Machine learning approaches for the prediction of polygenic obesity using genome-wide genotyping data

In the last decade, Polygenic Risk Scores (PRSs) have been widely used to identify individuals at high risk of being obese. In the present study, we propose to consider a variety of Machine Learning (ML) algorithms to estimate the prediction of obesity. We benchmarked PRS, Random Forest (RF), Gradient Boosting (GB) and Support Vector Machine (SVM) on a database of thousands of patients and millions of Single Nucleotide Polymorphisms (SNPs). To deal with the high-dimensional data we needed to deploy a HDP (Hortonworks Data Platform) cluster composed of multiple machines in order to distribute our data and processes. We used the most popular tools for processing Big Data, Apache Hadoop and Apache Spark, which allow for the parallelization across the cluster