A Machine Learning Method to Identify Genetic Variants Potentially Associated With Alzheimer's Disease.

Bradley Monk,Aleks Rajkovic,Andrei Rajkovic,Terry Gaasterland,Roberto Malinow,Semar Petrus

doi:10.3389/fgene.2021.647436

Abstract

There is hope that genomic information will assist prediction, treatment, and understanding of Alzheimer’s disease (AD). Here, using exome data from ∼10,000 individuals, we explore machine learning neural network (NN) methods to estimate the impact of SNPs (i.e., genetic variants) on AD risk. We develop an NN-based method (netSNP) that identifies hundreds of novel potentially protective or at-risk AD-associated SNPs (along with an effect measure); the majority with frequency under 0.01. For case individuals, the number of “protective” (or “at-risk”) netSNP-identified SNPs in their genome correlates positively (or inversely) with their age of AD diagnosis and inversely (or positively) with autopsy neuropathology. The effect measure increases correlations. Simulations suggest our results are not due to genetic linkage, overfitting, or bias introduced by netSNP. These findings suggest that netSNP can identify SNPs associated with AD pathophysiology that may assist with the diagnosis and mechanistic understanding of the disease.

Highlights

Alzheimer’s disease (AD), the most common form of dementia, is heritable [58–79%, estimated from twin studies (Gatz et al, 2006)], and highly polygenic (Cauwenberghe et al, 2015)
A large variant call format (VCF) datafile [∼200 GB; Alzheimer’s Disease Sequencing Project, ADSP (Beecham et al, 2017)] containing single nucleotide polymorphism (SNPs) information on ∼11,000 individuals over the age of 60 (Northern European descent; ∼6,000 diagnosed with AD, and ∼5,000 aged non-AD controls), was organized into a more manageable file (∼2 GB; N.B.: a VCF datafile contains mainly zeros – indicating reference alleles – since >95% of minor allele frequencies are
The minor frequency allele (MFA) and reference allele count were determined at each SNP locus, separately for case and control groups

Summary

Introduction

Alzheimer’s disease (AD), the most common form of dementia, is heritable [58–79%, estimated from twin studies (Gatz et al, 2006)], and highly polygenic (Cauwenberghe et al, 2015). Mutations in three genes (APP, PS1, PS2) cause rare forms of the disease [accounting for ∼1% of AD cases (Mendez, 2017)], which shows autosomal dominant transmission with high penetrance and displays an early onset [generally before age 60 (Carmona et al, 2018)]. In the more common form of the disease, late onset AD (LOAD), APOE has been established unequivocally as a susceptibility gene (Saunders et al, 1993) with several dozen other genetic loci receiving genetic support (Carmona et al, 2018; Jansen et al, 2019; Kunkle et al, 2019). The neuropathological progression of disease has been best described using the Braak staging scheme (I–VI) (Braak et al, 2006).

Objectives

Methods

Results

Conclusion