An improved polygenic score for Parkinson’s disease partly explains variable penetrance of genetic Parkinson’s disease

An improved polygenic score for Parkinson’s disease partly explains variable penetrance of genetic Parkinson’s disease | 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 An improved polygenic score for Parkinson’s disease partly explains variable penetrance of genetic Parkinson’s disease View ORCID Profile Sebastian Sendel , Zied Landoulsi , Katja Lohmann , Björn-Hergen Laabs , Meike Kasten , Eva-Juliane Vollstedt , Tatiana Usnich , Alexander Balck , Daniela Berg , Dheeraj Reddy Bobbili , Max Borsche , COURAGE-PD Consortium , Andre Franke , Henrike Hanssen , View ORCID Profile Emadeldin Hassanin , Andrew A. Hicks , Ulrike M. Krämer , Rejko Krüger , Gregor Kuhlenbäumer , Lara M. Lange , View ORCID Profile Wolfgang Lieb , Brit Mollenhauer , NCER-PD Consortium , Miriam Neis , Peter P. Pramstaller , Jannik Prasuhn , Eva Schaeffer , Manu Sharma , Meike Steinbach , Claudia Trenkwalder , Norbert Brüggemann , Ana Westenberger , Michael Wittig , Inke R. König , View ORCID Profile Patrick May , Christine Klein , Amke Caliebe doi: https://doi.org/10.1101/2025.02.28.25323076 Sebastian Sendel 1 Institute of Medical Informatics and Statistics, Christian-Albrechts-University of Kiel, University Hospital Schleswig-Holstein , Campus Kiel, Kiel, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Sebastian Sendel For correspondence: koch{at}medinfo.uni-kiel.de Zied Landoulsi 2 Luxembourg Centre for Systems Biomedicine, University of Luxembourg , Esch-sur-Alzette, Luxembourg 3 Luxembourg Institute of Health , Strassen, Luxembourg Find this author on Google Scholar Find this author on PubMed Search for this author on this site Katja Lohmann 4 Institute of Neurogenetics, University of Lübeck , Lübeck, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Björn-Hergen Laabs 5 Institute of Medical Biometry and Statistics, University of Lübeck , Lübeck, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Meike Kasten 4 Institute of Neurogenetics, University of Lübeck , Lübeck, Germany 6 Department of Psychiatry, University Hospital Schleswig-Holstein , Campus Lübeck, Lübeck, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Eva-Juliane Vollstedt 4 Institute of Neurogenetics, University of Lübeck , Lübeck, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Tatiana Usnich 4 Institute of Neurogenetics, University of Lübeck , Lübeck, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Alexander Balck 4 Institute of Neurogenetics, University of Lübeck , Lübeck, Germany 7 Department of Neurology, University Hospital Schleswig-Holstein , Campus Lübeck, Lübeck, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Daniela Berg 8 Department of Neurology, Christian-Albrechts-University of Kiel, University Hospital Schleswig-Holsten , Campus Kiel, Kiel, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Dheeraj Reddy Bobbili 2 Luxembourg Centre for Systems Biomedicine, University of Luxembourg , Esch-sur-Alzette, Luxembourg Find this author on Google Scholar Find this author on PubMed Search for this author on this site Max Borsche 4 Institute of Neurogenetics, University of Lübeck , Lübeck, Germany 7 Department of Neurology, University Hospital Schleswig-Holstein , Campus Lübeck, Lübeck, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Andre Franke 9 Institute of Clinical Molecular Biology, Christian-Albrechts-University of Kiel , Kiel, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Henrike Hanssen 7 Department of Neurology, University Hospital Schleswig-Holstein , Campus Lübeck, Lübeck, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Emadeldin Hassanin 2 Luxembourg Centre for Systems Biomedicine, University of Luxembourg , Esch-sur-Alzette, Luxembourg Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Emadeldin Hassanin Andrew A. Hicks 10 Institute for Biomedicine, Eurac Research , Bolzano, Italy Find this author on Google Scholar Find this author on PubMed Search for this author on this site Ulrike M. Krämer 11 Institute of Medical Psychology, Center of Brain , Behavior and Metabolism, University of Lübeck , Lübeck, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Rejko Krüger 2 Luxembourg Centre for Systems Biomedicine, University of Luxembourg , Esch-sur-Alzette, Luxembourg 3 Luxembourg Institute of Health , Strassen, Luxembourg 12 Parkinson Research Clinic, Centre Hospitalier de Luxembourg , Strassen, Luxembourg Find this author on Google Scholar Find this author on PubMed Search for this author on this site Gregor Kuhlenbäumer 8 Department of Neurology, Christian-Albrechts-University of Kiel, University Hospital Schleswig-Holsten , Campus Kiel, Kiel, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Lara M. Lange 4 Institute of Neurogenetics, University of Lübeck , Lübeck, Germany 7 Department of Neurology, University Hospital Schleswig-Holstein , Campus Lübeck, Lübeck, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Wolfgang Lieb 13 Institute of Epidemiology, Christian-Albrechts-University of Kiel , Kiel, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Wolfgang Lieb Brit Mollenhauer 14 Paracelsus Elena Clinic , Kassel, Germany 15 Department of Neurology, University Medical Center, Georg August University , Göttingen, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Miriam Neis 7 Department of Neurology, University Hospital Schleswig-Holstein , Campus Lübeck, Lübeck, Germany 16 Department of Midwifery Science, University of Lübeck , Lübeck, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Peter P. Pramstaller 10 Institute for Biomedicine, Eurac Research , Bolzano, Italy Find this author on Google Scholar Find this author on PubMed Search for this author on this site Jannik Prasuhn 4 Institute of Neurogenetics, University of Lübeck , Lübeck, Germany 7 Department of Neurology, University Hospital Schleswig-Holstein , Campus Lübeck, Lübeck, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Eva Schaeffer 8 Department of Neurology, Christian-Albrechts-University of Kiel, University Hospital Schleswig-Holsten , Campus Kiel, Kiel, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Manu Sharma 17 Centre for Genetic Epidemiology, Institute for Clinical Epidemiology and Applied Biometry, University of Tübingen , Tübingen, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Meike Steinbach 8 Department of Neurology, Christian-Albrechts-University of Kiel, University Hospital Schleswig-Holsten , Campus Kiel, Kiel, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Claudia Trenkwalder 14 Paracelsus Elena Clinic , Kassel, Germany 18 Department of Neurosurgery, University Medical Center Göttingen , Göttingen, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Norbert Brüggemann 4 Institute of Neurogenetics, University of Lübeck , Lübeck, Germany 7 Department of Neurology, University Hospital Schleswig-Holstein , Campus Lübeck, Lübeck, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Ana Westenberger 4 Institute of Neurogenetics, University of Lübeck , Lübeck, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Michael Wittig 9 Institute of Clinical Molecular Biology, Christian-Albrechts-University of Kiel , Kiel, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Inke R. König 5 Institute of Medical Biometry and Statistics, University of Lübeck , Lübeck, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Patrick May 2 Luxembourg Centre for Systems Biomedicine, University of Luxembourg , Esch-sur-Alzette, Luxembourg Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Patrick May Christine Klein 4 Institute of Neurogenetics, University of Lübeck , Lübeck, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Amke Caliebe 1 Institute of Medical Informatics and Statistics, Christian-Albrechts-University of Kiel, University Hospital Schleswig-Holstein , Campus Kiel, Kiel, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Abstract Full Text Info/History Metrics Supplementary material Preview PDF Abstract While genetic causes are identified in up to 15% of all Parkinson’s disease (PD) patients, the remaining idiopathic PD (iPD) patients are attributed to polygenic risk, environmental and lifestyle factors, and interactions thereof. We applied five advanced polygenic score (PGS) tools to data from 1,762 iPD patients and 4,227 healthy controls of European ancestry, resulting in the development of a novel iPD-PGS with significantly improved discriminative performance compared to existing models with an AUC of 0.680 (95% confidence interval (-CI): [0.665, 0.695]). Validation in independent cohorts confirmed its robustness. Notably, patients with early-onset iPD exhibited markedly high PGS values when compared to late-onset iPD patients and healthy controls, underlining the high polygenetic burden in these individuals. We subsequently applied our novel iPD-PGS to carriers of heterozygous PRKN variants and GBA1 coding risk variants. In both cases, our findings suggest that part of the penetrance in these genetic forms of PD can be explained by the same polygenic alterations as observed mitigating iPD. The discriminative potential was greater for GBA1 than for PRKN ( GBA1 : AUC=0.639, 95%-CI=[0.590, 0.687], PRKN : AUC=0.594, 95%-CI=[0.501, 0.687]). Our study highlights the potential of advanced PGSs in PD research, particularly for understanding varying penetrance in genetic PD and identifying high-risk individuals. Introduction Parkinson’s disease (PD) is the fastest growing neurological disease with an almost threefold increase in PD prevalence from 1990 to 2021 1 – 5 and presents one of the greatest health-related challenges due to the demographic development towards an aging population. In idiopathic PD (iPD) the exact etiology is unknown and is believed to be influenced by genetic and environmental factors. Also interactions between these components have been identified 6 – 9 . Its heritability is estimated at around 27% 10 and genome-wide association studies (GWAS) underlined a highly polygenic nature of the heritability of iPD. This polygenicity can be captured by polygenic scores (PGSs). PGSs can estimate the genetic disease predisposition of each individual by adding the effects of a large number of genome-wide risk variants into a single value. They enable and improve research about the genetic background and overlap of diseases, disease subtypes and symptoms, and gene-environment interactions, as well as the identification of individuals at high risk 11 , 12 . PGSs are based upon GWAS. The largest European GWAS to date included 56,306 European PD patients and proxy cases and 1,417,791 healthy individuals and identified 90 genome-wide significant independent loci 13 . Recently, an increasing number of GWAS had been performed on non-European or multi-ancestry cohorts 14 – 18 . Along with performing the aforementioned largest European PD-GWAS, an advanced PGS for PD was also developed in this study. This PD-PGS demonstrates a good discriminative performance between PD patients and healthy controls and was validated in an independent population 6 , 7 , 13 , 19 . It consists of 1805 single nucleotide polymorphisms (SNPs) and was generated by a clumping-and-thresholding method 20 , 21 . In the past years, various more refined statistical methods based on larger numbers of SNPs for PGS development have become available, raising the possibility of developing an even better discriminating PGS. To our knowledge, no comparison between different PGS methods has been undertaken for iPD yet. All listed iPD-PGSs in the PGS catalogue 22 – 24 ( www.pgscatalog.org ) use fewer than 2000 SNPs, thus potentially missing relevant parts of the polygenic heritability of iPD. Especially, novel promising Bayesian PGS approaches, such as LDPred2 25 , 26 , PRS-CSx 27 , 28 , and LDAK 29 , have not yet been applied to their full potential. Approximately fifteen percent of PD patients have a genetic form of PD (genetic PD) and carry a genetic causative or risk variant (when including GBA1 coding risk variant carriers) 30 , 31 . For example, the SNCA gene encoding the α-synuclein protein, was the first to be linked to a form of monogenic PD 32 , 33 . The penetrance of the variants differs considerably from nearly full penetrance (around 100% for homozygous or compound-heterozygous mutation carriers in PRKN or PINK1 ) to low penetrance (around 15% for GBA1 coding risk variant carriers). Up to now, the contribution of the polygenic component of iPD to the penetrance of genetic PD remains uncertain. In this study, we aimed to to generate a novel PGS with higher discrimination between iPD patients and healthy controls than previous models by comparing five advanced PGS construction tools utilizing a dataset of 1,762 patients and 4,227 controls of European ancestry. The role of eight genes related to genetic PD in this iPD-PGS was a main focus. The performance of the novel iPD on GBA1 coding risk variant carriers and on PRKN heterozygous carriers was especially relevant to the study’s objective. Overall, our study aimed to investigate the potential of advanced PGSs for iPD and genetic PD research. In the light of the PD pandemic, which calls for new measures for early detection, prevention, and disease-modifying therapy strategies, it is especially relevant to identify high-risk individuals. These can then be targeted for primary prevention in combination with environmental and lifestyle factors, e.g., exposure to neurotoxic agents or smoking. Results Development of an improved iPD-PGS For this study, we used individual genotyping data from the ProtectMove consortium ( www.protect-move.de ). After quality control, 5,623,154 SNPs of 6,826 individuals with genetically determined European ancestry were available. Part of the individuals carried variants of genetic PD on eight PD-related genes ( LRRK2 , SNCA , VPS35 , CHCHD2 , PRKN , PINK1 , PARK7, GBA1 ) and will be termed variant carriers in the following (including variants rated as variants of unknown significance (VUS), risk-variants (on coding regions of GBA1) , likely pathogenic and pathogenic variants; heterozygous carriers included as well). We split our dataset into a late-onset iPD dataset (n=4,227 controls (mean age-at-sampling (AAS)=55 years) ; n=1,762 cases (mean AAS=70 years, mean age-at-onset (AAO)=63 years), both without variant carriers and with an AAO above 40 years for patients), an early-onset iPD dataset (n=66 cases (mean AAS=45 years, mean AAO=35 years), without variant carriers and with an AA0 < 40 years) and a genetic PD dataset (variants carriers: n=436 controls (non-manifesting carriers, mean AAS=53 years), n=335 cases (mean AAS=67 years, mean AAO=60 years) ( Fig. 1 , Table 1 ). Download figure Open in new tab Figure 1: Workflow for PGS model development, validation and application in Parkinson’s disease cohorts Our ProtectMove cohort ( https://protect-move.de/ , n = 6,826), comprised European individuals in three datasets: late-onset idiopathic Parkinson’s disease (iPD), early-onset iPD, and variant carriers (PD-related variants (on the genes LRRK2 , SNCA , VPS35 , CHCHD2 , PRKN , PINK1 , PARK7 , GBA1 ) with heterozygous Parkin carriers and GBA1 carriers as largest subgroups, genetic PD). Five tools for generating polygenic scores (PGSs) were applied to the late-onset iPD dataset to derive five final PGS models, based on the GWAS summary statistics from Nalls et al. 13 and using five-fold cross-validation (CV). The final models were validated in two independent, external cohorts: CouragePD and LuxPark. The best-performing model was identified based on the performance on the late-onset iPD data and the external cohorts (CouragePD, LuxPark). This model was then applied to the genetic PD and early-onset iPD datasets of ProtectMove. View this table: View inline View popup Download powerpoint Table 1: Characteristics of the ProtectMove datasets used in this study (after quality control) For the development of a new iPD-PGS, the late-onset iPD dataset was used. We applied the following five PGS generating tools, utilizing different statistical approaches, to develop new iPD-PGSs : PRSice-2 20 , 21 , LDpred2 25 , 26 (inf, auto, grid), lassosum2 34 , 35 , LDAK 29 (BayesR-SS, Bolt-SS, Ridge-SS), PRS-CSx 27 , 28 (Supplementary Methods, Supplementary Table 1). For this, our late-onset iPD dataset was divided into five datasets with equal distribution in case-control status, cohort affiliation, sex and age (AAO, AAS) for five-fold cross validation (CV) (Supplementary Table 2). For each tool, several PGSs were developed and we selected the best-performing one for comparison. Based on all genome-wide variants, the LDpred2 tool demonstrated the best discrimination between PD-patients and healthy controls as measured by the area under the receiver operating characteristic curve (AUC) ( Table 2 , Fig. 2 ; for details on the individual methods, see Supplementary Tables 3-7). A mean AUC of 0.679 was achieved on the validation datasets ([minimum, maximum] ([min, max]) across the five datasets: [0.662, 0.691]) and of 0.680 (95% confidence interval (CI): [0.665, 0.695]) on our whole late-onset iPD dataset (n=1,386,717 SNPs in PGS). As comparison, we calculated the AUC for the currently established PGS 13 (n=1,805 SNPs originally in PGS, 1,097 SNPs available in late-onset iPD dataset) generated with the tool PRSice-2. Here, the resulting AUC was considerably lower, with 0.623 in both the validation and the whole late-onset iPD dataset ([min, max]: validation [0.606, 0.646], whole late-onset iPD 95%-CI: [0.607, 0.638]). The tools LDAK, lassosum2, and PRS-CSx showed a slightly worse performance than LDpred2 with AUCs around 0.66-0.67, whereas PRSice-2 dropped considerably in performance (mean AUC validation: 0.637, [min, max]: [0.623, 0.651]; AUC whole iPD: 0.641, 95%-CI: [0.626, 0.656]). View this table: View inline View popup Download powerpoint Table 2: Performance of constructed iPD-PGSs on the ProtectMove late-onset iPD dataset Download figure Open in new tab Figure 2: Comparison of performance of constructed iPD-PGSs on the late-onset iPD dataset Summary of model performance across the five cross-validation (CV) folds and the whole idiopathic late-onset Parkinson’s disease (iPD) dataset, using the area under the receiver operating characteristic curve (AUC) as measure. For the CV approach, the mean AUCs across all five combinations of folds for training and validation were calculated. The error bars indicate the minimum and maximum across the five training or validation datasets. Different colors correspond to the selected tools for polygenic score (PGS) generation with the established PGS by Nalls et al. 13 as comparison. The overall best-performing and selected final tool (LDpred2) is highlighted in bold. The best-performing PGS for each tool was applied to two external validation datasets: LuxPark (822 patients and 814 healthy individuals) and CouragePD 36 (10,854 patients, 8,626 healthy individuals) ( Table 3 , Fig. 3 ). Applied to the large CouragePD dataset, the performance of all PGSs was even improved compared to the ProtectMove late-onset iPD dataset. LDAK achieved the highest performance (AUC: 0.717, 95%-CI: [0.709, 0.724]), slightly better than LDpred2 (AUC: 0.715, 95%-CI: [0.708, 0,722]). On the smaller LuxPark dataset, the PGS of lassosum2, LDpred2 and PRS-CSx performed best (AUC: 0.672-0.673), while LDAK (AUC: 0.627) and PRSice-2 (AUC: 0.636) resulted in lower performance. The currently established PGS 13 showed a solid performance (CouragePD: AUC: 0.663, SNPs available: 1,730; LuxPark: AUC: 0.64, SNPs available: 1,801), even slightly better than PRSice-2 on both datasets. View this table: View inline View popup Download powerpoint Table 3: Performance of constructed iPD-PGSs on external validation datasets Download figure Open in new tab Figure 3: Comparison of performance of constructed iPD-PGSs on external validation datasets Summary of model performance across the two external validation datasets CouragePD and LuxPark for idiopathic Parkinson’s disease (iPD), using the area under the receiver operating characteristic curve (AUC) as measure. The error bars indicate the 95% confidence interval. Different colors correspond to the selected tools for polygenic score (PGS) generation with the established PGS by Nalls et al. 13 as comparison. The overall best-performing and selected final tool (LDpred2) is highlighted in bold. The correlation between PGS values generated by different tools varied considerably ( Fig. 4 ). For example, while a high correlation between the PGSs of LDpred2 with those of lassosum2 and PRS-CSx was observed (Spearman correlation coefficient r of 0.91 and 0.94, respectively, p < 0.001), the PGS values of PRSice-2 and LDAK showed only moderate correlations with the other PGS values (r between 0.51 and 0.80, p < 0.001). Download figure Open in new tab Figure 4: Correlation between PGS values based on different tools Given are Spearman correlation coefficients calculated for our late-onset idiopathic Parkinson’s disease (iPD) dataset. The established polygenic score (PGS) by Nalls et al. 13 is shown for comparison. Performance of best iPD-PGS The iPD-PGS generated by LDpred2 demonstrated exceptionally high predictive performance on the ProtectMove late onset iPD and the two external datasets, showing robust results across all three datasets. Therefore, this iPD-PGS served as the basis for our subsequent evaluations in this study ( Table 2 ). Fig. 5A shows the density curves of this PGS for patients and healthy individuals in our whole ProtectMove late-onset iPD dataset. The capability of the PGS to distinguish between patients and healthy individuals was further investigated by dividing the late-onset iPD dataset into deciles of PGS values. We then calculated odds ratios (ORs) by comparing individuals with the lowest PGS values (i.e., in the first decile) with those with higher PGS values (i.e., in the second to tenth deciles) ( Fig. 5B ). ORs varied between 1.463 (second decile, 95%-CI: [1.051, 2.046]) and 9.342 (tenth decile, 95%-CI: [6.973, 12.664]). The PGS also showed an association with AAO ( Fig. 5C ). An AUC of 0.698 (95%-CI: [0.637, 0.756]) was achieved when differentiating patients of the first and tenth decile of AAO (AAO75) based on the PGS. The correlation between PGS and AAO was -0.178 (95%-CI: [-0.225, -0.129], p < 0.001) and -0.167 (95%-CI: [-0.215, -0.118], p < 0.001) for the Pearson and Spearman correlation, respectively. Download figure Open in new tab Figure 5: Evaluation of the best-performing iPD-PGS The best-performing iPD-PGS (LDpred2) was applied to the whole late-onset idiopathic Parkinson’s disease (iPD) dataset. A: Density curves of the standardized polygenic score (PGS) for patients and healthy individuals. The discriminative ability of the PGS for these groups is measured by the area under the receiver operating characteristic curve (AUC). B: Odds ratio (OR) of case-control status for each PGS decile, compared to the first PGS decile. The error bars indicate the 95% confidence interval. C: Density curves of the standardized PGS for cases within the highest and lowest age-at-onset (AAO) decile. D: Density curves of the full logistic regression model (influence variables PGS, AAS, sex and the first 20 principal components) for cases and controls. The discriminative performance of our derived PGS and combined with other covariables is displayed in Table 4 . We investigated age (AAS), sex, and the first 20 principal components (PCs). As expected, age alone showed a high discriminative potential (R 2 =0.302, AUC=0.787), whereas the association of either sex or the PCs with PD was lower (sex: R 2 =0.019, PCs: R 2 =0.047, AUC=0.614). Including all three influencing factors (age, sex, PCs) improved model performance (R 2 =0.340, AUC=0.808). Notably, adding our PGS on top of these three factors further enhanced this performance, and a high AUC of 0.840 was achieved (95%-CI: [0.830, 0.851], R 2 =0.407, Fig. 5D ), underscoring the strong predictive potential of the combined model (Supplementary Figure 6). View this table: View inline View popup Download powerpoint Table 4: Performance of the best-performing PGS (LDpred2) combined with covariables on the whole ProtectMove late-onset iPD dataset Influence of PD-related genes We next investigated to what extent the predictive potential of our developed PGSs was due to genetic variability in genes related to genetic PD ( LRRK2 , SNCA , VPS35 , CHCHD2 , PRKN , PINK1 , PARK7 , GBA1 ). We therefore excluded all variants in these eight genes (33,285 out of 5,623,154 variants) from the set of genome-wide variants and recalculated PGSs with all five PGS tools based on this reduced set of variants. The performance of the developed PGSs was slightly reduced, when the PD-related genes were excluded from the construction process ( Table 2 , column Whole late-onset iPD dataset, AUC no PD-genes). This trend was consistent across all five tools and was also observed in the two external validation datasets ( Table 3 ). For instance, the AUC of our best-performing PGS decreased marginally from 0.680 to 0.672 in the ProtectMove late-onset iPD dataset. To further elucidate the specific role of SNCA , we constructed PGSs using a set of variants where those in the SNCA gene were retained (4,926), while variants from the other seven PD genes were excluded (28,359). Interestingly, the resulting PGSs demonstrated an equivalent performance to the PGSs including all variants, again demonstrating robustness over all five tools and in the external datasets LuxPark and CouragePD ( Table 2 , column Whole late-onset iPD dataset, AUC no PD-genes, with SNCA , Table 3 ). The slight reduction in the predictive performance when variants in the eight PD-related genes were excluded is, thus, due to variants in SNCA . Application of the newly developed iPD-PGS on genetic PD and early-onset iPD Next, we assessed the role of polygenic background in the variability of penetrance of PD among variant carriers. Our genetic PD dataset consisted of 436 healthy individuals (non-manifesting carriers) and 335 PD patients, both with one (heterozygous) or two (biallelic) genetic causative or risk variants in one of the eight PD-related genes ( Table 1 , Supplementary Tables 8 and 9.) Most genetic variants were detected in GBA1 (n=495) and in PRKN (n=168 heterozygous and n=14 biallelic). Based on power calculations, we restricted our analysis to these two groups (see Statistical analysis). We considered only carriers of heterozygous genetic variants for PRKN because of the difference in disease susceptibility, clinical presentation, and AAO in the biallelic PRKN carrier group 37 (causality can only be assumed in the latter group). Our best-performing PGS based on all variants demonstrated increased values for GBA1 variant carriers (both PD patients and healthy individuals; data not shown). This was due to the more common GBA1 coding risk variants included in the PGS. Therefore, we applied our PGS without variants in PD-related genes to GBA1 and PRKN variant carriers to determine whether the case/control status (and thus the penetrance) could be predicted. The AUC for manifestation of both PRKN and GBA1 variant carriers based on the PGS was significantly larger than 0.5, indicating that the PGS contributes to the varying penetrance in these two groups ( Table 5 ). For GBA1, the discriminative potential was larger than for PRKN ( GBA1 : AUC=0.639, 95%-CI=[0.590, 0.687], p < 0.001, PRKN : AUC=0.594, 95%-CI=[0.501, 0.687], p < 0.0265) (Supplementary Figure 7). Additionally, we examined the group of early-onset iPD. Interestingly, the AUC for iPD patients with an early AAO was even larger than for the late-onset iPD dataset (AUC=0.725, 95%-CI=[0.665, 0.784]). View this table: View inline View popup Download powerpoint Table 5: Application of our best-performing PGS without variants on PD-related genes on variant carriers and early-onset iPD patients Discussion In this study, we developed a new PGS for iPD that substantially outperformed previously established PGSs 13 , 38 , 39 . We achieved this by applying five tools and using three different statistical approaches: Bayesian inference, penalized regression, and clumping-and-thresholding. The performance of PGS models varied substantially depending on the tool used. Among the tools evaluated, LDpred2 consistently demonstrated the best overall performance. It excelled across all five CV sets, the entire training dataset, and both external validation datasets, underlining its robustness across cohorts of the same ancestry. The PRS-CSx’s strong performance is comparable to LDpred2 but is slightly less robust on the CV datasets. The advanced Bayesian framework shared by both tools likely explains their superior performance. It accounts for the LD structure, and thus adjusts the GWAS summary statistic weights and incorporates more markers and information into the PGS. The penalized regression tool lassosum2 delivered robust and reliable results across all datasets but slightly underperformed compared to LDpred2 and PRS-CSx. LDAK (Ridge-SS), another tool based on penalized regression, performed similarly to lassosum2 in most cases but exhibited unique characteristics: it achieved the highest AUC on the CouragePD dataset—outperforming even LDpred2—while delivering the poorest results on the LuxPark dataset. These simpler penalized regression frameworks may offer advantages when computational efficiency is required 34 . In contrast, PRSice-2 consistently performed the worst among the evaluated methods. This highlights the limitations of more straightforward clumping-and-thresholding methods, which may struggle to capture the entire polygenic architecture of iPD by discarding informative markers 25 . The results of this study exhibit the potential to improve the predictive capability of PGS by advanced statistical methodology. The number of SNPs included in a PGS model heavily relies on the method used for its development and significantly impacts its performance. Our study demonstrated that models with fewer SNPs, such as those based on the clumping-and-thresholding tool PRSice-2, exhibit the poorest discriminative ability. This finding aligns with previous literature 26 suggesting that limited SNP inclusion may be insufficient for assessing the complete polygenic architecture of complex diseases. On the other hand, PGS models with fewer SNPs have the advantage of being easier to calculate and potentially simpler to transfer to external datasets that use different SNP arrays. Moreover, a performance drop could occur when applying models with high SNP inclusion on external datasets due to overfitting and differing imputation quality. In our study, the PGS models comprising a high number of variables were robustly validated across two independent datasets. Importantly, this robustness was maintained even without using proxy SNPs reinforcing the integrity of the models. Furthermore, we tested a sparse version of LDpred2, using only 800,971 SNPs (instead of 1,386,717), which did not yield improved performance on the external validation datasets, suggesting that the inclusion of a broader SNP set is advantageous for predictive accuracy on external data for PD. Genetic variability in eight known PD-related genes does not drive the high predictive potential of our newly derived PGS. Indeed, there was only a slight decrease in the prediction performance when variants on PD-related genes were excluded. Interestingly, there was no decrease at all, when variants on SNCA were included while variants on the other seven PD-related genes were excluded. Variants on SNCA were thus causal for this small performance drop. SNCA is long known to play a major role in monogenic PD, but for our study it seems to have a significant influence in the polygenic inheritance of iPD as well probably due to the risk variants on this gene 40 . Risk variants on GBA1 , on the other hand, did not contribute to the PGS accuracy. The performance of our best iPD-PGS varied among different groups. The discriminative ability between PD patients and healthy individuals, measured by AUC, was highest for early AAO patients, followed by non-carriers (late-onset iPD cohort), GBA1 carriers, and PRKN carriers. The iPD-PGS’s ability to distinguish between PD-patients and healthy individuals, even in carriers of GBA1 and PRKN variants, reveals its role in modulating the penetrance of rare causal genetic variants or coding risk variants on GBA1 by capturing the polygenic background. The lower AUC values observed for GBA1 and PRKN carriers compared to non-carriers highlight the predominant role of rare mutations or specific risk variants in driving disease in these individuals. This study therefore could identify the polygenic background as one factor that explains the varying penetrance in variant carriers of genetic PD, at least for the genes GBA1 and PRKN . This also highlights that genetic and idiopathic PD are not completely separated entities but share the polygenic background as a common risk factor. A previous study found that a PGS for iPD modulated PD disease risk for GBA1 carriers 41 . However, the applied PGS consisted only of 90 independent genome-wide significant variants previously identified 13 . In contrast, our PGS was especially derived for iPD with advanced statistical methodology and includes more than one million variants. Another study investigated the interplay of monogenic causal variants and the polygenic etiological component for familial hypercholesterolemia, hereditary breast and ovarian cancer syndrome, and Lynch syndrome. They also found that the polygenic background modified the penetrance of monogenic variants and suggested more research in this important area to bridge the translation to patients 42 . The superior performance of the iPD-PGS in early-onset individuals suggests that these patients have, on average, an elevated individual score. It is therefore unlikely that the early onset of this group is mainly due to undetected variants in single genes. Instead, they carry a high burden of common risk variants that accelerate disease onset. The polygenic background is therefore particularly influential for these patients, underscoring the predictive potential of iPD-PGSs for individuals with exceptionally high PGS values. This was also targeted in an earlier study by calculating a PGS for early-onset (≤ 50 years) PD patients with and without a monogenic cause. However, the small number of patients (n=90) and the lack of healthy controls render their findings primarily exploratory 43 . The relevance of PGSs in addition to rare variants in understanding the genetic contribution to PD has high imperative for clinical practice. For non-variant carriers, the iPD-PGS serves as a valuable risk stratification tool. Those individuals with very high PGS values (in the tenth decile) have ninefold increased odds of being a PD patient than persons with very low PGS values (in the first decile). Although transfer from diagnostic to prognostic ability is limited without longitudinal data, this finding shows the potential for the PGS to serve as a tool in personalized medicine. In addition to differentiating between patients and healthy individuals, the iPD-PGS performs well in discriminating early from late AAO patients. The increased AUC in early-onset individuals underscores the need for targeted research into genetic and environmental factors contributing to early-onset disease. For variant carriers of genetic PD, the PGS provides insights into modifiers within the polygenic background. It is important to interpret our findings in light of potential limitations. First, we only investigated variant carriers in GBA1 and heterozygous carriers in PRKN . The sample size of variant carriers of other genes was not sufficient for meaningful statistical analyses. Initiatives to increase the number of mutation carriers, such as the Global Parkinson’s Genetics Program (GP2), are ongoing and should result in promising follow-up studies in the future 44 . Second, we defined mutation carriers by single nucleotide variants (SNVs) or copy number variants (CNVs) rated as pathogenic, likely pathogenic, or VUS without differentiation between pathogenicity scoring or variant type (SNV or CNV). This suggests another avenue for future valuable research. A further fundamental limitation is our restriction to European individuals. PGSs are known to be highly ancestry-sensitive 45 , 46 . Because most GWAS data are available for individuals of European ancestry, this might exacerbate healthcare differences globally 47 . The awareness of the requirement of sufficient genetic data for minorities, diverse and admixed populations, and adequate statistical methodology has been increased in recent years 48 . Several initiatives have also started to address this challenge 49 for PD, such as GP2 44 . Developing advanced PGSs for other or multi ancestries or methods for transferring PGSs across ancestries could provide equitable PGS-related medical care in the future. Methods Study cohorts and variant carrier identification For this study, we utilized data from nine cohorts (Supplementary Table 10) which were collected in the ProtectMove consortium ( https://protect-move.de/ ). Further information on the cohorts included can be found in the Supplement. For validation, we had access to two independent datasets. Both datasets were quality controlled to include only individuals of European ancestry. The CouragePD 36 dataset comprised 10,854 patients and 8,626 healthy individuals. This dataset did not include information on the carrier status of variants on PD-related genes. The LuxPark dataset included 822 patients and 814 healthy individuals, with LRRK2 , SNCA , VPS35 , CHCHD2 , PRKN , PINK1 , PARK7 and GBA1 carriers removed 50 , 51 . For the definition of PD-related genes, we restricted our analysis to the confirmed PD-linked genes LRRK2 , SNCA , VPS35 , CHCHD2 , PRKN , PINK1 , PARK7 , and GBA1 . This selection is based on the published “classical parkinsonism” and “early onset parkinsonism” lists 52 . Additionally, we included carriers of coding risk variants in GBA1 . Genetic variants (SNVs and CNVs) within these genes were assessed for pathogenicity using the Franklin ( https://franklin.genoox.com - Franklin by Genoox), Varsome 53 and MDSGene 54 , 55 assessments. The possible values for pathogenicity are “benign”, “likely benign”, “VUS “, “risk variant” (for GBA1 only), “likely pathogenic” and “pathogenic”, in increasing order of severity. Individuals were defined as variant carriers of genetic PD if they had at least one allele or one CNV in one of these genes with a pathogenicity rating of at least “VUS”. The final pathogenicity rating for an individual was determined by the highest-ranked mutation present in that individual 50 , 52 , 56 . An overview of the identified variant carriers and the pathogenicity of detected variants can be found in Supplementary Tables 8 and 9. Ethics statements The study was conducted according to the guidelines of the Declaration of Helsinki. Informed consent was obtained from all subjects involved in the study. It was approved by the Ethics Committees of the University of Lübeck, Germany (protocol code 16-039, date of approval 27 September 2019) and the P2N supervisory board, Kiel University, Germany (protocol code 2021-037, date of approval 16 September 2021). The DeNoPa study was approved by the ethics committee of the Physician’s Board Hesse, Germany (approval no. FF89/2008). The Ethics Committee of the Bolzano Health District approved the GESSPARK protocol on 05 June 2008 (30/2008), with an update approved by the Ethics Committee of the Healthcare System of the Autonomous Province of Bolzano-South Tyrol on 11 October 2017. The Ethics Committee of the Bolzano Health District approved the DISP protocol on 25 July 2012 (62/2012), with an update approved by the Ethics Committee of the Healthcare System of the Autonomous Province of Bolzano-South Tyrol on 11 October 2017. Genotyping and quality control For the ProtectMove data, genomic DNA was extracted from peripheral blood leukocytes and genotyped using the Infinium Global Screening Array with Custom Content (GSA; Illumina Inc., San Diego, CA, USA), which targets 645,896 variants. Quality control was performed using PLINK 1.9 57 , 58 , PLINK 2.0 58 , 59 , and the R package plinkQC 60 . The following thresholds were applied at the SNP level: minor allele frequency (MAF) ≥ 0.01, SNP call rate ≥ 0.98, and Hardy-Weinberg equilibrium test ≥ 10 -50 for the software-issued p-value. After this process, a total of 431,738 variants passed quality control and were used for imputation with SHAPEIT2 61 and IMPUTE2 62 , leveraging the publicly available portion of the HRC reference panel (release 1.1, The European Genome-Phenome Archive, EGAS00001001710). Imputation generated genotyping data for 39,106,911 variants. Variants with MAF < 0.01 or an info score < 0.7 were excluded, leaving a final dataset of 7,804,284 variants for downstream analysis. We performed quality control for genotyped individuals by removing individuals with increased heterozygosity rate, related individuals and individuals with non-European ancestry (Supplementary Methods). The dataset was then divided into three groups: late-onset iPD (patients and unaffected individuals without variant carriers and with an AAO above 40 years for patients), early-onset iPD (patients without variant carriers and with an AA0 < 40 years) and a genetic PD (affected and not affected variants carriers). See Table 1 for further information on the three groups. The final ProtectMove study population consisted of 6,826 individuals. Subsequent SNP selection resulted in 5,623,154 SNPs for PGS construction. PGS construction a) PGS algorithms and GWAS We constructed PGS models for iPD according to five different tools (PRSice-2 20 , 21 , LDpred2 25 , 26 (inf, auto, grid), lassosum2 34 , 35 , LDAK 29 (BayesR-SS, Bolt-SS, Ridge-SS), PRS-CSx 27 , 28 ). While PRSice-2 is based on clumping and thresholding and both lassosum2 and LDAK’s Ridge-SS utilize penalized regression, LDpred2, LDAK’s BayesR-SS, LDAK’s Bolt-SS and PRS-CSx rely mainly on Bayesian approaches. All of the methods use a multitude of parameters in their models, which we varied to obtain the best-performing tool-specific PGS (for more details see Supplementary Methods and Supplementary Table 1). As GWAS summary statistics, we applied the largest GWAS for PD in individuals of European ancestry, including 37,688 cases, 18,618 UK Biobank 63 proxy-cases, and 1,417,791 controls 13 . The GWAS encompassed a total of 7,784,415 SNPs (6,611,727 without ambiguous SNPs) and were kindly provided by Mike Nalls (with approval by the 23andMe project, agreement University of Kiel SOW #1). The GWAS summary statistics were harmonized to fit the format and effect direction for the ProtectMove late-onset iPD dataset. Ambiguous SNPs, SNPs with a MAF below 0.01 or with a different alternate allele than in the late-onset iPD set were removed. For SNPs with a swapped major and minor allele compared to the late-onset iPD data, the effect size was multiplicated with -1. b) Construction of cross-validation datasets To identify the best and most robust PGSs for iPD, we performed a five-fold CV. For this, we decomposed the late-onset iPD dataset into the groups “cases with known sex and known AAO”, “controls with known sex and known AAS” and the remainder. We then used the fold() function from the R-package “groupdata2” 64 on each group to create five equal-sized CV sets. The three groups (cases, controls and remainder) were each divided into five groups of equal size, with each of these five groups being equally distributed according to cohort affiliation, sex and age (mean AAO for cases, mean AAS for controls) and with equally distributed case-control status for the remainder. The resulting CV sets of these three groups were then combined. See Supplementary Table 2 for a more detailed description of the CV sets. c) Construction of PGS models without PD-related genes To develop a PGS without SNPs from the PD-related genes LRRK2 , SNCA , VPS35 , CHCHD2 , PRKN , PINK1 , PARK7 and GBA1 , we removed all SNPs within the gene regions with an additional distance of 1,000,000 base-pairs to either side in our genetic data of the late-onset iPD cohort. We also created PGS models in which the SNCA gene was retained while the other seven PD-related genes were removed. Statistical analysis a) Identifying the best-performing PGS model We performed five-fold CV by dividing the dataset into five groups of the same size and the same structure (see above). On each CV step, one set served as the validation set for the PGS models trained on the combined remaining four sets. We carried out the following steps to determine the highest-performing PGS model: For each CV step, various models for each tool (PRSice2, LDpred2, lassosum2, PRS-CSx, LDAK) were calculated depending on the variation of parameters per tool (Supplementary Methods). Based on the AUC performance, one resulting best model per tool and per training dataset was selected. The selected models were then applied to the respective validation datasets and mean AUC over the five validation sets calculated to estimate the prediction performance on an independent internal dataset. We then trained the prediction tool on the whole iPD cohort. The best models of each tool trained on the whole iPD cohort were applied to our two external validation datasets LuxPark and CouragePD. The best-performing PGS was then identified by the performance on the internal validation datasets of the CV and on the two external validation datasets. b) Statistical measures, models, tests and software Descriptive measures used in our study were frequencies for categorical variables and median and interquartile range for continuous variables. All statical calculations were performed using the statistics software R 65 , version 4.4.1. All applied statistical tests were two-sided and a significance level of 0.05 was used. All statistical figures have been created using the R package “ggplot2” 66 , unless stated otherwise. Individual PGS scores were calculated for the training and validation sets using Plink2, based on the selected SNPs and their corresponding weights. The scores were standardized to have a mean of 0 and a standard deviation of 1 among healthy individuals. The receiver operating characteristics (ROC) curve and the AUC, along with 95%-CIs, were computed using the R package “pROC” 67 , the respective p-values were calculated using the R package “verification” 68 . Additionally, we evaluated PGS performance by calculating ORs for individuals within the PGS deciles compared to the first PGS decile for the best-performing model. For the correlation between PGS and AAO, the correlation coefficients and 95%-CIs were calculated using base R (for Pearson) and the SpearmanRho() function from the R package “DescTools” 69 (for Spearman). Furthermore, the densities of individual PGS values for the first versus the tenth AAO decile of the PD-patients were plotted and the corresponding AUC between these AAO groups calculated. To evaluate the role of covariables, we applied six logistic regression models for PD status (case/control) as outcome using R’s glm() function, with the following influence variables: (1)-(4) Each of the following variables individually: PGS, age (AAS), sex and the first twenty PCs (PCs included together), (5) all variables except PGS and (6) all aforementioned variables together. Nagelkerke’s R² for each logistic regression model was calculated using the r2_nagelkerke() function from the R package “performance” 70 . To estimate the concordance of the PGS values generated by the five different PGS tools, we calculated Spearman correlation coefficients and used the R package “corrplot” 71 to visualized the results. For the analysis of variant carriers, we restricted our statistical analysis to the two largest groups, i.e. variants on the genes GBA1 and PRKN . SNVs, as well as CNVs, were considered (see the respective subsection above for the definition of variant carriers). For PRKN , only heterozygous carriers were analyzed due to the small number of homozygous carriers and the distinct differences in disease susceptibility, clinical presentation, and AAO between the two groups 37 . Variant carriers with variants on PD-related genes in more than one gene were excluded from this analysis. We applied our best-performing PGS, constructed without variants from PD-related genes, to ensure that the observed performance was attributable to the polygenic component rather than the influence of variants in the gene under investigation, as intended by this analysis. This PGS was also applied to our dataset of early-onset iPD patients (AA0 < 40) and compared against healthy individuals of the late-onset iPD dataset. We calculated the median and interquartile range (IQR) of the PGS for each group, separately for patients and healthy individuals, and the AUC with corresponding 95%-CI. c) Power calculation for analysis of variant carriers We performed power calculations for variant carriers of different PD-related genes to evaluate which genes were meaningful to analyze. As basis we utilized the Wilcoxon rank-sum test for comparison of PGS values between patients and healthy individuals carrying variants on a specific PD-related gene and assumed a significance level of 0.05. For GBA1, 222 patients and 273 healthy individuals were available which resulted in a power of >99% for a medium effect (Cohen’s d=0.5) and of 80% for a small effect (Cohen’s d=0.3). For PRKN we had 51 patients and 117 healthy individuals yielding a power of 82% for a medium effect and of 41% for a small effect. For all other genes, the power was <20% for a medium effect and <10% for a small effect. Power calculations were performed with the statistics software G*Power 3.1.9.4 72 . Data availability The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions. Funding information This research was funded by the German Research Foundation (FOR2488 to D.B., N.B., A.C., A.H., M.K., C.K., I.R.K., K.L., P.M., P.P.P., A.W. and TR-CRC134 to U.M.K., M.K., C.K.). The DISP and GESSPARK studies were funded by the Autonomous Province of Bolzano-South Tyrol - Department of Innovation, Research, University and Museums. Author information Sebastian Sendel and Zied Landoulsi are shared first authors, Amke Caliebe and Christine Klein are shared last authors. Contributions Conceptualization, A.C., C.K., I.R.K., K.L., P.M., S.S.; methodology, A.C., I.R.K., B.-H.L., P.M., S.S.; formal analysis, D.R.B., Z.L., S.S.; interpretation of data, A.C., C.K., I.R.K., B.-H.L., K.L., P.M., S.S.; resources, A.B., D.B., M.B., N.B., A.F., H.H., E.H., A.A.H., U.M.K., G.K., R.K., L.M.L., W.L., K.L., B.M., M.N., P.P.P., J.P., E.S., M.S. (Manu Sharma), M.S. (Meike Steinbach), C.T., T.U., A.W., M.W.; data curation, M.K., B.-H. L., Z.L., K.L., E.-J.V.; writing—original draft preparation, A.C., S.S.; writing—review and editing, A.B., M. B., N.B., A.C., C.K., I.R.K., L.M.L., K.L., S.S., C.T.; visualization, S.S.; supervision, A.C., P.M.; project administration, C.K.; funding acquisition, A.C., C.K. All authors reviewed and approved the final manuscript. Ethics declaration Competing interests Z.L. reports grants from the Luxembourg National Research Fund (FNR) (FNR11264123) and INTER/DFG/19/14429377. K.L. reports grants from the the German Research Foundation (DFG, LO 1555/11-2) and DMRF. M.K. reports grants from the German Research Foundation (DFG) (FOR2488). N.B. received honaria from Abbott, Abbvie, Biogen, Esteve, Ipsen, Merz, Teva, Zambon. N.B. was supported by the DFG, Michael J Fox Foundation and the EU Joint Programme - Neurodegenerative Disease Research (JPND). A.B. reports grants from the Else Kröner-Fresenius-Stiftung and the German Research Foundation (DFG) as well as stock ownership in medically-related fields (Novo Nordisk, Stryker, and Sartorius). D.B. received grants from Biohaven, BMBF, Deutsche Forschungsgemeinschaft (DFG), Else-Kröner-Forschungskolleg (EKFK), Hoffmann La Roche AG, Jan von Appen Stiftung, Lundbeck, Michael J. Fox Foundation (MJFF), UCB Pharma GmbH, EU, Novartis Pharma GmbH, honoraria from UCB Pharma GmbH; Lilly Germany GmbH; Ottawa Hospital Research Institute, and consultancies from UCB Pharma GmbH; Lilly Germany GmbH; Ottawa Hospital Research Institute. M.B. reports grants from the Else Kröner-Fresenius-Stiftung and from the University of Lübeck and received honoraria from Bial and Ipsen Pharma. L.M.L. received honoraria from the Movement Disorder Society (Faculty at MDS Teaching School) and received support from the Bachmann-Strauss Foundation for Dystonia Research. E.S. received grants from the University of Kiel (intramural research funding) and Germany society for Parkinson’s Disease (DPG e.V.), and speaker honoraria from Bayer Vital GmbH, Novartis Pharma GmbH, BIAL GmbH, Zambon and the Movement Disorder Society outside the submitted work. I.R.K. reports grants from the German Research Foundation (DFG), BMBF and German Cancer Aid. P.M. reports grants from Luxembourg National Research Fund (FNR) and BMBF. C.K. reports grants from the German Research Foundation (DFG), Michael J Fox Foundation and ASAP and received speakers’ honoraria from Bial and royalities from Oxford University Press. C.K. serves as a medical advisor to Centogene, Biogen, Takeda, and the Lundbeck Foundation. A.C. reports grants from the German Research Foundation (DFG) (FOR2488, CA 1620/5-1, CO 992/10-2) and of Kiel University (SEA-EU Research Seed Fund 2024). The other authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results. Acknowledgments We sincerely thank all study participants for their invaluable contributions to this research. We also acknowledge the cohorts that provided data and samples for this study and the investigators who facilitated access to these resources. The funding bodies had no role in the study design, data collection, analysis, interpretation, or manuscript preparation. We thank Mike A. Nalls for providing us with the list of the 1805 SNPs included in their published PRS (together with reference alleles and effect sizes β). We thank Manon Sendel for consultation on the discussion and review of language and content of the manuscript. Footnotes ↵ § joint first authors ↵ † joint last authors References 1. ↵ Steinmetz , J. D. et al. Global, regional, and national burden of disorders affecting the nervous system, 1990–2021: a systematic analysis for the Global Burden of Disease Study 2021 . Lancet Neurol . 23 , 344 – 381 ( 2024 ). OpenUrl CrossRef PubMed 2. ↵ Ben-Shlomo , Y. et al. The epidemiology of Parkinson’s disease . Lancet Lond. Engl . 403 , 283 – 292 ( 2024 ). OpenUrl 3. Willis , A. W. et al. Incidence of Parkinson disease in North America . NPJ Park. Dis . 8 , 170 ( 2022 ). 4. Dorsey , E. R. , Sherer , T. , Okun , M. S. & Bloem , B. R. The Emerging Evidence of the Parkinson Pandemic . J. Park. Dis. 8 , S3 – S8 ( 2018 ). OpenUrl 5. ↵ Bloem , B. R. , Okun , M. S. & Klein , C . Parkinson’s disease . The Lancet 397 , 2284 – 2303 ( 2021 ). OpenUrl 6. ↵ Lüth , T. et al. Interaction of Mitochondrial Polygenic Score and Lifestyle Factors in LRRK2 p.Gly2019Ser Parkinsonism . Mov. Disord. Off. J. Mov. Disord. Soc. 38 , 1837 – 1849 ( 2023 ). OpenUrl 7. ↵ Gabbert , C. et al. The combined effect of lifestyle factors and polygenic scores on age at onset in Parkinson’s disease . Sci. Rep . 14 , 14670 ( 2024 ). OpenUrl PubMed 8. Huang , Y. et al. Risk factors associated with age at onset of Parkinson’s disease in the UK Biobank . Npj Park. Dis . 10 , 1 – 8 ( 2024 ). OpenUrl 9. ↵ Reynoso , A. et al. Gene–Environment Interactions for Parkinson’s Disease . Ann. Neurol . 95 , 677 – 687 ( 2024 ). OpenUrl 10. ↵ Goldman , S. M. et al. Concordance for Parkinson’s disease in twins: A 20-year update . Ann. Neurol . 85 , 600 – 605 ( 2019 ). OpenUrl CrossRef 11. ↵ Wray , N. R. et al. Research Review: Polygenic methods and their application to psychiatric traits . J. Child Psychol. Psychiatry 55 , 1068 – 1087 ( 2014 ). OpenUrl CrossRef PubMed 12. ↵ Lewis , C. M. & Vassos , E . Polygenic risk scores: from research tools to clinical instruments . Genome Med . 12 , 44 ( 2020 ). 13. ↵ Nalls , M. A. et al. Identification of novel risk loci, causal insights, and heritable risk for Parkinson’s disease: a meta-analysis of genome-wide association studies . Lancet Neurol . 18 , 1091 – 1102 ( 2019 ). OpenUrl CrossRef PubMed 14. ↵ Verma , A. et al. Diversity and scale: Genetic architecture of 2068 traits in the VA Million Veteran Program . Science 385 , eadj1182 ( 2024 ). OpenUrl CrossRef PubMed 15. Park , K. W. et al. Ethnicity- and sex-specific genome wide association study on Parkinson’s disease . NPJ Park. Dis . 9 , 141 ( 2023 ). 16. Pan , H. et al. Genome-wide association study using whole-genome sequencing identifies risk loci for Parkinson’s disease in Chinese population . Npj Park. Dis . 9 , 1 – 11 ( 2023 ). OpenUrl 17. Foo , J. N. et al. Identification of Risk Loci for Parkinson Disease in Asians and Comparison of Risk Between Asians and Europeans: A Genome-Wide Association Study . JAMA Neurol . 77 , 746 – 754 ( 2020 ). OpenUrl 18. ↵ Kim , J. J. et al. Multi-ancestry genome-wide association meta-analysis of Parkinson’s disease . Nat. Genet . 56 , 27 – 36 ( 2024 ). OpenUrl CrossRef PubMed 19. ↵ Koch , S. et al. Validity and Prognostic Value of a Polygenic Risk Score for Parkinson’s Disease . Genes 12 , 1859 ( 2021 ). OpenUrl 20. ↵ Euesden , J. , Lewis , C. M. & O’Reilly , P. F . PRSice: Polygenic Risk Score software . Bioinformatics 31 , 1466 – 1468 ( 2015 ). OpenUrl CrossRef PubMed 21. ↵ Choi , S. W. & O’Reilly , P. F . PRSice-2: Polygenic Risk Score software for biobank-scale data . GigaScience 8 , giz082 ( 2019 ). OpenUrl CrossRef PubMed 22. ↵ Lambert , S. A. et al. Enhancing the Polygenic Score Catalog with tools for score calculation and ancestry normalization . Nat. Genet . 56 , 1989 – 1994 ( 2024 ). OpenUrl PubMed 23. Lambert , S. A. et al. The Polygenic Score Catalog as an open database for reproducibility and systematic evaluation . Nat. Genet . 53 , 420 – 425 ( 2021 ). OpenUrl CrossRef PubMed 24. ↵ PGS Catalog - Parkinson disease [MONDO:0005180] (Polygenic Trait) . https://www.pgscatalog.org/trait/MONDO_0005180/ ( 2024 ). 25. ↵ Vilhjálmsson , B. J. et al. Modeling Linkage Disequilibrium Increases Accuracy of Polygenic Risk Scores . Am. J. Hum. Genet . 97 , 576 ( 2015 ). OpenUrl CrossRef PubMed 26. ↵ Privé , F. , Arbel , J. & Vilhjálmsson , B. J . LDpred2: better, faster, stronger . Bioinformatics 36 , 5424 – 5431 ( 2021 ). OpenUrl CrossRef PubMed 27. ↵ Ge , T. , Chen , C.-Y. , Ni , Y. , Feng , Y.-C. A. & Smoller , J. W . Polygenic prediction via Bayesian regression and continuous shrinkage priors . Nat. Commun . 10 , 1776 ( 2019 ). OpenUrl CrossRef PubMed 28. ↵ Ruan , Y. et al. Improving Polygenic Prediction in Ancestrally Diverse Populations . Nat. Genet . 54 , 573 ( 2022 ). 29. ↵ Zhang , Q. , Privé , F. , Vilhjálmsson , B. & Speed , D . Improved genetic prediction of complex traits from individual-level data or summary statistics . Nat. Commun . 12 , 4192 ( 2021 ). OpenUrl PubMed 30. ↵ Westenberger , A. , Brüggemann , N. & Klein , C . Genetics of Parkinson’s Disease: From Causes to Treatment . Cold Spring Harb. Perspect. Med . a041774 ( 2024 ) doi: 10.1101/cshperspect.a041774 . OpenUrl Abstract / FREE Full Text 31. ↵ Cook , L. et al. Parkinson’s disease variant detection and disclosure: PD GENEration, a North American study . Brain J. Neurol . 147 , 2668 – 2679 ( 2024 ). OpenUrl PubMed 32. ↵ Polymeropoulos , M. H. et al. Mutation in the alpha-synuclein gene identified in families with Parkinson’s disease . Science 276 , 2045 – 2047 ( 1997 ). OpenUrl Abstract / FREE Full Text 33. ↵ Klein , C. , Hattori , N. & Marras , C. MDSGene: Closing Data Gaps in Genotype-Phenotype Correlations of Monogenic Parkinson’s Disease . J. Park. Dis. 8 , S25 – S30 . 34. ↵ Mak , T. S. H. , Porsch , R. M. , Choi , S. W. , Zhou , X. & Sham , P. C . Polygenic scores via penalized regression on summary statistics . Genet. Epidemiol . 41 , 469 – 480 ( 2017 ). OpenUrl CrossRef PubMed 35. ↵ Privé , F. , Vilhjálmsson , B. J. & Mak , T. S. H. lassosum2: an updated version complementing LDpred2. 2021.03.29.437510 Preprint at doi: 10.1101/2021.03.29.437510 ( 2021 ). OpenUrl Abstract / FREE Full Text 36. ↵ Grover , S. et al. Genome-wide Association and Meta-analysis of Age at Onset in Parkinson Disease: Evidence From the COURAGE-PD Consortium . Neurology 99 , e698 – e710 ( 2022 ). OpenUrl PubMed 37. ↵ Klein , C. , Lohmann-Hedrich , K. , Rogaeva , E. , Schlossmacher , M. G. & Lang , A. E . Deciphering the role of heterozygous mutations in genes associated with parkinsonism . Lancet Neurol . 6 , 652 – 662 ( 2007 ). OpenUrl CrossRef PubMed Web of Science 38. ↵ Bobbili , D. R. , Banda , P. , Krüger , R. & May , P . Excess of singleton loss-of-function variants in Parkinson’s disease contributes to genetic risk . J. Med. Genet . 57 , 617 – 623 ( 2020 ). OpenUrl Abstract / FREE Full Text 39. ↵ Chairta , P. P. et al. Prediction of Parkinson’s Disease Risk Based on Genetic Profile and Established Risk Factors . Genes 12 , 1278 ( 2021 ). OpenUrl 40. ↵ Campêlo , C. L. das C. & Silva , R. H. Genetic Variants in SNCA and the Risk of Sporadic Parkinson’s Disease and Clinical Outcomes: A Review . Park. Dis . 2017 , 4318416 ( 2017 ). OpenUrl 41. ↵ Blauwendraat , C. , Nalls , M. A. & Singleton , A. B . The genetic architecture of Parkinson’s disease . Lancet Neurol . 19 , 170 – 178 ( 2020 ). OpenUrl CrossRef PubMed 42. ↵ Fahed , A. C. et al. Polygenic background modifies penetrance of monogenic variants for tier 1 genomic conditions . Nat. Commun . 11 , 3635 ( 2020 ). OpenUrl CrossRef PubMed 43. ↵ Kukkle , P. L. et al. Genome-Wide Polygenic Score Predicts Large Number of High Risk Individuals in Monogenic Undiagnosed Young Onset Parkinson’s Disease Patients from India. Adv . Biol . 6 , e2101326 ( 2022 ). OpenUrl 44. ↵ Global Parkinson’s Genetics Program . GP2: The Global Parkinson’s Genetics Program . Mov. Disord. Off. J. Mov. Disord. Soc. 36 , 842 – 851 ( 2021 ). OpenUrl 45. ↵ Schultz , L. M. et al. Stability of polygenic scores across discovery genome-wide association studies . Hum. Genet. Genomics Adv . 3 , 100091 ( 2022 ). 46. ↵ Ding , Y. et al. Polygenic scoring accuracy varies across the genetic ancestry continuum . Nature 618 , 774 – 781 ( 2023 ). OpenUrl CrossRef PubMed 47. ↵ Martin , A. R. et al. Current clinical use of polygenic scores will risk exacerbating health disparities . Nat. Genet . 51 , 584 – 591 ( 2019 ). OpenUrl CrossRef PubMed 48. ↵ Lim , S.-Y. et al. Uncovering the genetic basis of Parkinson’s disease globally: from discoveries to the clinic . Lancet Neurol . 23 , 1267 – 1280 ( 2024 ). OpenUrl 49. ↵ All of Us Research Program Genomics Investigators. Genomic data in the All of Us Research Program . Nature 627 , 340 – 346 ( 2024 ). OpenUrl CrossRef PubMed 50. ↵ Landoulsi , Z. et al. Genetic landscape of Parkinson’s disease and related diseases in Luxembourg . Front. Aging Neurosci . 15 , ( 2023 ). 51. ↵ Pachchek , S. et al. Accurate long-read sequencing identified GBA1 as major risk factor in the Luxembourgish Parkinson’s study . Npj Park. Dis . 9 , 1 – 14 ( 2023 ). OpenUrl 52. ↵ Lange , L. M. et al. Nomenclature of Genetic Movement Disorders: Recommendations of the International Parkinson and Movement Disorder Society Task Force – An Update . Mov. Disord . 37 , 905 – 935 ( 2022 ). OpenUrl CrossRef PubMed 53. ↵ Kopanos , C. et al. VarSome: the human genomic variant search engine . Bioinformatics 35 , 1978 – 1980 ( 2019 ). OpenUrl CrossRef PubMed 54. ↵ Kasten , M. et al. Genotype-Phenotype Relations for the Parkinson’s Disease Genes Parkin, PINK1, DJ1: MDSGene Systematic Review . Mov. Disord. Off. J. Mov. Disord. Soc. 33 , 730 – 741 ( 2018 ). OpenUrl 55. ↵ Trinh , J. et al. Genotype-phenotype relations for the Parkinson’s disease genes SNCA, LRRK2, VPS35: MDSGene systematic review . Mov. Disord. Off. J. Mov. Disord. Soc. 33 , 1857 – 1870 ( 2018 ). OpenUrl 56. ↵ Landoulsi , Z. et al. Genome-wide association study of copy number variations in Parkinson’s disease . medRxiv 2024.08.21.24311915 ( 2024 ) doi: 10.1101/2024.08.21.24311915 . OpenUrl Abstract / FREE Full Text 57. ↵ Purcell , S. & Chang , C. C. PLINK 1.9 . www.cog-genomics.org/plink/1.9/ . 58. ↵ Chang , C. C. et al. Second-generation PLINK: rising to the challenge of larger and richer datasets . GigaScience 4 , s13742-015-0047–8 ( 2015 ). 59. ↵ Purcell , S. & Chang , C. C. PLINK 2.0 . www.cog-genomics.org/plink/2.0/ . 60. ↵ Meyer , H . plinkQC: Genotype Quality Control with ‘PLINK’ . ( 2021 ). 61. ↵ O’Connell , J. et al. A General Approach for Haplotype Phasing across the Full Spectrum of Relatedness . PLOS Genet . 10 , e1004234 ( 2014 ). OpenUrl CrossRef PubMed 62. ↵ Howie , B. N. , Donnelly , P. & Marchini , J . A Flexible and Accurate Genotype Imputation Method for the Next Generation of Genome-Wide Association Studies . PLoS Genet . 5 , e1000529 ( 2009 ). OpenUrl CrossRef PubMed 63. ↵ Sudlow , C. et al. UK Biobank: An Open Access Resource for Identifying the Causes of a Wide Range of Complex Diseases of Middle and Old Age . PLoS Med . 12 , e1001779 ( 2015 ). OpenUrl CrossRef PubMed 64. ↵ Olsen , L. R . groupdata2: Creating Groups from Data . R package version 2.0.5 ( 2024 ). 65. ↵ R Core Team . R: A Language and Environment for Statistical Computing . R Foundation for Statistical Computing , Vienna, Austria ( 2024 ). 66. ↵ Wickham , H . Ggplot2: Elegant Graphics for Data Analysis . ( Springer International Publishing, Cham , 2016 ). doi: 10.1007/978-3-319-24277-4 . OpenUrl CrossRef 67. ↵ Robin , X. et al. pROC: an open-source package for R and S+ to analyze and compare ROC curves . BMC Bioinformatics 12 , 77 ( 2011 ). 68. ↵ Gilleland , E. verification: Weather Forecast Verification Utilities . R package version 1.44 ( 2024 ). 69. ↵ Signorell , A . DescTools: Tools for Descriptive Statistics . R package version 0.99.58 ( 2024 ). 70. ↵ Lüdecke , D. , Ben-Shachar , M. S. , Patil , I. , Waggoner , P. & Makowski , D. performance: An R Package for Assessment, Comparison and Testing of Statistical Models . J. Open Source Softw . 6 , 3139 ( 2021 ). OpenUrl CrossRef 71. ↵ Wei , T. & Simko , V . R package ‘corrplot’: Visualization of a Correlation Matrix (Version 0.92) . ( 2024 ). 72. ↵ Faul , F. , Erdfelder , E. , Lang , A.-G. & Buchner , A . G*Power 3: a flexible statistical power analysis program for the social, behavioral, and biomedical sciences . Behav. Res. Methods 39 , 175 – 191 ( 2007 ). OpenUrl CrossRef PubMed Web of Science View the discussion thread. Back to top Previous Next Posted March 04, 2025. Download PDF Supplementary Material 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 An improved polygenic score for Parkinson’s disease partly explains variable penetrance of genetic Parkinson’s disease 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 An improved polygenic score for Parkinson’s disease partly explains variable penetrance of genetic Parkinson’s disease Sebastian Sendel , Zied Landoulsi , Katja Lohmann , Björn-Hergen Laabs , Meike Kasten , Eva-Juliane Vollstedt , Tatiana Usnich , Alexander Balck , Daniela Berg , Dheeraj Reddy Bobbili , Max Borsche , COURAGE-PD Consortium , Andre Franke , Henrike Hanssen , Emadeldin Hassanin , Andrew A. Hicks , Ulrike M. Krämer , Rejko Krüger , Gregor Kuhlenbäumer , Lara M. Lange , Wolfgang Lieb , Brit Mollenhauer , NCER-PD Consortium , Miriam Neis , Peter P. Pramstaller , Jannik Prasuhn , Eva Schaeffer , Manu Sharma , Meike Steinbach , Claudia Trenkwalder , Norbert Brüggemann , Ana Westenberger , Michael Wittig , Inke R. König , Patrick May , Christine Klein , Amke Caliebe medRxiv 2025.02.28.25323076; doi: https://doi.org/10.1101/2025.02.28.25323076 Share This Article: Copy Citation Tools An improved polygenic score for Parkinson’s disease partly explains variable penetrance of genetic Parkinson’s disease Sebastian Sendel , Zied Landoulsi , Katja Lohmann , Björn-Hergen Laabs , Meike Kasten , Eva-Juliane Vollstedt , Tatiana Usnich , Alexander Balck , Daniela Berg , Dheeraj Reddy Bobbili , Max Borsche , COURAGE-PD Consortium , Andre Franke , Henrike Hanssen , Emadeldin Hassanin , Andrew A. Hicks , Ulrike M. Krämer , Rejko Krüger , Gregor Kuhlenbäumer , Lara M. Lange , Wolfgang Lieb , Brit Mollenhauer , NCER-PD Consortium , Miriam Neis , Peter P. Pramstaller , Jannik Prasuhn , Eva Schaeffer , Manu Sharma , Meike Steinbach , Claudia Trenkwalder , Norbert Brüggemann , Ana Westenberger , Michael Wittig , Inke R. König , Patrick May , Christine Klein , Amke Caliebe medRxiv 2025.02.28.25323076; doi: https://doi.org/10.1101/2025.02.28.25323076 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 Genetic and Genomic Medicine 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 (15229) Forensic Medicine (30) Gastroenterology (1124) Genetic and Genomic Medicine (6600) Geriatric Medicine (668) Health Economics (997) Health Informatics (4536) Health Policy (1368) Health Systems and Quality Improvement (1613) Hematology (541) 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 (9231) 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:'a0079d586d9fe2c5',t:'MTc3OTU3NjU3NA=='};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())}}}})();