Fitting Rank Order Data in the Age of Context

Abstract

With the advent of high performance computing and commensurate increase in data, the opportunity to capture the overall distribution of values ("context") by means of nonparametric curve fitting (LOESS) enables the identification of exceptional points in large datasets. We revisit the assumptions of the kernel functions for nonparametric curve fitting of biological and biomedical data exhibiting these rare or exceptional instances. We propose a new linear asymmetric kernel function and evaluate its ability to fit rank order data in the domain of protein-protein interaction prediction. The proposed linear kernel significantly improved predictive performance (p < 0.001) of two state-of-the-art predictors.