Neural networks ensemble for automatic DNA microarray spot classification
- 222 Downloads
In this work, a new step for the DNA microarray image analysis pipeline is proposed using neural computing techniques. We perform the classification of the spots into morphology-derived classes in order to assist the segmentation procedure that is traditionally performed after the gridding process. Our method consists of extracting multiple features from each individual spot area (or cell—derived from the gridding process) that are then reduced to a presumably optimal subset using a feature selection process, the sequential forward selection algorithm. Classification is then realized by means of a neural network ensemble with a tree-like structure, made up of seven multi-layer perceptron networks. The architecture of each neural network has been obtained through an exhaustive automatic searching process that optimizes the size of the network as a function of the classification error rate. The neural ensemble classifier is tested on two sub-grids extracted from real microarray DNA images and is shown to achieve high accuracy rates over the seven different classes of spot. In addition, a dataset with more than 1000 samples of classes of spot has been generated and made freely available.
KeywordsDNA microarray images Spot classification Neural networks ensemble Optimization Sequential forward selection Image processing
The authors are grateful for the resources made available by the “Laboratorio de Investigaciones Tecnológicas en Reconocimiento de Patrones”, Universidad Católica del Maule, Talca, Chile.
Compliance with ethical standards
Conflict of interest
The authors declare that they have no conflict of interest.
- 7.Álvarez-Ramos C, Nino E, Santos M (2013) Automatic classification of Nosema pathogenic agents through machine vision techniques and kernel based vector machines. In: Computing Colombian conference (8CCC). IEEE, pp 1–5Google Scholar
- 14.Daskalakis A, Cavouras D, Bougioukos P, Kostopoulos S, Georgiadis P, Kalatzis I, Nikiforidis G (2007) Effective quantification of gene expression levels in microarray images using a spot-adaptive compound clustering-enhancement-segmentation scheme. In Computational science and its applications–ICCSA 2007. Springer, Berlin, pp 555–565Google Scholar
- 29.Gonzalez R, Woods R (2008) Digital image processing, 3rd edn. Prentice-Hall, Upper Saddle RiverGoogle Scholar
- 33.Foresee FD, Hagan MT (1997) Gauss–Newton approximation to Bayesian learning. In International conference on neural networks, 1997, vol 3. IEEE, pp 1930–1935Google Scholar
- 35.Daskalakis A, Glotsos D, Kostopoulos S, Cavouras D, Nikiforidis G (2009) A comparative study of individual and ensemble majority vote cDNA microarray image segmentation schemes, originating from a spot-adjustable based restoration framework. Comput Methods Programs Biomed 95(1):72–88CrossRefGoogle Scholar