Learning to Discover Faulty Spots in cDNA Microarrays

Gene expression ratios obtained from microarray images are strongly affected by the algorithms used to process them as well as by the quality of the images. Hundreds of spots often suffer from quality problems caused by the manufacturing process and many must be discarded because of lack of reliability. Recently, several computational models have been proposed in the literature to identify defective spots, including the powerful Support Vector Machines (SVMs). In this paper we propose to use different strategies based on aggregation methods to classify the spots according to their quality. On one hand we apply an ensemble of classifiers, in particular three boosting methods, namely Discrete, Real and Gentle AdaBoost. As we use a public dataset which includes the subjective labeling criteria of three human experts, we also evaluate different ways of modeling consensus between the experts. We show that for this problem ensembles achieve improved classification accuracies over alternative state-of-the-art methods.