Mild-cognitive impairment (MCI) occurs in up to one-fifth of individuals over the age of 65, with approximately a third of MCI individuals converting to dementia in later life. There is a growing necessity for early identification for those at risk of dementia as pathological processes begin decades before onset of symptoms. A cohort of 122 individuals diagnosed with MCI and followed up for a 36-month period for conversion to late-onset Alzheimer’s disease (LOAD) were genotyped on the NeuroChip array along with pathologically confirmed cases of LOAD and cognitively normal controls. Polygenic risk scores (PRS) for each individual were generated using PRSice-2, derived from summary statistics produced from the International Genomics of Alzheimer’s Disease Project (IGAP) genome-wide association study. Predictability models for LOAD were developed incorporating the PRS with APOE SNPs (rs7412 and rs429358), age and gender. This model was subsequently applied to the MCI cohort to determine whether it could be used to predict conversion from MCI to LOAD. The PRS model for LOAD using area under the precision-recall curve (AUPRC) calculated a predictability for LOAD of 82.5%. When applied to the MCI cohort predictability for conversion from MCI to LOAD was 61.0%. Increases in average PRS scores across diagnosis group were observed with one-way ANOVA suggesting significant differences in PRS between the groups ( p < 0.0001). This analysis suggests that the PRS model for LOAD can be used to identify individuals with MCI at risk of conversion to LOAD.

Sporadic early-onset Alzheimer’s disease (sEOAD) exhibits the symptoms of late-onset Alzheimer’s disease but lacks the familial aspect of the early-onset familial form. The genetics of Alzheimer’s disease (AD) identifies APOE ε4 to be the greatest risk factor; however, it is a complex disease involving both environmental risk factors and multiple genetic loci. Polygenic risk scores (PRSs) accumulate the total risk of a phenotype in an individual based on variants present in their genome. We determined whether sEOAD cases had a higher PRS compared to controls. A cohort of sEOAD cases was genotyped on the NeuroX array, and PRSs were generated using PRSice. The target data set consisted of 408 sEOAD cases and 436 controls. The base data set was collated by the International Genomics of Alzheimer’s Project consortium, with association data from 17,008 late-onset Alzheimer’s disease cases and 37,154 controls, which can be used for identifying sEOAD cases due to having shared phenotype. PRSs were generated using all common single nucleotide polymorphisms between the base and target data set, PRS were also generated using only single nucleotide polymorphisms within a 500 kb region surrounding the APOE gene. Sex and number of APOE ε2 or ε4 alleles were used as variables for logistic regression and combined with PRS. The results show that PRS is higher on average in sEOAD cases than controls, although there is still overlap among the whole cohort. Predictive ability of identifying cases and controls using PRSice was calculated with 72.9% accuracy, greater than the APOE locus alone (65.2%). Predictive ability was further improved with logistic regression, identifying cases and controls with 75.5% accuracy.

In addition, polygenic risk scores (PRS) were able to distinguish between cases and controls with 83.8% accuracy using 3268 variants, sex, age at death and APOE ε4 and ε2 status as predictors.

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

Synapse loss is an early event in late-onset Alzheimer's disease (LOAD). In this study we have assessed the capacity of a polygenic risk score (PRS) restricted to synapse-encoding loci to predict LOAD. We used summary statistics from the IGAP genome-wide association meta-analysis of 74,046 subjects for model construction and tested the "Synaptic PRS" in two independent datasets of controls and pathologically-confirmed LOAD. The mean Synaptic PRS was 2.3-fold higher in LOAD compared to controls (p<0.0001) with a predictive accuracy of 72% in the target dataset (n=439) and 73% in the validation dataset (n=136), a 5-6% improvement compared to the APOE locus (p<0.00001). The model comprises 8 variants from 4 previously-identified (BIN1, PTK2B, PICALM, APOE) and 2 novel (DLG2, MINK1) LOAD loci involved in glutamate signaling (p=0.01) or APP catabolism/tau binding (p=0.005). As the simplest PRS model with good predictive accuracy to predict LOAD, we conclude that synapse-encoding genes are enriched for LOAD risk-modifying loci.The Synaptic PRS could be used to identify individuals at risk of LOAD before symptom onset.