Abstract
AbstractBackgroundGenome wide association study (GWAS) presented that a cohort having one or two copies of APOE4 show about 2.84 or 8.07 times, respectively, higher risk of having Alzheimer’s disease (AD) than who do not. However, it is unclear how much impact does APOE4 have in progressing AD for an individual. We implemented a deep learning model to identify the APOE4’s risk impact at the individual level.MethodWe applied a novel deep learning model to the 266,161 SNP chromosome 19 data of 313 AD and 457 cognitively unimpaired (CU) ADNI patients. The whole dataset was divided into train, validation, and test set with a ratio of 60%, 20%, and 20% for model development. In the test set, 63.93% of AD participants were APOE4 carriers as opposed to 30.11% of CU participants. Therefore, APOE4 alone has a 67.5% diagnostic accuracy.ResultThe model achieved 68.18% in classifying CU vs. AD and recognized SNPs in APOC1 (rs12721051, rs56131196, rs12721046, and rs4420638) as strong AD risk factors. On the other hand, SNPs in APOE did not show a risk impact such that removal or replacement of the SNPs of the APOE haplotype (rs429358 and rs7412) did neither increase nor decrease the likelihood of AD for any of the 154 test set participants. Furthermore, rs429358 and rs7412 did not show any interaction effects with other SNPs in chromosome 19. Further closer examination of these results confirmed that the APOE haplotype (rs429358 and rs7412) and the APOC1 SNPs (rs12721051, rs56131196, rs12721046, and rs4420638) are in strong linkage disequilibrium (normalized correlation coefficients range 0.685‐0.844) which explains why only one of the two genes was selected by our model. These results were reproduced by 5‐fold cross validation as well as with the implementation of a different deep learning model architecture.ConclusionAPOC1 variants were ranked higher and selected by our deep learning algorithm over the APOE haplotype that has been established as the strongest sporadic AD risk factor to date. APOC1 might be a major AD risk gene that needs further exploration.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have