GASVeM: A New Machine Learning Methodology for Multi-SNP Analysis of GWAS Data Based on Genetic Algorithms and Support Vector Machines

Fidel Díez Díaz,Víctor Moreno,Fernando Sánchez Lasheras,Ferran Moratalla-Navarro,Vicente Martín Sánchez,Antonio José Molina de la Torre

doi:10.3390/math9060654

Abstract

Genome-wide association studies (GWAS) are observational studies of a large set of genetic variants in an individual’s sample in order to find if any of these variants are linked to a particular trait. In the last two decades, GWAS have contributed to several new discoveries in the field of genetics. This research presents a novel methodology to which GWAS can be applied to. It is mainly based on two machine learning methodologies, genetic algorithms and support vector machines. The database employed for the study consisted of information about 370,750 single-nucleotide polymorphisms belonging to 1076 cases of colorectal cancer and 973 controls. Ten pathways with different degrees of relationship with the trait under study were tested. The results obtained showed how the proposed methodology is able to detect relevant pathways for a certain trait: in this case, colorectal cancer.

Highlights

The results of the Human Genome Project [1] and the International HapMap Project [2]made it possible to find genes linked to traits and health problems
After having fixed the genetic algorithm (GA) parameters at a population size of 5500, 6000 iterations for each cycle and a 100% crossover with a mutation rate of 1%, the proposed algorithm was applied to 10 different pathways with different degrees of relationship with the trait under study
This paper presents a novel method called genetic algorithms support vector machines methodology (GASVeM), which is based on two wellknown machine learning methodologies—genetic algorithms, and support vector machines

Summary

Introduction

The results of the Human Genome Project [1] and the International HapMap Project [2]. Made it possible to find genes linked to traits and health problems. Genome-wide association studies (GWAS) have contributed to several new discoveries in human genetics. GWAS exploit the fact that genetic variants that are close together tend to be statistically correlated, something which in genetics is known as linkage disequilibrium [3]. The advances in genome arrays of genetic variations have led to the discovery of many DNA variants associated with complex traits such as those related to diseases. It can be said that until the development of GWAS it was not possible to find any gene linked to schizophrenia [5]

Objectives

Methods

Discussion

Conclusion