Abstract

For complex disorders, estimating the age-related penetrance associated with rare variants of strong effect is essential before a putative use for genetic counseling or disease prevention. However, rarity and co-occurrence with other risk factors make such estimations difficult. In the context of Alzheimer disease, we present a survival model to estimate the penetrance of SORL1 rare (allele frequency< 1%) Loss-of-Function variants (LoF) while accounting for APOE-{varepsilon}4, the main risk factor (allele frequency[~] 14% in Caucasians). We developed an efficient strategy to compute penetrance estimates accounting for both common and rare genetic variants based on available penetrance curves associated with common risk factors and using incomplete pedigree data to quantify the additional risk conferred by rare variants. Our model combines: (i) a baseline for non-carriers of SORL1 LoF variants, stratified by APOE genotypes derived from the Rotterdam study and (ii) an age-dependent proportional hazard effect for SORL1 LoF variants estimated from pedigrees with a proband carrying such a variant. We embed this model into an Expectation-Maximisation algorithm to accommodate for missing genotypes. Confidence intervals were computed by bootstraps. To correct for ascertainment bias, proband phenotypes were omitted. We obtained penetrance curves associated with SORL1 LoF variants at the digenic level. By age 70, we estimate a 100% penetrance of SORL1 LoF variants only among APOE-{varepsilon}4{varepsilon}4 carriers, while penetrance is 56%[40% - 72%] among {varepsilon}4 heterozygous carriers and 37%[26% - 51%] among {varepsilon}4 non-carriers. We conclude that rare SORL1 LoF variants should not be used for genetic counseling regardless of the APOE status.