Artificial Neural Network Analysis of DNA Microarray-based Prostate Cancer Recurrence

Abstract

-DNA microarray-based gene expression profiles have been established for a variety of adult cancers. This paper addresses application of an artificial neural network (ANN) with leave-oneout testsing and 8-fold cross-validation for analyzing DNA microarray data to identify genes predictive of recurrence after prostatectomy. Among 725 genes screened for ANN input, a 16-gene model resulted in 99-100% diagnostic sensitivity and specificity: DGCR5,

Artificial neural network classification of microarray data using new hybrid gene selection method

“…Furthermore, for those methods, selecting fewer genes than the ICA + ABC algorithm, their classification accuracy is less than that of the ICA + ABC algorithm. Therefore, it is clear from the experimental result that the proposed ICA + ABC algorithm achieves the highest classification accuracy and the lowest average of selected genes when tested using all data sets, as compared to the original ABC algorithm and previously proposed (Fernández-Navarro et al, 2012;Peterson et al, 2005;Sahu and Mishra, 2012) algorithm for ANN classifier. Hence, the ICA + ABC algorithm is a promising approach for solving gene selection problems and improves the cancer classification performance of ANN.…”

“…ANN has been widely applied in DNA microarrays for supervised and unsupervised learning (Tong and Schierz, 2011). For example, Peterson et al (2005) used a Multilayer Perceptron (MLP) with back-propagation learning and a dimensional reduction method based on k-means and Principal Component Analysis (PCA) techniques for prostate cancer data sets of microarray. Authors classify and predict cancers using Small Round Blue Cell Tumours (SRBCT) with PCA as dimension reduction and train ANN models with no hidden layers for classification (Khan et al, 2001).…”

Artificial neural networks as classification and diagnostic tools for lymph node-negative breast cancers

“…Second, it provides models of the molecular mechanisms underlying metastasis. Finally, network-based classification achieves higher accuracy in prediction, as ascertained by selecting markers from one data set and applying them to a second independent validation data set [23]. Neural networks are used here for classification of lymph node negative breast cancer genes based on the information content of microarray data [1] due to its ability of training easily and tractability to tackle higher amount of information with good generalization ability, resulting in cost-effective and flexible modelling of large data sets.…”

“…Prevalent studies have been established for classification by the statisticians, but the networkbased method emerging as a new area of research is viewed to provide better efficiency for classification problems. We have used neural networks because they have numerous applications in handling large databases [22,23], which includes biochemical engineering, biomedical science and bioinformatics [2,3,21], DNA sequence analysis and biological pattern recognition [24]. There are a number of reasons to incorporate prognostic markers with ANN, which has been thought to provide an efficient diagnostic methodology; First, prognostic signature belongs to many functional classes, which suggests that different paths could lead to disease progression, hence providing better means of detection.…”

Cancer Recurrence Prediction Using Machine Learning

“…There have been some attempts to predict cancer recurrence using machine learning techniques. Peterson et al [18] applied old-fashioned artificial neural networks (ANNs) with back propagation to predict prostate cancer recurrence of patients after undergoing radical prostatectomy. After gene screening and optimization, they claimed to have achieved 0.99 to 1.0 diagnostic sensitivity and specificity.…”