Analysis of premalignant pancreatic cancer mass spectrometry data for biomarker selection using a group search optimizer

This paper presents a novel group search optimizer (GSO)-based biomarker discovery method for pancreatic cancer diagnosis using mass spectrometry (MS) data. The GSO was inspired by animal social searching behaviour. It has been shown that the global search performance of the GSO is competitive to other biologically inspired optimization algorithms. In this study, we applied a GSO as a feature selection method to MS data analysis for premalignant pancreatic cancer biomarker discovery. We first applied a smooth non-linear energy operator to detect peaks. Then a GSO with linear discriminant analysis was used to select a parsimonious set of peak windows (biomarkers) that can distinguish cancer. After selecting a set of biomarkers, a support vector machine was then applied to build a classifier to diagnosis premalignant cancer cases. We compared the GSO algorithm with a genetic algorithm, evolution strategies, evolutionary programming and a particle swarm optimizer. The results showed that the GSO-based feature selection algorithm is capable of selecting a parsimonious set of biomarkers to achieve better classification performance than other algorithms. The source code of the proposed GSO-based feature selection algorithm is available at www.cs.bham.ac.uk/similar to szh.