Xicota Laura, Cheng Rong, Barral Sandra, Honig Lawrence S, Schupf Nicole, Gu Yian, Andersen Stacy, Cosentino Stephanie, Zmuda Joseph, Perls Thomas, Province Michael, Lee Joseph H
Department of Neurology, Columbia University Irving Medical Center, 630W 168th street, New York City, New York, 10032, USA.
Gertrude H. Sergievsky Center, Columbia University Irving Medical Center, 630W 168th street, New York City, New York, 10032, USA.
medRxiv. 2025 Jul 21:2025.07.21.25331660. doi: 10.1101/2025.07.21.25331660.
Polygenic risk scores (PRS) have been used to assess an individual's risk for various diseases, including Alzheimer's disease (AD). This study applied the PRS approach to a cohort of families ascertained for healthy aging, that have shown a reduced risk of AD. Using the SNPs identified as significantly associated with AD in the study by Kunkle and colleagues, we examined the utility of PRS for predicting AD risk in a cohort ascertained for familial healthy aging.
We restricted the study to US LLFS study participants who have been evaluated for AD and have available whole genome sequencing (WGS) data. AD diagnosis was based on consensus diagnosis, and for those without consensus diagnosis, we used the algorithm based on standardized memory scores. PRS were calculated using a published weighted formula. To further examine the predictability of PRS, we assessed the relationship between PRS and AD biomarkers, including Aβ, Aβ, NfL, and GFAP. Mixed effects models were used to adjust for confounders as well as relatedness among family members. Given the age at onset of common late onset AD, the present study included those who were at least 65 years of age.
We observed that PRS had limited predictive power for AD in this healthy aging cohort. Yet, allele frequencies for the SNPs used in PRS estimation differed between the two studies in a small number (9.7%) of SNPs, suggesting that the lack of effect of the PRS is likely to be due to the small number of AD associated SNPs (12.3% and 16.1%). Subsequent analysis observed no significant association between PRS and biomarkers. This was explained by the low number of SNPs significantly associated with each of the biomarkers.
This study highlights the importance of ascertainment of study population in interpreting PRS. In LLFS, a population at reduced risk of AD, PRS based on genetic variants identified from the general population may be inadequate to explain the variability in AD risk. Our results suggest that genetic risk variants, the basis of PRS, may need to be adjusted according to the study population of interest.
多基因风险评分（PRS）已被用于评估个体患包括阿尔茨海默病（AD）在内的各种疾病的风险。本研究将PRS方法应用于一组因健康衰老而被确定的家庭队列，这些家庭显示患AD的风险较低。利用在Kunkle及其同事的研究中确定与AD显著相关的单核苷酸多态性（SNP），我们在一个因家族性健康衰老而被确定的队列中检验了PRS预测AD风险的效用。
我们将研究限制在美国纵向衰老研究（LLFS）的参与者中，这些参与者已接受AD评估并拥有全基因组测序（WGS）数据。AD诊断基于共识诊断；对于没有共识诊断的参与者，我们使用基于标准化记忆评分的算法。使用已发表的加权公式计算PRS。为了进一步检验PRS的可预测性，我们评估了PRS与AD生物标志物之间的关系，包括淀粉样β蛋白（Aβ）、Aβ、神经丝轻链（NfL）和胶质纤维酸性蛋白（GFAP）。使用混合效应模型来调整混杂因素以及家庭成员之间的相关性。鉴于常见晚发性AD的发病年龄，本研究纳入了至少65岁的参与者。
我们观察到，在这个健康衰老队列中，PRS对AD的预测能力有限。然而，用于估计PRS的SNP的等位基因频率在两项研究中存在少量差异（9.7%），这表明PRS缺乏效应可能是由于与AD相关的SNP数量较少（分别为12.3%和16.1%）。后续分析未观察到PRS与生物标志物之间存在显著关联。这可以通过与每种生物标志物显著相关的SNP数量较少来解释。
本研究强调了在解释PRS时确定研究人群的重要性。在LLFS中，一个患AD风险较低的人群中，基于从一般人群中鉴定出的遗传变异的PRS可能不足以解释AD风险的变异性。我们的结果表明，作为PRS基础的遗传风险变异可能需要根据感兴趣的研究人群进行调整。
