Comparing logistic regression, support vector machines, and permanental classification methods in predicting hypertension

Abstract

In this paper, we compare logistic regression and 2 other classification methods in predicting hypertension given the genotype information. We use logistic regression analysis in the first step to detect significant single-nucleotide polymorphisms (SNPs). In the second step, we use the significant SNPs with logistic regression, support vector machines (SVMs), and a newly developed permanental classification method for prediction purposes. We also detect rare variants and investigate their impact on prediction. Our results show that SVMs and permanental classification both outperform logistic regression, and they are comparable in predicting hypertension status.

Comparing machine learning and logistic regression methods for predicting hypertension using a combination of gene expression and next-generation sequencing data

“…Following previous work on this data set [3], we use different combinations of the first 150 simulated data sets (SIMPHEN.1 to SIMPHEN.150) to select genes of interest and used 3 other, arbitrarily chosen, simulated data sets (SIMPHEN.197 to SIMPHEN.199) from the remaining simulated sets, as classification data sets. We now provide details of the selection and classification steps.…”

“…The model is an extension of a recently proposed model [3] but using (a) variant collapsing across the gene and (b) adding main effect and interaction terms for gene expression data.where Y  = 1 indicates that an individual is hypertensive, Sex and Smoke are indicator variables for the respondent’s Sex and Smoking status, respectively, and Age is a continuous measure of the respondent’s age. Pedigree information was incorporated into the model via the use of an indicator variable for each distinct pedigree as has been done previously with this data set [3].…”

“…The model is an extension of a recently proposed model [3] but using (a) variant collapsing across the gene and (b) adding main effect and interaction terms for gene expression data.where Y  = 1 indicates that an individual is hypertensive, Sex and Smoke are indicator variables for the respondent’s Sex and Smoking status, respectively, and Age is a continuous measure of the respondent’s age. Pedigree information was incorporated into the model via the use of an indicator variable for each distinct pedigree as has been done previously with this data set [3]. G i is a continuous measure of gene expression at the gene of interest, i , and S i is an indicator of the presence of any rare (minor allele frequency <5 %) alleles at any location within the same gene of interest, i ; collapsing is done in the spirit of combined multivariate and collapsing [5].…”

“…We extend a recently proposed supervised machine learning approach [3] in order to further understand the behavior and performance of MLMs on sequence data. We incorporated a recent statistical model proposed for the joint analysis of gene expression data and genotype data in evaluating disease risk [4], along with explicit consideration of the analysis of rare variants using a collapsing (burden) approach [5].…”

Predicting hypertension without measurement: A non-invasive, questionnaire-based approach

“…Kesselmeier et al (2014) used data from the Genetic Analysis Workshop 18 to evaluate the performance of the standard logistic regression methods, and found their strong dependence on a few observations that deviated from the majority of the data. Huang, Xu, and Yang (2014) developed a two-stage hypertension prediction approach using the genotype information. They first detected significant single-nucleotide polymorphisms (SNPs) and then developed a permanental classifier for prediction purposes.…”