.2022.0914 OpenUrl CrossRef PubMed 42. ↵ Peterson RE , Kuchenbaecker K , Walters RK , et al. Genome-wide association studies in ancestrally diverse populations: Opportunities, methods, pitfalls, and recommendations . Cell . 2019 ; 179 ( 3 ): 589 – 603 . doi: 10.1016/j.cell.2019.08.051 OpenUrl CrossRef PubMed 43. ↵ Fatumo S , Chikowore T , Choudhury A , Ayub M. A roadmap to increase diversity in genomic studies . Nat Med. Published online 2022 . https://www.nature.com/articles/s41591-021-01672-4 44. ↵ Leucht S , Chaimani A , Krause M , et al. The response of subgroups of patients with schizophrenia to different antipsychotic drugs: a systematic review and meta-analysis . Lancet Psychiatry . 2022 ; 9 ( 11 ): 884 – 893 . doi: 10.1016/S2215-0366(22)00304-2 OpenUrl CrossRef PubMed 45. ↵ McCutcheon RA , Pillinger T , Mizuno Y , et al. The efficacy and heterogeneity of antipsychotic response in schizophrenia: A meta-analysis . Mol Psychiatry . 2021 ; 26 ( 4 ): 1310 – 1320 . doi: 10.1038/s41380-019-0502-5 OpenUrl CrossRef View the discussion thread. Back to top Previous Next Posted February 24, 2026. Download PDF Supplementary Material Data/Code Email Thank you for your interest in spreading the word about medRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following Development and validation of polygenic risk profiles of schizophrenia Message Subject (Your Name) has forwarded a page to you from medRxiv Message Body (Your Name) thought you would like to see this page from the medRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share Development and validation of polygenic risk profiles of schizophrenia Ersoy Kocak , Joonas Naamanka , Tobias Gradinger , Fiona Klaassen , Johannes Nitsche , Philipp Grotehusmann , Kristina Adorjan , Linda A. Antonucci , Giuseppe Blasi , Monika Budde , Piergiuseppe Di Palo , Maria Heilbronner , Gianluca C. Kikidis , Alba Navarro-Flores , Mojtaba Oraki Kohshour , Sergi Papiol , Alessandra Raio , Antonio Rampino , Daniela Reich-Erkelenz , Eva C. Schulte , Fanny Senner , Leonardo Sportelli , Alessandro Bertolino , Peter Falkai , Urs Heilbronner , Giulio Pergola , Thomas G. Schulze , Andreas Meyer-Lindenberg , Fabian Streit , Emanuel Schwarz , FinnGen medRxiv 2025.04.29.25326656; doi: https://doi.org/10.1101/2025.04.29.25326656 Share This Article: Copy Citation Tools Development and validation of polygenic risk profiles of schizophrenia Ersoy Kocak , Joonas Naamanka , Tobias Gradinger , Fiona Klaassen , Johannes Nitsche , Philipp Grotehusmann , Kristina Adorjan , Linda A. Antonucci , Giuseppe Blasi , Monika Budde , Piergiuseppe Di Palo , Maria Heilbronner , Gianluca C. Kikidis , Alba Navarro-Flores , Mojtaba Oraki Kohshour , Sergi Papiol , Alessandra Raio , Antonio Rampino , Daniela Reich-Erkelenz , Eva C. Schulte , Fanny Senner , Leonardo Sportelli , Alessandro Bertolino , Peter Falkai , Urs Heilbronner , Giulio Pergola , Thomas G. Schulze , Andreas Meyer-Lindenberg , Fabian Streit , Emanuel Schwarz , FinnGen medRxiv 2025.04.29.25326656; doi: https://doi.org/10.1101/2025.04.29.25326656 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Psychiatry and Clinical Psychology Subject Areas All Articles Addiction Medicine (568) Allergy and Immunology (863) Anesthesia (300) Cardiovascular Medicine (4435) Dentistry and Oral Medicine (444) Dermatology (382) Emergency Medicine (608) Endocrinology (including Diabetes Mellitus and Metabolic Disease) (1509) Epidemiology (15228) Forensic Medicine (30) Gastroenterology (1124) Genetic and Genomic Medicine (6597) Geriatric Medicine (668) Health Economics (997) Health Informatics (4536) Health Policy (1368) Health Systems and Quality Improvement (1613) Hematology (540) HIV/AIDS (1264) Infectious Diseases (except HIV/AIDS) (15916) Intensive Care and Critical Care Medicine (1103) Medical Education (623) Medical Ethics (146) Nephrology (667) Neurology (6599) Nursing (346) Nutrition (998) Obstetrics and Gynecology (1144) Occupational and Environmental Health (957) Oncology (3332) Ophthalmology (974) Orthopedics (369) Otolaryngology (420) Pain Medicine (436) Palliative Medicine (130) Pathology (663) Pediatrics (1693) Pharmacology and Therapeutics (691) Primary Care Research (711) Psychiatry and Clinical Psychology (5447) Public and Global Health (9230) Radiology and Imaging (2198) Rehabilitation Medicine and Physical Therapy (1370) Respiratory Medicine (1196) Rheumatology (593) Sexual and Reproductive Health (712) Sports Medicine (530) Surgery (712) Toxicology (99) Transplantation (289) Urology (265) (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'a004b745af4206f3',t:'MTc3OTU0NjE3OQ=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();