Machine learning for selection of SNPs related to the diagnosis of Alzheimer's Disease

Abstract

Single nucleotide polymorphisms (SNPs) are variations at unique positions in the nucleotide sequence where the DNA is created. Due to the genetic alteration, SNPs are extremely important for human health studies. By them, it is possible to predict the individual response to some particular medications, for example. Furthermore, SNPs can be associated with more complex diseases, like cardiovascular diseases, diabetes, cancer, and Alzheimer's Disease. This kind of association can be done using some machine learning methods like the multiple variables regressions, for example. Thus, this research project intends to investigate the relation between SNPs and Alzheimer's Disease in people who are more than 65 years old (late-onset Alzheimer's Disease, ou LOAD) using machine learning. To do so, some data sets will be created from the intersection of Alzheimer's Disease Neuroimaging Initiative (ADNI) information and genomics bank piece of information from the National Center for Biotechnology Information (NCBI) and the ALZGENE. Anonymous individuals' SNPs and health diagnostic (normal, mild cognitive impairment, Alzheimer's Disease) will be used to create these data sets. For the purpose of researching which SNPs are most related to Alzheimer's Disease, we intend to use regularization strategies from the Ridge and Lasso regression algorithms to select the most important model variables. (AU)