A Note on Applications of Support Vector Machine

TL;DR

The study treats SNP variation data as numeric vectors in $R^n$ and frames disease association as a binary classification problem solvable via Support Vector Machines (SVM). It grounds the method in maximal-margin hyperplane optimization and connects it to Kuhn–Tucker conditions, arguing for SVM's broad applicability in biology. Preliminary cardio-patient experiments using Joachims' SVM implementation show separability between disease and non-disease classes across several tests with varying $C$, illustrating practical feasibility. The work points to versatile, data-driven biomedical decision support that can integrate genotype-phenotype data alongside non-numeric inputs such as imaging.

Abstract

We describe in a rudimentary fashion how SVM(support vector machine) plays the role of classifier in a mathematical setting. We then discuss its application in the study of multiple SNP(single nucleotide polymorphism) variations. Also presented is a set of preliminary test results with clinical data.