111 Risk of dementia related to APOE ε4 and a polygenic risk score were prepared for imputation using provided scripts (HRC-1000G-check-bim.pl, https://www.well.ox.ac.uk/~wrayner/tools/) to compare variant ID, strand and allele frequencies to the Haplotype Reference Panel (HRC v1.1, April 2016)(39). Finally, all autosomal variants were submitted to the Sanger imputation server (https:// imputation.sanger.ac.uk/). The server uses MACH3 to phase data and imputation to the reference panel was performed with Positional Burrows-Wheeler Transform (PBWT) (40). APOE was determined based on imputation. We calculated a weighted individual PRS based on the 39 genetic variants that showed genome-wide significant evidence of association with AD (Supplementary Table 1). APOE was analyzed separately and was not included in the PRS. The selected variants were directly genotyped (median genotyping rate = 1) or imputed with high quality (median imputation score R² = 0.98). The PRSs were generated by multiplying the genotype dosage of each risk allele for each variant by its respective weight and then summing across all variants (41). We weighted this by the effect size from previous International Genomics of Alzheimer’s project (IGAP) studies (Supplementary Table 1) (8, 9). The PRS was normalized (mean = 0, standard deviation = 1). Statistics We performed logistic regression analysis to investigate the relationship between APOE ε4 or the PRS and biomarkers in the ATN classification. The first set of models included APOE ε4 (per allele) or PRS (z-score) as predictor, and A status, T status or N status as outcome (separate models). In addition, we performed multinomial logistic regression analyses with first the eight-profile ATN classification as outcome, and second the three clustered ATN categories as outcome (A-T-N- reference for both analyses). We created a second set of logistic regression models which included both APOE ε4 and PRS as predictors in the same model. Next, we performed Cox proportional hazards analyses to assess the relationship between APOE ε4 or PRS and the risk of dementia. We studied progression to all-type dementia, AD dementia and non-AD dementia as outcome separately. In the analyses with AD dementia as outcome, the individuals that progressed to other types of dementia were censored, and vice versa. We created a second set of Cox regression models which included both APOE ε4 and PRS as predictors in the same model. We additionally investigated the combined predictive value of APOE ɛ4 and PRS. We first stratified the PRS into low and high (based on the median) and then constructed a four-level variable: APOEε4-PRSlow (reference), APOEε4-PRShigh, APOEε4+PRSlow, APOEε4+PRShigh. We ran an additional analysis with this four-level variable as predictor. 5
RkJQdWJsaXNoZXIy MjY0ODMw