Inference of causal relationships based on the genetics of cardiometabolic traits and conditions unique to females in >50,000 participants

27 28

Cardiometabolic diseases are highly comorbid and associated with poor health 29 outcomes. However, the investigation of the relationship between the genetic predisposition 30 to cardiometabolic diseases with the risk of conditions unique to females such as breast 31 cancer, endometriosis and pregnancy -related complications is highly understudied. This 32 study ai med to estimate the cross -trait genetic overlap and influence of genetic burden of 33 cardiometabolic traits on health conditions unique to females. 34

We obtained data for female participants in the Penn Medicine BioBank (PMBB; 35 21,837 samples ) and the electronic MEdical Records and GEnomics (eMERGE ; 49,171 36 samples) network. We examined the relationship between four cardiometabo lic phenotypes 37 (body mass index (BMI), coronary artery disease (CAD), type 2 diabetes (T2D) and 38 hypertension (through blood pressure measurements )) and 23 female health conditions by 39 performing four analyses: 1) C ross-trait genetic correlation analyses t o compare genetic 40 architecture. 2) Polygenic risk scores (PRS) -based association tests to characterize shared 41 genetic effects on disease risk . 3) Mendelian randomization (MR) for significant associations 42 to assess cross -trait causal relationships. 4) Chronology analyses to visualize the timeline of 43 events unique to groups of females with high and low genetic burden for cardiometabolic 44 traits and highlight the disease prevalence in risk groups by age. 45

We observed high genetic correlation among cardiometabolic and female health 46 conditions. PRS meta -analysis identified 29 significant associations reflecting potential 47 shared biology among common cardiometabolic phenotypes and female health conditions. 48 Significant associations include PRSBMI with endometrial cancer and polycystic ovarian 49 syndrome (PCOS), PRSCAD with breast cancer, and the PRST2D with gestational diabetes and 50 PCOS. Mendelian randomization provided additional evidence of independent causal effects 51 between T2D and gestational diabetes and CAD and with breast cancer. Our results reflected 52 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 3 inverse association between PRS CAD and breast cancer. Lastly, as visualized from chronology 53 analyses, individuals with high PRS are also more likely to develop conditions such as PCOS 54 and gestational hypertension at earlier ages. 55

Polygenic susceptibility to cardiometabolic traits is associated with conditions 56 unique to females . Several of these associations are likely to resu lt from the complex 57 pathophysiology of cardiometabolic risk, and others may reflect potential pleiotropic effects 58 that go beyond cardiometabolic health in females. 59 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 4

60 Cardiometabolic diseases such as coronary artery disease (CAD), obesity, 61 hypertension, and type 2 diabetes (T2D) are profoundly prevalent and among the leading 62 causes of death in the world 5,6,7. Cardiometabolic conditions are highly comorbid and 63 considered as risk factors for s equelae of many diseases, such as depression, anxiety, chronic 64 obstructive pulmonary disease (COPD), and cancer, to name a few 8,9. These conditions 65 affect female individuals disproportionately because they are also linked to disorders of the 66 reproductive system and adverse outcomes of pregnancy and childbirth such as preeclampsia, 67 gestational diabetes, stillbirth, and pregnancy loss 10,11. Many studies suggest that the 68 pathophysiology of cardiometabolic diseases affects males and females differently 14. There 69 are multiple shreds of evidence supporting the relationship between female health conditions 70 and cardiometabolic diseases. For instance , people who develop preeclampsia during 71 pregnancy are more likely to develop cardiovascular diseases and hypertension after 72 pregnancy15,16. Additionally, obesity and polycystic ovarian syndrome (PCOS) are closely 73 linked conditions, and people with PCOS are at high risk of developing T2D17-19. People with 74 endometriosis are also at high risk for developing oncological cancers and cardiovascular 75 diseases such as myocardial infarction and ischemic heart disease 20,21. However, the 76 relationship between female health and cardiometabolic phenotypes, particularly the potential 77 for any shared genetic burden between them, is still highly understudied. The investigation of 78 shared ge netic burden could lead to identifying obesity and cardiometabolic cluster -related 79 risk factors associated with female health conditions to modify standard screening practices 80 in people at high risk for many diseases. 81 Genome-wide association studies (GWAS) in the past couple of decades have 82 exposed common cross-trait connections. Disease traits such as type 2 diabetes, obesity, 83 sleep apnea, hypertension, Alzheimer’s diseases, and many cancer types have been shown to 84 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 5 share genetic etiology 1-4. GWAS have identified >1000 loci with a likely impact on 85 cardiometabolic phenotypes, but the effect size of any single variant is generally tiny. Many 86

have identified relationships between different phenotypes and traits by calculating 87 their genetic correlation from GWAS effect sizes. In addition, to more accurately estimate an 88 individual’s overall risk for disease, researchers have used GWAS to calculate polygenic risk 89 scores (PRS), which sum up the effect of common single nucleotide polymorphisms (SNPs) 90 throughout the genome into a single score of overall genetic burden for a phe notype27. PRS 91 have been shown to predict the risk of many cardiometabolic diseases and its 92 comorbidities28,29. 93 Our approach to investigating the impact of genetic burden of cardiometabolic t raits 94 in female health conditions is to measure the genetic correlation and association of 95 cardiometabolic PRS with EHR -derived phenotypes. Many prior studies have successfully 96 used PRS as the genetic risk factor for identifying links between overall genetic risk for one 97 phenotype to other phenotypes31,32. PRS-based association tests make fewer assumptions and 98 have the benefit that they are based on unvarying risk factors (i.e., inherited genetic burden). 99 We hypothesize that the genetic burden for cardiometabolic phenotypes could further explain 100 the phenotypic variance in health conditions unique to females. This present study obtained 101 genotyped data and a wide range of female health conditions from the Penn Medicine 102 BioBank (PMBB) and the Electronic Medical Records and Genomics (eMERGE) Network. 103 We measured pairwise genetic correlation between cardiometabolic phenotypes and female 104 health conditions to identify the strength of shared genetic factors. The association of PRS 105 with multiple selected phenotypes is further used to estimate the significance of associations 106 between cardiometabolic genetic burden and female health conditions . We additionally 107 evaluated the causal relationship between significant associations by using Me ndelian 108 randomization (MR) approach es. Lastly, electronic health record (EHR) data provides the 109 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 6 opportunity to map female health by considering the disease prevalence by age. We 110 generated a chronological map of diseases in participants in high and low PRS groups to 111 understand the prevalence of diseases in the two risk groups at different ages. 112

The authors h ave individual level access to genotype and medical record data 113 from t he Penn Medicine BioBank and the Electronic Medical Records and Genomics 114 network datasets. 115 Study Populations 116 Penn Medicine BioBank: The Penn Medicine Bio Bank (PMBB) is a University of 117 Pennsylvania academic biobank which recruits patient-participants from the University of 118 Pennsylvania Health System around the greater Philadelphia area in the United States. 119 PMBB links an individual’s genotype data with detailed electronic health record ( EHR) 120 information. Currently, PMBB consists of genome-wide genotyped and whole -exome 121 sequenced data on ~45 ,000 samples. PMBB is a diverse cohort, with over 25% of 122 participants of African ancestry. PMBB genotyped data is imputed to TOPMED Reference 123 panel using the Michigan Imputation server. We included 21,837 female participants from 124 PMBB in this study (Table 1). The stratified analyses in this study only included participants 125 of European and African ancestry and excluded those from Asian and Hispanic ancestries 126 due to the limiting sample size. 127 128 eMERGE: The Electronic Medical Records and Genomics (eMERGE) Network is a 129 nationwide consortium containing individuals with genome-wide genotyped data linked to 130 EHRs from several health systems across the United States, with most participants from the 131 Geisinger Health System and Vanderbilt University. eMERGE data is imputed to Haplotype 132

Consortium panel using Michigan Imputation Server. Like PMBB, the eMERGE 133 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 7 cohort is diverse across ancestries and ages. This study was performed on 49,171 female 134 patients in eMERGE born after 2001 (Table 1). 135 Genome-wide association studies for cardiometabolic phenotypes 136 Most genetic correlation and PRS calculation methods require effect sizes of variants 137 on the phenotypes determined through large GWAS. Given the diverse nature of our study 138 population, w e obtained the largest publicly available multi -ancestry GWAS summary 139 statistics for six cardiometabolic phenotypes: obesity (measured through body mass index 140 (BMI), CAD, hypertension (measured through diastolic blo od pressure (DBP), systolic blood 141 pressure (SBP), and pulse pressure (PP)), and T2D. The summary statistics used for each 142 phenotype and its respective study is referenced in Table 2. 143 144 Genome-wide association for female health conditions 145 Large multi-ancestry GWAS are not available for most female health conditions 146 evaluated in this study. Thus, we used PLINK version 1.90 to conduct GWA S for the female 147 health conditions in PMBB and eMERGE datasets 33. We filtered the variants in the PMBB 148 imputed data sets to include only those with imputation quality R 2 > 0.3 and minor allele 149 frequency > 0.01. We then meta-analyzed the GWAS from our two cohorts in PLINK. 150 Genetic correlation calculation 151 We calculated pairwise genetic correlation between cardiometabolic phenotypes and 152 female health conditions using LD score regression (LDSC), with the large publicly available 153 cardiometabolic GWAS and the meta -analyzed GWAS for female health conditions from 154 PMBB and eMERGE as input GWAS 34. LDSC accounts for linkage disequilibrium (LD) 155 among SN Ps by using an external reference panel that should also match the anc estry 156 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 8 distribution of the GWAS . We generated a multi -ancestry LD reference panel using the 157 HapMap3 SNPs (~1M common variants) from the entire 1000 Genomes population. 158 Polygenic Risk Scores (PRS) 159 PRS calculated using GWAS performed in one ancestry group tend to perform poorly 160 in individuals from different ancestries 27,35,36. To accurately calculate a PRS for our diverse 161 target datasets (PMBB and eMERGE), we calculated a PRS for the six different 162 cardiometabolic phenotypes using the large publicly available multi-ancestry GWAS used in 163 the genetic correlation analyses. Weights for each SNP included in the PRS were calculate d 164 using PRS -CS (version from Apr 24, 2020) 38. Like LDSC, PRS-CS requires a reference 165 panel that match es the ancestry distribution of the target dataset, and we generated a similar 166 multi-ancestry LD reference panel using the HapMap SNPs from 1000 Genomes. We used 167 PLINK(v1.90) to identify LD blocks and calculate the LD between the SNPs in each bloc k. 168 For PRS -CS, the gl obal shrinkage parameter phi was fixed to 0.01, and default values were 169 selected for all other parameters. PRS was then calculated using these weights through 170 PLINK. Only the SNPs in the target dataset, summary statistics, and LD reference panel were 171 included in the PRS. The number of SNPs used for PRS calculation is listed in Table 2 . 172 Scores were then normalized to obtain meaningful beta coefficients. 173 To evaluate the power of PRS, we tested the performance of each PRS on predicting 174 the primary ph enotype of the summary statistics. We could not obtain any blood pressure 175 quantitative measurements for participants from eMERGE , and PP and SBP measurements 176 were not curated in PMBB. Therefore, w e evaluated the performance of the PRS from blood 177 pressure t raits ( SP, DBP , PP) in eMERGE on hypertension case-control phenotype and in 178 PMBB by predicting DBP or hypertension (for SBP and PP) as outcomes. We constructed 179 logistic regression models for binary phenotypes (CAD, T2D, and hypertension) , and 180 evaluated PRS performance based on the area under the receiver -operator curve (AUC) using 181 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 9 R’s pROC package. Similarly, we constructed linear regression models for continuous 182 phenotypes (BMI and DBP) and evaluated them based on the R2 using R’s glm function. The 183 regression models used birth year and the first five principal components (PCs) as covariates. 184 We tested PRS performance in all individuals as well as in only European and African 185 ancestry individuals. 186 Phenotype data 187 Cases and controls for each phenotype were defined using International Classification 188 of Disease (ICD) diagnosis codes. Participants were coded as cases for a phenotype if they 189 had at least one occurrence of the corresponding ICD codes. For pregnancy -related 190 phenotypes, participants were only considered as controls for a phenotype if they had at least 191 one occurrence of a pregnancy -related ICD code and no occurrences of codes for certain 192 complications during pregnancy (such as miscarriage) or any of the ca se ICD codes. 193 Participants were counted as controls for cardiometabolic and all other female health 194 conditions if they did not have any of the case ICD codes. The complete list of all ICD cod es 195 used to include or exclude participants as cases and controls can be found in Supplementary 196 Table 1. Using these definitions, we determined the sample size for each phenotype in 197 eMERGE and PMBB (Table 1). 198 PRS association analyses 199 We tested the association of each cardiometabolic PRS and female health conditions 200 by fitting separate logistic regression models, adjusting the models by birth year and the fi rst 201 five PCs. We conducted this analysis in all participants as well as ancestry -stratified analyses 202 for individuals of European and African ancestry. To account for biases from multiple 203 hypothesis testing, we determined if these associations passed an FDR -significance threshold 204 of 0.05, using the number of female health conditions (23) as the number of hypotheses 205 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 10 tested. The logistic regressions were performed using R’s glm function, and results from 206 PMBB and eMERGE were combined using the rma function from the metafor R package 207 under the restricted maximum -likelihood estimator model 37. We used PheWAS -View to 208 visualize our results 39. We then created prevalence plots for each significant association . 209 Participants were di vided into quintiles based on their PRS, and the percentage of cases for 210 the most significant female health conditions was calculated at each quintile. 211 Mendelian Randomization 212 To identify potential evidence of causality between cardiometabolic phenotypes and 213 female health conditions , we performed one-sample Mendelian r andomization (MR) for 29 214 significant associations from PRS analyses using the ivreg function from the ivpack R 215 package. Cardiometabolic PRS were used as genetic instruments in the one-sample MR, with 216 birth year and the first five PCs included as covariates. Results were combined through meta-217 analysis using the metafor R package as done in the PRS meta-analysis. Since the low sample 218 size for some of the female health conditions could limit the power of our analyses , we also 219 performed two-sample MR for the same associations using the inverse variance weighted 220 (IVW) method in the twoSampleMR package in R40. MR sensitivity analyses were conducted 221 using the weighted median and the MR Egger methods through the same package. Genetic 222 instruments for cardiometabolic phenotypes in the two -sample MR were defined as genome-223 wide significant SNPs in the GWAS summary statistics used to calculate the PRS. SNPs 224 were LD pruned according to LD patterns i n the 1000 Genomes HapMap SNPs, and the 225 representative SNPs were included in the analysis. Matching genetic instruments for female 226 health conditions were obtained from publicly available GWAS available through the 227 twoSampleMR package (Supplementary Table 2). 228 229 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 11 Chronology Analyses 230 We divided participants into high -risk and low -risk groups according to each 231 cardiometabolic PRS. High PRS was defined as the top quintile (PRS > 80 percentile), and 232 low PRS was defined as the bottom quintile (PRS < 20 percentile). We obtained the age at 233 the first occurrence of each female health condition according to the ICD record. For 234 pregnancy-related conditions, participants were split into three age groups: < 25, 25-40, and 235 40-55. We excluded participants who were over 55 at the first occurrence of pr egnancy-236 related conditions due to low sample sizes and potential errors in diagnosis coding. For all 237 other conditions, participants were split into five different age groups: 70. We examined the combined case prevalence in PMBB and eMERGE for female 239 health conditions in high and low PRS groups across all age groups. 240 241

242 Genetic correlation among cardiometabolic phenotypes and female health conditions 243 The six cardiometabolic phenotypes were significantly correlated with six different 244 female health conditions for a total of 13 statistically significant correlations (Figure 1A). 245 BMI was significantly correlated with breast cancer (R g=-0.179, p=0.011) and PCOS 246 (Rg=0.4, p=0.007), CAD with breast cancer (R g=-0.199, p=0.011), PCOS (R g=0.216, 247 p=0.04), and postpartum depression (R g=0.204, p=0.025), PP with gestational hypertension 248 (Rg=0.481, p=0.0025), SBP with breast cancer (R g=-0.153, p=0.041) and gestational 249 hypertension (R g=0.523, p=0.0017), and T2D with breast cancer (R g=-0.196, p= 0.0001), 250 excessive fetal growth (R g=0.126, p= 0.044), gestational diabetes (R g=0.529, p= 0.011), 251 gestational hypertension (Rg=0.237, p=0.028), and PCOS (Rg=0.316, p=0.0093). 252 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 12 PRS performance on primary phenotype 253 We calculated a PRS for six cardiometabolic phenotypes for individuals in PMBB 254 and eMERGE and checked t he distribution of the raw and normalized PRS (Supplementary 255 Figures 1-6). The full model for all PRS generally performed well in predicting the primary 256 phenotype across ancestry groups (Table 3). PRS was significantly (p < 0.05) associated with 257 the primary phenotype for all cardiometabolic phenotypes. The covariate only model (null 258 model) generally performed better in Afri can ancestry individuals than in European 259 individuals, but the PRS only model generally performed better in European ancestry 260 individuals. We calculated the difference between the full and null models for all 261 cardiometabolic PRS in PMBB and eMERGE and foun d that the PRS improved predictive 262 performance significantly more for European ancestry participants than for African ancestry 263 participants (p = 0.000488, Wilcox s igned rank test). As such, the cardiometabolic PRS were 264 more accurate in European ancestry in dividuals but still significantly improved predictive 265 performance in African ancestry individuals. 266 Association of cardiometabolic PRS association and female health conditions 267 We detected numerous associations between cardiometabolic PRS and female health 268 conditions in the meta -analyzed results (Figure 1 B-D). 29 associations were statistically 269 significant after correcting for multiple hypothesis burden through FDR significance ( Table 270 4). Most of these associations were also significantly associated in both PMBB and eMERGE 271 separately (Supplementary Figures 7-12). In the meta -analysis, for all and European ancestry 272 individuals, the PRS BMI was significantly associated with endometri al cancer (betaall=0.24, 273 seall=0.046, pall=9.4x10-8; beta eur=0.28, seeur=0.087, peur=0.0015), gestational diabetes 274 (betaall=0.23, seall=0.051, pall =6x10-6; beta eur=0.28, seeur=0.062, peur=8.7x10-6), and PCOS 275 ((betaall=0.27, seall=0.039, pall=2.4x10-12; betaeur=0.29, seeur=0.045 peur=6.8x10-11). The 276 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 13 PRSBMI was also significantly associated with breast cancer in all individuals (betaall=-0.071, 277 seall=0.022, pall=0.0016). These results suggest that an increased genetic burden for obesity 278 and high BMI also increases the risk for endometrial cancer, gestational diabetes, and PCOS 279 but decreases the risk for breast cancer . The PRS CAD was also significantly associated with 280 breast cancer for all and European ancestry individuals (betaall=-0.072, seall=0.015, pall=1x10-281 6; beta eur=-0.07, seall=0.017, peur=3.1x10-5). This highly significant negative association 282 suggests that individuals, particularly those of European ancestry, with high PRSCAD are at 283 relatively lower risk for breast cancer compared to individuals with low PRSCAD. The PRST2D 284 was also significantly associated with gestational diabetes in all and European ancestry 285 individuals (betaall=0.58, seall=0.063, pall=1.2x10-20; betaeur=0.68, seeur=0.076, peur=3.9x10-19 ) 286 and with PCOS in all, European ancestry, and African ancestry individuals (betaall=0.22, 287 seall=0.043, pall=1.9x10-7; beta eur=0.21, seeur=0.049, peur=1.6x10-5; beta afr=0.32, seafr=0.1, 288 pafr=0.0021). The link between T2D and these female related phenotypes is well known, and 289 our results support a potential genetic basis linking these phenotypes 18,19. Among all and 290 European ancestry individuals, the PRS T2D was also significantly associated with gestational 291 hypertension (betaall=0.18, seall=0.064, pall=0.0041; betaeur=0.18, seeur=0.067, peur=0.008) and 292 breast cancer (betaall=-0.073, seall=0.021, pall=0.00066; beta eur=-0.079, seeur=0.026, 293 peur=0.0027). 294 The three-blood pressure traits PRS ( PRSDBP, PRS SBP, and PRSPP) showed varying 295 significant associations with gestational hypertension and preeclampsia in the PMBB and 296 eMERGE meta-analysis. There was no sign of association between the PRSDBP and these 297 phenotypes in the meta -analysis. However, t he PRS DBP was significantly associated with 298 gestational hypertension in PMBB all and Africa n ancestry individuals ( p=1x10-05, 0.00041) 299 and nominally associated (p<0.01) with preeclampsia in African anc estry participants 300 (p=0.0063). The PRSPP was significantly associated with gestational hypertension for all and 301 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 14 European ancestry participants ( betaall=0.22, seall=0.044, pall=8.5x10-7; beta eur=0.24, 302 seeur=0.056, peur=2.5x10-5) and with preeclampsia for all participants (betaall=0.2, seall=0.058, 303 pall=0.00076). The PRSSBP was significantly associated with gestational hyper tension and 304 preeclampsia in all, European ancestry, and African ancestry individuals (Gestational 305 hypertension: betaall=0.33, seall=0.083, pall=5.6x10-5; betaeur=0.31, seall=0.066, peur=3.6x10-6; 306 betaafr=0.4, seafr=0.099, pafr=5.7x10-5; Preeclampsia: beta all=0.28, seall=0.08, pall=0.00057; 307 betaeur=0.23, seeur=0.071, peur=0.0013; betaafr=0.36, seafr=0.1, pafr=0.00052). 308 For each FDR-significant association in both PMBB and eMERGE , we looked at the 309 case prevalence of the female health condition per the associated PRS quintile (Figure 2B 310 and Supplementary Figures 13-17). The trends matched the results we obtained from the 311 association analyses across ancestry groups and in both datas ets. Distribution of the number 312 of cases increased as the PRS percentile increased for the positive associations in the 313 association analysis, such as PCOS with the PRSBMI. The number of cases for breast cancer 314 decreased when increasing the PRSCAD percentile, matching the inverse association between 315 breast cancer and the PRSCAD (Figure 2B). 316 317 Association of PRSCAD and breast cancer 318 To validate the inverse association between PRS CAD and breast cancer, we examined 319 the genetic correlation between CAD and breast cancer from the UKBB Genetic Correlation 320 Browser (Figure 2A) . The genetic correlation between I9_CHD (Major coronary heart 321 disease event) and C50 (Diagnoses -main ICD10: C5 0 Malignant neoplasm of breast) was 322 significantly negative (Rg=-0.24, p =0.0325). 323 Additionally, t o evaluate the risk of ascertainment biases and the reporting of 324 comorbidities in the EHR, we preformed the associations of CAD PRS with breast cancer in 325 females w ho are not diagnosed with CAD (n= 61,201). In the individuals who were not 326 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 15 diagnosed with CAD, we observed slightly less significant association between breast cancer 327 and CAD ( betaall=-0.0484, p all=0.0045; betaeur=-0.0464, p eur=0.0112) than in the full set of 328 individuals. T hese associations attenuated but were not completely diminished when 329 evaluating participants who were not diagnosed with CAD. Next, given the longitudinal 330 nature of availability of comorbidities in the EHR, we aimed to measure the impact of risk of 331 CAD on the first reported incidence of CAD and breast cancer in females. The overlap on 332 individuals who were diagnosed with CAD and breast cancer in our study population 333 suggests that females with a diagnosis of CAD are less likely to be diagnosed or report a 334 history of breast cancer (Figure 2C). 335 336 Mendelian Randomization 337 We performed one -sample MR in PMBB and eMERGE for the 29 associations that 338 were FDR-significant and combined them through a meta -analysis (Figure 3A). The majority 339 of analyses were significant (p < 0.05), and the direction of the estimated beta coefficient for 340 each association aligned with the direction of effect seen in the PRS association analysis. Our 341

support a potential causal relationship between many cardiometabolic phenotypes and 342 female health conditions , with the most significant ass ociations being between T2D and 343 gestational diabetes ( All: p=1.5x10-11, beta=1.52, se=0.23 ; EUR: p=2.4x10-10, beta=1.78, 344 se=0.28) and BMI and PCOS (All: p=1.1x10 -8, beta=0.0021, se=0.00037; EUR: p=3.6x10-8, 345 beta=0.0021, se=0.0038). 346 We also performed two -sample MR to leverage the power of larger GWAS for our 347 female health conditions ( Figure 3B). Most analyses were significant when us ing the IVW 348

but became less significant when using the MR Egger and the weighted median 349 methods. Some associations remained significant for all three methods, particularly CAD 350 with breast cancer (IVW: beta= -0.0587, se=0.021, p=0.00463; MR Egger: beta=-0.102, 351 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 16 se=0.049, p=0.04; Weighted median: beta= -0.0642, se=0.028, p=0.021) and T2D and 352 gestational diabetes (IVW: beta= 0.613, se=0.044, p=1.73x10-43; MR Egger: beta= 0.656, 353 se=0.11, p=8.8x10-10; Weighted median: beta=0.635, se=0.066, p=8.12x10-22). 354 355 Investigation of the role of population stratification 356 Population stratification could have confounded our results in the PRS a ssociation 357 and the MR analyses. Therefore, we tested the association of the PRS with the PCs 358 (Supplementary Table 3 ). The high R 2 for most cardiometabolic PR S shows that a large 359 amount of the variance in the PRS could be explained by the PCs when considering all multi -360 ancestry variables, and the R 2 decreases when analyzing only European or African ancestry 361 groups. The PCs explained more of the variance in PRS for African ancestry individuals than 362 European ancestry individuals, a finding that suggests there is more population stratification 363 among our African ancestry individuals and that there could be confounding factors 364 influencing some results. To overcome these biases, we accounted for PCs in both our PRS 365 association analyses and the one -sample MR analyses. Additionally, we ran one-sample MR 366 without adjusting for the covariates (PCs and birth year) and compared the results with the 367 adjusted ones (Supplementary Table 4). 368 369 Chronology analyses 370 EHR is a powerful resource to visualize the landscape of events in a patient ’s 371 diagnosis. Continuing on the theme of genetic correlation and the influence of shared genetic 372 burden of cross -trait analyses, we inv estigated if a n individual’s genetic burden for 373 cardiometabolic diseases may also affect the time at which they develop certain female health 374 conditions. Case prevalence for female health conditions was generally higher in the high 375 PRS group in the younger age groups (Figure 4 and Supplementary Figures 18-22). Many 376 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 17 pregnancy-related phenotypes were most prevalent in the 25 -40 years age group , but for 377 participants who developed those phen otypes in the < 25 years age group , the relative 378 difference in prevalence between the high -risk and low -risk PRS group was large r. The 379 cardiometabolic PRS used to define high -risk and low-risk groups did not seem to affect case 380 prevalence for most phenotypes, but phenotypes found to be associated with sp ecific 381 cardiometabolic PRS in the association analysis did show differences across PRS. Gestational 382 hypertension and preeclampsia were generally more prevalent in the younger age groups (<25 383 and 25-40) for participants with high blood pressure PRS than for participants with high PRS 384 for other phenotypes. Similar trends were observed for other traits, such as the PRS T2D with 385 gestational diabetes and the PRS BMI with PCOS. High prevalence of endometriosis and 386 uterine fibroid cases were also observed in individ uals with high-risk PRS, particularly in the 387 age group 25-40. 388

389 We have demonstrated high genetic correlation and shared genetic burden between 390 cardiometabolic traits and female health conditions spanning across many obstetrics and 391 gynecological disorders. Additionally, this study shows that genome -wide PRS of 392 cardiometabolic traits has potential translational effects in health conditions unique to 393 females. Initially, w e calculated the genetic correlation between phenotypes related to 394 cardiometabolic diseases and traits with female health conditions. Genetic correlation 395 analyses suggested a high overlap among the shared genetic etiology of all phenotypes tested 396 in this study . We further investigated the effect of this shared genetic burden by gen erating 397 PRS for cardiometabolic phenotypes and testing their association to various female-specific 398 disease conditions. PRS was generally predictive of the primary phenotype, and we identified 399 several FDR -significant associations with female health conditions that were statistically 400 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 18 significant after meta -analyzing both datasets . Prior research has established relationshi ps 401 (mainly non-genetic factors) between many of these associations, such as BMI with PCOS 402 and T2D with gestational diabetes 17,18. Epidemiological studies have also shown evidence 403 that people with obesity are at high risk for endometrial cancer 41. The associations identified 404 in this study between female health conditions and cardiometabolic PRS suggest genetic basis 405 for these cross -trait categories. In addition, we performed Mendelian randomizat ion and 406 identified potential causal relationships. 407 Unexpectedly, all of our analyses showed that t he PRS CAD was inversely associated 408 with breast cancer in European ancestry participants. CAD and breast cancer share many 409 common risk factors, such as smoking and diet 42. However, our results suggest that a high 410 genetic burden for CAD is protective for breast cancer . Participants in our cohorts with high 411 PRSCAD had lower incidence of breast cancer. When we evaluated the individuals with high 412 PRSCAD but no CAD diagnoses, we still saw a moderate but decreased risk for breast ca ncer. 413 Several factors might contribute to these associations. First, these patterns might reflect 414 ascertainment biases and competing risks. Individuals with higher risk of CAD likely live 415 shorter lives and thus are not diagnosed with breast cancer. Second, the treatments and drugs 416 given for treating CAD might also protect against breast cancer. Lastly, genetic mechanisms 417 predisposing individuals to CAD might show protective effects for breast cancer. Briefly, 418 among the SNPs that passed genome -wide significa nce in the CAD GWAS we used to 419 calculate the PRS CAD, 125/220 SNPs had opposite directions of effects in the breast cancer 420 GWAS we used in the two -sample MR . For example, rs1011970 (CAD: beta= -0.0407, 421 p=6.94x10-10; breast cancer: beta= 0.053, p= 3.1x10-5) is a known risk variant for both CAD 422 and breast cancer and maps to the CDKN2B-AS1 gene, which helps silence genes 423 epigenetically in the CDKN2A-CDKN2B cluster43. CDKN2B-AS1 knockdown suppresses 424 breast cancer progression, and decreased expression of CDKN2B increases development of 425 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 19 atherosclerotic plaques44-46. Variants that promote decreased CDKN2B-AS1 expression could 426 increase breast cancer risk but increase expression of CDKN2B and subsequently decrease 427 CAD risk. A recent study also shows this protective effect of a PRS CAD on breast cancer, and 428 more research should be undertaken to uncover the mechanisms behind this association47. 429 The differences between the three PRSblood pressure suggests that analyzing the genetic 430 burden for all three blood pressure traits might elucidate the understanding of hypertension 431 during pregnancy . Only the PRS SBP was significantly associated with gestational 432 hypertension and preeclampsia in African ancestry individuals . Prior studies have identified 433 variability in the predictive performance when using different blood pressure measurements 434 to predict hypertension and other diseases, and our results warrant further studies to replicate 435 this association in other large multi-ancestry datasets to reach a more definite conclusion on 436 the role blood pressure measurements play in predicting hypertensive diseases during 437 pregnancy48,49. 438 To further investigate the effect of cardiometabolic genetic burden on conditions 439 unique to females, we drew information from EHRs to capture the landscape of trajectories of 440 diseases presented at different ages. We found that a high genetic burden for most 441 cardiometabolic phenotypes could increase risk of developing female health conditions at 442 earlier ages, even if there was no overall association between the PRS and the condition. For 443 example, the PRS BMI was not associated with many pregnancy -related complications such as 444 gestational hypertension and ectopic pregnancy. However, in the youngest a ge group (<25), 445 these phenotypes were more prevalent in the high PRS BMI group than the low PRSBMI group. 446 Similarly, there was no association overall between the PRS SBP and PCOS or endometriosis, 447 but participants who developed PCOS before 25 years old or endometriosis between 25 -40 448 years old were mo re likely to have high PRS. The difference in case prevalence between high 449 and low PRS groups for these phenotypes decreased in older age groups. Patients with high 450 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 20 genetic burden for cardiometabolic phenotypes could be at higher risk to develop female 451 health conditions at younger ages . These findings can be particul arly of importance to 452 prioritize patients for early screening. 453 The l imitations of our study point to many promising future directions. First, our 454 study establishes a link between the genetic burden for different cardiometabolic phenotypes 455 and several diseases unique to females. However, we estimate the genetic burden by using 456 PRS alone, which are calculated based on common variants and do not include the effects of 457 other genetic risk factors such as rare variants and copy number variations. Second, our 458 analysis does not consider clinical or environmental factors that could influence the 459 associations between cardiometabolic and female health conditions . We found that 460 population stratification is po tentially present in our datasets and accounted PCs in our 461 analyses accordingly. However, important social and environmental risk factors such as 462 education level and socioeconomic status were missing and could not be properly accounted 463 for in our current analyses. Our focus on PRS alone reflects the current limitations of 464

for multi -modal risk model predictions. Current widespread interest in building 465 integrative risk models suggests that this gap can be closed in the near future 27,50. 466 Furthermore, the weaker performance o f PRS in African ancestry individuals contributed to 467 the lack of power to identify African ancestry specific associations and suggests the urgent 468 need for expanding studies to include more racially and ethnically diverse cohorts. 469 The low sample size for so me conditions may have limited the power of our one -470 sample MR analyses. In addition, w hile our two-sample MR uses GWAS from larger cohorts, 471 they were performed on populatio ns of primarily European ancestry and thus lack the power 472 to detect causality for th e non-European ancestry participants in our cohort. The effects we 473 see in our MR results may be biased due to h orizontal pleiotropy, in which genetic variants 474 associated with the exposure (cardiometabolic phenotypes) affect other traits that 475 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 21 subsequently influence the outcome (female health conditions). The effects of pleiotropy can 476 be seen from the decreased statistical significance when using the weighted median and MR 477 Egger methods in the two-sample MR. For example, the relationship between T2D and PCOS 478 became insignificant when using the MR Egger method (p = 0.109) compared to the IVW 479

(p = 0.000148). These methods account for pleiotropy and other confounding factors 480 but may not have captured all their effects. 481 EHRs are particularly advantageous in investigating disease trajectories and 482 progression. Our analyses in this study provided a big picture visualization of the burden of 483 early diagnoses of disease unique to females at early ages among the hi gh cardiometabolic 484 PRS group. However, these analyses might point to the ascertainment biases of EHR and 485 including confounding factors for statistically informed findings. This study illustrates the 486 influence of cardiometabolic geneti c burden on diverse ph enotypes. Our findings serve as the 487 initial basis for presenting the clinical utility of cardiometabolic PRS as non -modifiable risk 488 factors for screening and early diagnostic tools for a variety of obstetr ic and gynecological 489 conditions. To improve the pow er of cardiometabolic PRS for predicting risk for female 490 health conditions and better understand the relationship between these phenotypes, future 491 studies should incorporate PRS with other genetic and non-genetic risk factors and study their 492 effects on larger and more diverse multi-ancestry populations. 493 494 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 22

We acknowledge the Penn Medicine BioBank (PMBB) for 495 providing data and thank the patient -participants of Penn Medicine who consented to 496 participate in this research program. We would also like to thank the Penn Medicine BioBank 497 team and Regeneron Genetics Center for providing genetic variant dat a for analysis. The 498 PMBB is approved under IRB protocol #813913 and supported by Perelman School of 499 Medicine at University of Pennsylvania. Approval for PMBB was given from the University 500 of Pennsylvania Institutional Review Board. The PMBB study has been determined to pose 501 minimal risk to subjects. 502 503 Sources of Funding : The eMERGE Network was initiated and funded by NHGRI 504 through the following grants: 505 Phase IV: U01HG011172 (Cincinnati Children’s Hospital Medical Center); U01HG011175 506 (Children’s Hospital of Philadelphia); U01HG008680 (Columbia University); U01HG011176 507 (Icahn School of Medicine at Mount Sinai); U01HG008685 (Mass General Brigham); 508 U01HG006379 (Mayo Clinic); U01HG011169 (Northwestern University); U01HG011167 509 (University of Alabama at Birmingha m); U01HG008657 (University of Washington Medical 510 Center, Seattle ); U01HG011181 (Vanderbilt University Medical Center); U01HG011166 511 (Vanderbilt University Medical Center serving as the Coordinating Center) 512 Phase III: U01HG8657 ( Kaiser Permanente Washingtom/University of Washington Medical 513 Center); U01HG8685 (Brigham and Women’s Hospital); U01HG8672 (Vanderbilt University 514 Medical Center); U01HG8666 (Cincinnati Children’s Hospital Medical Center); U01HG6379 515 (Mayo Clinic); U01HG8679 (Geisinger Clinic); U01HG 8680 (Columbia University Health 516 Sciences); U01HG8684 (Children’s Hospital of Philadelphia); U01HG8673 (Northwestern 517 University); U01HG8701 (Vanderbilt University Medical Center serving as the Coordinating 518 Center); U01HG8676 (Partners Healthcare/Broad Inst itute); and U01HG8664 (Baylor 519 College of Medicine). 520 521 Disclosures: None 522

523 1. Chang C, Zhang K, Veluchamy A, et al. A Genome-Wide Association Study Provides 524 New Evidence That CACNA1C Gene is Associated With Diabetic Cataract. Invest 525 Ophthalmol Vis Sci. 2016;57(4):2246-2250. 526 2. Patel SR. Shared genetic risk factors for obstructive sleep apnea and obesity. J Appl 527 Physiol (1985). 2005;99(4):1600-1606. 528 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 23 3. Zhuang QS, Zheng H, Gu XD, Shen L, Ji HF. Detecting the genetic link between 529 Alzheimer's disease and obesity using bioinformatics analysis of GWAS 530 data. Oncotarget. 2017;8(34):55915-55919. 531 4. Rashkin SR, Graff RE, Kachuri L, et al. Pan-cancer study detects genetic risk variants 532 and shared genetic basis in two large cohorts. Nat Commun . 2020;11(1):4423. 533 Published 2020 Sep 4. 534 5. Blüher M. Obesity: global epidemiology and pathogenesis. Nat Rev Endocrinol . 535 2019;15(5):288-298. 536 6. Saeedi P, Petersohn I, Salpea P, et al. Global and regional diabetes prevalence 537 estimates for 2019 and projections for 2030 and 2045: Results from the International 538 Diabetes Federation Diabetes Atlas, 9 th edition. Diabetes Res Clin Pract . 539 2019;157:107843. 540 7. GBD 2015 Mortality and Causes of Death Collaborators. Global, regional, and 541 national life expectancy, all -cause mortality, and cause -specific mortality for 249 542 causes of death, 1980 -2015: a systematic analysis for the Global Burden of Disease 543 Study 2015 [published correction appears in Lancet. 2017 Jan 544 7;389(10064):e1]. Lancet. 2016;388(10053):1459-1544. 545 8. Buddeke J, Bots ML, van Dis I, et al. Comorbidity in patients with cardiovascular 546 disease in primary care: a cohort study with routine health care data. Br J Gen Pract . 547 2019;69(683):e398-e406. 548 9. Kendir C, van den Akker M, Vos R, Metsemakers J. Cardiovascular disease patients 549 have increased risk for comorbidity: A cross -sectional study in the Netherlands. Eur J 550 Gen Pract. 2018;24(1):45-50. 551 10. Kulie T, Slattengren A, Redmer J, Counts H, Eglash A, Schrager S. Obesity and 552 women's health: an evidence-based review. J Am Board Fam Med. 2011;24(1):75-85. 553 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 24 11. Fox R, Kitt J, Leeson P, Aye CYL, Lewandowski AJ. Preeclampsia: Risk Factors, 554 Diagnosis, Management, and t he Cardiovascular Impact on the Offspring. J Clin Med. 555 2019;8(10):1625. Published 2019 Oct 4. 556 12. Say L, Chou D, Gemmill A, et al. Global causes of maternal death: a WHO 557 systematic analysis. Lancet Glob Health. 2014;2(6):e323-e333. 558 13. Baharuddin M, Amelia D, Suh owatsky S, Kusuma A, Suhargono MH, Eng B. 559 Maternal death reviews: A retrospective case series of 90 hospital -based maternal 560 deaths in 11 hospitals in Indonesia. Int J Gynaecol Obstet. 2019;144 Suppl 1:59-64. 561 14. Gerdts E, Regitz -Zagrosek V. Sex differences in cardiometabolic disorders. Nat Med. 562 2019;25(11):1657-1666. 563 15. Garovic VD, August P. Preeclampsia and the future risk of hypertension: the pregnant 564 evidence. Curr Hypertens Rep. 2013;15(2):114-121. 565 16. Tooher J, Thornton C, Makris A, Ogle R, Korda A, Hennessy A. All Hypertensive 566 Disorders of Pregnancy Increase the Risk of Future Cardiovascular Disease. 567 Hypertension. 2017 Oct;70(4):798-803. 568 17. Gambineri A, Pelusi C, Vicennati V, Pagotto U, Pasquali R. Obesity and the 569 polycystic ovary syndrome. Int J Obes Relat Metab Disord. 2002;26(7):883-896. 570 18. Kwak SH, Jang HC, Park KS. Finding genetic risk factors of gestational 571 diabetes. Genomics Inform. 2012;10(4):239-243. 572 19. Gambineri A, Patton L, Altieri P, et al. Polycystic ovary syndrome is a risk factor for 573 type 2 diabetes: results from a long -term prospective study. Diabetes. 574 2012;61(9):2369-2374. doi:10.2337/db11 -1360 575 20. Mu F, Rich -Edwards J, Rimm EB, Spiegelman D, Missmer SA. Endometriosis and 576 Risk of Coronary Heart Disease. Circ Cardiovasc Qual Outcomes . 2016;9(3) :257-577 264. 578 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 25 21. Dawson A, Fernandez ML, Anglesio M, Yong PJ, Carey MS. Endometriosis and 579 endometriosis-associated cancers: new insights into the molecular mechanisms of 580 ovarian cancer development. Ecancermedicalscience. 2018;12:803. 581 22. Paul A, Paul S. The breast cancer susceptibility genes (BRCA) in breast and ovarian 582 cancers. Front Biosci (Landmark Ed). 2014;19:605-618. 583 23. Nilsson E, Salonen Ros H, Cnattingius S, Lichtenstein P. The importance of genetic 584 and environmental effects for pre -eclampsia and gestati onal hypertension: a family 585 study. BJOG. 2004;111(3):200-206. 586 24. Vink JM, Sadrzadeh S, Lambalk CB, Boomsma DI. Heritability of polycystic ovary 587 syndrome in a Dutch twin -family study. J Clin Endocrinol Metab . 2006;91(6):2100-588 2104. doi:10.1210/jc.2005-1494 589 25. Vujkovic M, Keaton JM, Lynch JA, et al. Discovery of 318 new risk loci for type 2 590 diabetes and related vascular outcomes among 1.4 million participants in a multi -591 ancestry meta-analysis. Nat Genet. 2020;52(7):680-691. 592 26. Justice AE, Winkler TW, Feitosa MF, et al. Genome-wide meta-analysis of 241,258 593 adults accounting for smoking behaviour identifies novel loci for obesity traits. Nat 594 Commun. 2017;8:14977. 595 27. Lewis CM, Vassos E. Polygenic risk scores: from research tools to clinical 596 instruments. Genome Med. 2020;12(1):44. Published 2020 May 18. 597 28. Khera AV, Chaffin M, Aragam KG, et al. Genome -wide polygenic scores for 598 common diseases identify individuals with risk equivalent to monogenic 599 mutations. Nat Genet. 2018;50(9):1219-1224. 600 29. Steinthorsdottir V, McGinnis R, Williams NO, et al. Genetic predisposition to 601 hypertension is associated with preeclampsia in European and Central Asian women. 602 Nat Commun. 2020;11(1):5976. 603 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 26 30. Verma A, Ritchie MD. Current Scope and Challenges in Phenome -Wide Association 604 Studies. Curr Epidemiol Rep. 2017;4(4):321-329. 605 31. Fritsche LG, Gruber SB, Wu Z, et al. Association of Polygenic Risk Scores for 606 Multiple Cancers in a Phenome -wide Study: Results from The Michigan Genomics 607 Initiative. Am J Hum Genet. 2018;102(6):1048-1061. 608 32. Shen X, Howard D M, Adams MJ, et al. A phenome -wide association and Mendelian 609 Randomisation study of polygenic risk for depression in UK Biobank. Nat Commun . 610 2020;11(1):2301. 611 33. Chang CC, Chow CC, Tellier LC, Vattikuti S, Purcell SM, Lee JJ. Second -generation 612 PLINK: rising to the challenge of larger and richer datasets. Gigascience. 2015;4:7. 613 34. Bulik-Sullivan BK, Loh PR, Finucane HK, et al. LD Score regression distinguishes 614 confounding from polygenicity in genome -wide association studies. Nat Genet . 615 2015;47(3):291-295. 616 35. Martin A R, Kanai M, Kamatani Y, Okada Y, Neale BM, Daly MJ. Clinical use of 617 current polygenic risk scores may exacerbate health disparities. Nat Genet . 618 2019;51(4):584-591. 619 36. Duncan L, Shen H, Gelaye B, et al. Analysis of polygenic risk score usage and 620 performance in diverse human populations. Nat Commun. 2019;10(1):3328. 621 37. Viechtbauer, W. Conducting Meta -Analyses in R with the metafor Package. Journal 622 of Statistical Software. 2010;36(3):1-48. 623 38. Ge T, Che n CY, Ni Y, Feng YA, Smoller JW. Polygenic prediction via Bayesian 624 regression and continuous shrinkage priors. Nat Commun. 2019;10(1):1776. 625 39. Pendergrass SA, Dudek SM, Crawford DC, Ritchie MD. Visually integrating and 626 exploring high throughput Phenome-Wide Association Study (PheWAS) results using 627 PheWAS-View. BioData Min. 2012;5(1):5. 628 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 27 40. Hemani G, Zheng J, Elsworth B, et al. The MR -Base platform supports systematic 629 causal inference across the human phenome. Elife. 2018;7:e34408. 630 41. Onstad MA, Schmandt RE, Lu KH. A ddressing the Role of Obesity in Endometrial 631 Cancer Risk, Prevention, and Treatment. J Clin Oncol. 2016;34(35):4225-4230. 632 42. Mehta LS, Watson KE, Barac A, et al. Cardiovascular Disease and Breast Cancer: 633 Where These Entities Intersect: A Scientific Statement From the American Heart 634 Association [published correction appears in Circulation. 2019 Aug 635 27;140(9):e543]. Circulation. 2018;137(8):e30-e66. 636 43. Huang K, Zhong J, Li Q, et al. Effects of CDKN2B -AS1 polymorphisms on the 637 susceptibility to coronary heart disease. Mol Genet Genomic Med. 2019;7(11):e955. 638 44. Qin S, Ning M, Liu Q, Ding X, Wang Y, Liu Q. Knockdown of long non -coding 639 RNA CDKN2B -AS1 suppresses the pr ogression of breast cancer by miR -122-640 5p/STK39 axis. Bioengineered. 2021;12(1):5125-5137. 641 45. Kojima Y, Downing K, Kundu R, et al. Cyclin -dependent kinase inhibitor 2B 642 regulates efferocytosis and atherosclerosis. J Clin Invest. 2014;124(3):1083-1097. 643 46. Motterle A, Pu X, Wood H, et al. Functional analyses of coronary artery disease 644 associated variation on chromosome 9p21 in vascular smooth muscle cells. Hum Mol 645 Genet. 2012;21(18):4021-4029. 646 47. Clark SL, Parham M, Shadyab AH, et al. Broad Clinical Manifestations of Po lygenic 647 Risk for Coronary Artery Disease in the Women’s Health Initiative. medRxiv 648 2021.06.15.21258993 649 48. Kanegae H, Oikawa T, Okawara Y, Hoshide S, Kario K. Which blood pressure 650 measurement, systolic or diastolic, better predicts future hypertension in normotensive 651 young adults?. J Clin Hypertens (Greenwich). 2017;19(6):603-610. 652 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 28 49. Sesso HD, Stampfer MJ, Rosner B, et al. Systolic and diastolic blood pressure, pulse 653 pressure, and mean arterial pressure as predictors of cardiovascular disease risk in 654 Men. Hypertension. 2000;36(5):801-807. 655 50. Kachuri L, Graff RE, Smith -Byrne K, et al. Pan -cancer analysis demonstrates that 656 integrating polygenic risk scores with modifiable risk factors improves risk 657 prediction. Nat Commun. 2020;11(1):6084. 658 Figure Legends: 659 Figure 1: Genetic correlation and the influence of shared genetic burden of 660 cardiometabolic traits and health conditions unique to females. Panel A shows a heatmap 661 of genetic correlation from cross -trait LDSC analyses : Blue represent s negative correlation 662 and red re presents positive correlation. An * in each box suggests statistical significance of 663

based on p -values. Panels B, C and D refer to PRS based meta -analyses between 664 cardiometabolic PRS and case/controls status of female -specific diseases in overall, EUR and 665 AFR ancestry individuals respectively. The first panel in these plots corresponds to p -values 666 and second panel represents beta estimates. The color of each point refers to the PRS for 667 cardiometabolic traits. 668 669 Figure 2: Inverse relationship between Coronary Artery Disease (CAD) and Breast 670 Cancer. Panel A shows the heatmap of negative genetic correlation between CAD and breast 671 cancer from the UKBB Genetic Correlation Browser dataset. The gradient of color refers to 672 positive (red) to negative (blue) correlation, and the text in each box refers to the genetic 673 correlation coefficient. Panel B shows the distribution of breast cancer per each PRSCAD 674 quintile. The x-axis represents each PRS quintile, and the y-axis is the disease prevalence. 675 The color of each point refers to the ancestry group (All, European, and African), and the 676 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 29 shape indicates the target dataset (eMERGE (circle) or PMBB (triangle)). Panel C is a Venn 677 diagram representing the overlap of CAD and breast cancer cases in PMBB and eMERGE 678 datasets based on disease diagnosis. 679 680 Figure 3: Mendelian Randomization (MR) results for 29 significant PRS based 681 associations. This figure shows forest plot s of association between female health conditions 682 as outcomes and cardiometabolic phenotypes as exposures in a one -sample MR analyses 683 shown in Panel A and two -sample MR analyses shown in Panel B. Genetic instruments are 684 the cardiometabolic PRS in the one-sample MR and genome -wide significant SNPs from 685 GWAS in the two -sample MR. In panel A, each point refers to beta outcome/SD exposure for 686 PRSBMI and log(OR) outcome for all other exposure variables for each tests performed as 687 separated by ancestry. In panel B, each point refers to b eta outcome/SD exposure for BMI 688 and beta outcome/SD log(OR) exposure for all other variables across all methods used in 689 sensitivity analyses for two-sample MR. P-values are reported in last column in both panels. 690 691 Figure 4: Chronology analyses for the visualization of events from the EHR. Circular 692 plot showing disease prevalence among high PRSBMI (in yellow) and low PRSBMI (in blue) 693 categories. General female health conditions are shown in panel A and pregnancy and 694 childbirth-related phenotypes are shown in panel B. The circular plots are divided into five 695 age categories (70) for general female health conditions and 696 three age categories (<25, 25-40, and 40-55) for pregnancy-related phenotypes697 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 30 Tables: Table 1: Sample size in PMBB and eMERGE datasets, for overall and stratified by ancestry analyses. Phenotype PMBB Sample Size (N Cases) eMERGE Sample Size (N Cases) All EUR AFR All EUR AFR Cardiometabolic Phenotypes BMI 20,209 12,344 6,744 39,403 31,875 5,250 CAD 21,837 (3,002) 13,515 (1,956) 7,039 (955) 49,171 (9,597) 37,003 (7,524) 8,308 (1,308) DBP 21,612 13,343 6,994 NA NA NA Hypertension 21,837 (10,278) 13,515 (5,640) 7,039 (4,286) 49,171 (27,685) 37,003 (20,820) 8,308 (4,695) T2D 21,837 (4,388) 13,515 (1,969) 7,039 (2,221) 49,171 (12,403) 37,003 (8,311) 8,308 (2,725) Women’s Health Phenotypes Breast cancer 21,837 (1,621) 13,515 (1,621) 7,039 (415) 49,171 (4,148) 37,003 (3,532) 8,308 (408) . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 31 Cervical cancer 21,837 (105 13,515 (53) 7,039 (50) 49,171 (332) 37,003 (252) 8,308 (52) Ectopic pregnancy 2,808 (1,779) 1,201 (827) 1,319 (757) 3,078 (332) 1,975 (521) 641 (153) Endometrial cancer 21,837 (286) 13,515 (183) 7,039 (90) 49,171 (771) 37,003 (653) 8,308 (74) Endometriosis 21,837 (701) 13,515 (320) 7,039 (340) 49,171 (2,314) 37,003 (1,805) 8,308 (330) Excessive fetal growth 693 (85) 293 (37) 329 (42) 2,433 (627) 1,618 (377) 437 (150) Gestational diabetes 2,655 (523) 1,111 (193) 1,262 (248) 3,174 (762) 2,005 (414) 719 (244) Gestational hypertension 2,666 (631) 1,118 (256) 1,275 (324) 3,135 (703) 1,980 (383) 711 (235) Intrauterine death 685 (57) 289 (23) 324 (27) 2,156 (64) 1,438 (38) 358 (17) Miscarriage 2,934 (389) 1,273 (199) 1,360 (151) 3,568 (823) 2,323 (547) 740 (182) Ovarian cancer 21,837 (305) 13,515 (191) 7,039 (89) 49,171 (1,589) 37,003 (1,324) 8,308 (177) . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 32 Placenta abruption/previa 1,192 (396) 482 (157) 586 (195) 2,674 (997) 1,697 (616) 583 (240) Polycystic ovarian syndrome 21,837 (736) 13,515 (387) 7,039 (272) 49,171 (1,006) 37,003 (690) 8,308 (200) Poor fetal growth 792 (202) 325 (78) 388 (107) 2,209 (167) 1,478 (102) 360 (21) Postpartum depression 924 (384) 374 (136) 466 (226) 2,417 (555) 1,579 (319) 457 (157) Postpartum hemorrhage 846 (283) 350 (108) 408 (146) 2,320 (401) 1,544 (250) 408 (104) Preeclampsia 2,631 (452) 1,100 (149) 1,264 (272) 3,132 (702) 1,974 (377) 713 (13) Preterm birth 687 (66) 284 (21) 332 (40) 2,144 (57) 1,433 (30) 350 (13) Stillbirth 649 (17) 270 (3) 311 (12) 2,123 (8) 1,417 (6) 347 (0) Uterine cancer 21,837 (113) 13,515 (66) 7,039 (43) 49,171 (418) 37,003 (340) 8,308 (58) Uterine fibroid 21,837 (1,570) 13,515 (516) 7,039 (984) 49,171 (5,711) 37,003 (3,912) 8,308 (1,347) . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 33 Table 2: GWAS summary statistics datasets that are used for the calculation of genetic correlations and polygenic risk scores and the number of SNPs used from each GWAS to calculate the corresponding PRS. Vaginal cancer 21,837 (23) 13,515 (13) 7,039 (9) 49,171 (109) 37,003 (86) 8,308 (17) Vulvar cancer 21,837 (41) 13,515 (25) 7,039 (14) 49,171 (120) 37,003 (92) 8,308 (19) Phenotype Ancestries Included Source Sample Size (Number of Cases) PMID N SNPs included in PRS Type 2 Diabetes EUR, AFR, EAS, SAS, HIS Vujkovic et al., Nat Gen, 2020, MVP 1,407,282 (228,499) 32541925 PMBB: 1,023,697 eMERGE: 716,330 Body Mass Index EUR, AFR, EAS, SAS, HIS Justice et al., Nat Com, 2017, GIANT 241,258 28443625 PMBB: 885,143 eMERGE: 614,668 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 34 **DBP= Diastolic Blood Pressure; SBP= Systolic Blood Pressure; PP= Pulse Pressure; EUR= European ancestry; EAS= East Asian ancestry; SAS= South Asian ancestry; HIS= Hispanic ancestry; AFR= African ancestry; MVP= Million Veterans Program; GIANT=The Genetic Investigation of ANthropometric Traits; UKB= UK BioBank Hypertension (DBP, SBP, PP) EUR, AFR, EAS, SAS, HIS Giri et al., Nat Gen, 2019, MVP 318,891 30578418 PMBB: 1,024,567 eMERGE: 715,471 Coronary Artery Disease EUR, EAS, SAS, HIS, AFR van der Harst et al., Circ Res, 2018, CARDIoGRAMplusC4D + UKBB 547,261 (122,733) 29212778 PMBB: 981,480 eMERGE: 681,029 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 35 Table 3: Effect estimates for testing association of PRS with its primary phenotype in eMERGE and PMBB datasets Phenotype Ancestry N Total N Cases R2 or AUC of Full Model Beta SE P-value R2 or AUC of Null Model R2 or AUC of PRS only Model eMERGE BMI All 39,403 0.08494 2.378 0.05044 0 0.0334 0.0593 EUR 31,875 0.07208 2.336 0.05144 0 0.0121 0.0583 AFR 5,250 0.07424 2.76 0.25 5.01E-28 0.0529 0.0201 CAD All 49,171 9,597 0.784 0.3091 0.01318 1.62E-121 0.7745 0.5474 EUR 37,003 7,524 0.7722 0.3096 0.01427 2.82E-104 0.7608 0.5614 AFR 8,308 1,308 0.8299 0.308 0.04789 1.24E-10 0.827 0.5442 DBP (Hypertension) All 49,171 27,685 0.8109 0.1387 0.01347 7.17E-25 0.8099 0.529 EUR 37,003 20,820 0.7961 0.1464 0.01487 7.02E-23 0.7946 0.5339 AFR 8,308 4,695 0.8634 0.1503 0.04098 0.000245 0.8628 0.5114 PP (Hypertension) All 49,171 27,685 0.8115 0.1644 0.01344 2.15E-34 0.8099 0.526 EUR 37,003 20,820 0.7969 0.1712 0.01438 1.13E-32 0.7946 0.5273 AFR 8,308 4,695 0.8629 0.09689 0.04871 0.0467 0.8628 0.5049 SBP (Hypertension) All 49,171 27,685 0.8131 0.2977 0.01665 1.86E-71 0.8099 0.5352 EUR 37,003 20,820 0.7992 0.3156 0.01793 2.28E-69 0.7946 0.5428 AFR 8,308 4,695 0.8637 0.2845 0.06227 4.91E-06 0.8628 0.5116 T2D All 49,171 12,403 0.7193 0.6334 0.01914 3.02E-240 0.691 0.6163 EUR 37,003 8,311 0.6869 0.6311 0.0205 3.56E-208 0.6414 0.6033 AFR 8,308 2,725 0.7795 0.6318 0.06546 4.83E-22 0.7728 0.5597 PMBB . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 36 BMI All 20,209 0.1542 2.658 0.08244 2.05E-222 0.1108 0.1417 EUR 12,344 0.07473 2.621 0.08679 2.66E-193 0.006422 0.06782 AFR 6,744 0.03203 2.708 0.2089 5.75E-38 0.008037 0.02711 CAD All 21,837 3,002 0.8235 0.3571 0.02354 5.73E-52 0.8148 0.5619 EUR 13,515 1,956 0.8303 0.3894 0.02732 4.44E-46 0.8185 0.5929 AFR 7,039 955 0.799 0.2566 0.05214 8.60E-07 0.7955 0.5389 DBP All 21,612 0.0425 1.036 0.06661 3.05E-54 0.03183 0.02905 EUR 13,343 0.01534 0.9964 0.08064 7.01E-35 0.004144 0.01213 AFR 6,994 0.04559 1.104 0.129 1.35E-17 0.03571 0.01375 PP (Hypertension) All 21,837 10,278 0.8125 0.1942 0.02328 7.18E-17 0.8112 0.602 EUR 13,515 5,640 0.7797 0.1918 0.02638 3.53E-13 0.7778 0.5272 AFR 7,039 4,286 0.8389 0.1886 0.05376 0.00045 0.8383 0.5391 SBP (Hypertension) All 21,837 10,278 0.8144 0.327 0.02523 2.08E-38 0.8112 0.6097 EUR 13,515 5,640 0.7813 0.279 0.02846 1.11E-22 0.7778 0.5397 AFR 7,039 4,286 0.8415 0.4505 0.05923 2.81E-14 0.8383 0.5649 T2D All 21,837 4,388 0.754 0.8136 0.03523 5.49E-118 0.7287 0.6713 EUR 13,515 1,969 0.712 0.8418 0.04533 5.55E-77 0.6634 0.6323 AFR 7,039 2,221 0.7342 0.7896 0.05947 3.11E-40 0.7168 0.5919 **BMI= Body Mass Index; CAD= Coronoray artery disease, T2D=Type 2 Diabetes; DBP= Diastolic Blood Pressure; SBP= Systolic Blood Pressure; PP= Pulse Pressure ; EUR= European ancestry; AFR= African ancestry; Beta coefficients and standard errors are per SD PRS in the full model. Covariates included: birth year and the first five PCs. . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 37 Table 4: FDR-significant (FDR P-value < 0.05) associations between cardiometabolic PRS and female health conditions identified in the PMBB and eMERGE meta-analysis PRS Association Ancestry Beta SE OR 95% CI P-value BMI Breast cancer All -0.071 0.0225 0.93 0.891-0.973 0.00159 Endometrial cancer All 0.244 0.0458 1.28 1.17-1.4 9.4x10-8 EUR 0.276 0.087 1.32 1.11-1.56 0.00152 Gestational diabetes All 0.23 0.0508 1.26 1.14-1.39 6x10-6 EUR 0.277 0.0622 1.32 1.17-1.49 8.69x10-6 PCOS All 0.272 0.0387 1.31 1.22-1.42 2.37x10-12 EUR 0.294 0.0451 1.34 1.23-1.47 6.76x10-11 CAD Breast cancer All -0.0718 0.0147 0.931 0.904-0.958 9.96x10-7 EUR -0.0699 0.0168 0.932 0.902-0.964 3.11x10-5 Postpartum depression EUR 0.178 0.0577 1.19 1.07-1.34 0.00201 PP Gestational hypertension All 0.218 0.0443 1.24 1.14-1.36 8.51x10-7 EUR 0.235 0.0558 1.26 1.13-1.41 2.54x10-5 Preeclampsia All 0.196 0.0581 1.22 1.09-1.36 0.000762 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 38 SBP Gestational hypertension All 0.332 0.0825 1.39 1.19-1.64 5.61x10-5 EUR 0.306 0.0661 1.36 1.19-1.55 3.61x10-6 AFR 0.398 0.099 1.49 1.23-1.81 5.7x10-5 SBP Preeclampsia All 0.275 0.0799 1.32 1.13-1.54 0.000567 EUR 0.229 0.0712 1.26 1.09-1.45 0.0013 AFR 0.358 0.102 1.43 1.17-1.75 0.000517 T2D Breast cancer All -0.0726 0.0213 0.93 0.892-0.97 0.000657 EUR -0.0787 0.0263 0.924 0.878-0.973 0.00272 Gestational diabetes All 0.587 0.063 1.8 1.59-2.03 1.19x10-20 EUR 0.678 0.0759 1.97 1.7-2.29 3.88x10-19 Gestational hypertension All 0.184 0.0642 1.2 1.06-1.36 0.00414 EUR 0.178 0.0671 1.19 1.05-1.36 0.00798 PCOS All 0.223 0.0428 1.25 1.15-1.36 1.93x10-7 EUR 0.209 0.0485 1.23 1.12-1.36 1.56x10-5 AFR 0.32 0.104 1.38 1.12-1.69 0.00206 Postpartum depression All 0.18 0.07 1.2 1.04-1.37 0.0101 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 39 **BMI= Body Mass Index; CAD= Coronoray artery disease, T2D=Type 2 Diabetes; PCOS=Polycystic ovarian syndrome; EUR= European ancestry; P= PMBB; E= eMERGE; Beta coefficients are per SD PRS. Results were adjusted for birth year and the first five PCs. Figures: . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 40 Figure 1: . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 41 Figure 2: A. B. C. . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 42 Figure 3: . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint 43 Figure 4: . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted February 4, 2022. ; https://doi.org/10.1101/2022.02.02.22269844doi: medRxiv preprint