Identifying Patients with Atrioventricular Septal Defect in Down Syndrome Populations by Using Self-Normalizing Neural Networks and Feature Selection

Abstract

Atrioventricular septal defect (AVSD) is a clinically significant subtype of congenital heart disease (CHD) that severely influences the health of babies during birth and is associated with Down syndrome (DS). Thus, exploring the differences in functional genes in DS samples with and without AVSD is a critical way to investigate the complex association between AVSD and DS. In this study, we present a computational method to distinguish DS patients with AVSD from those without AVSD using the newly proposed self-normalizing neural network (SNN). First, each patient was encoded by using the copy number of probes on chromosome 21. The encoded features were ranked by the reliable Monte Carlo feature selection (MCFS) method to obtain a ranked feature list. Based on this feature list, we used a two-stage incremental feature selection to construct two series of feature subsets and applied SNNs to build classifiers to identify optimal features. Results show that 2737 optimal features were obtained, and the corresponding optimal SNN classifier constructed on optimal features yielded a Matthew’s correlation coefficient (MCC) value of 0.748. For comparison, random forest was also used to build classifiers and uncover optimal features. This method received an optimal MCC value of 0.582 when top 132 features were utilized. Finally, we analyzed some key features derived from the optimal features in SNNs found in literature support to further reveal their essential roles.

Integrative Analysis of Methylation and Gene Expression in Lung Adenocarcinoma and Squamous Cell Lung Carcinoma

“…After the irrelevant features were removed, the relevant methylation and expression features were ranked based on their importance evaluated with MCFS (Monte Carlo Feature Selection) (Draminski et al, 2008). The MCFS was a widely used method to rank features based on classification trees (Chen et al, , 2019Pan et al, 2018Pan et al, , 2019aLi et al, 2019). First, for the d features, we selected s subsets and each subset included m features (m was much smaller than d).…”

The Functional Effects of Key Driver KRAS Mutations on Gene Expression in Lung Cancer

“…We tried several different classifiers: (1) SVM (Support Vector Machine) (Jiang et al, 2019;Yan et al, 2019;Chen et al, 2019a;Li et al, 2019a;Pan et al, 2019a;Wang and Huang, 2019b;Chen et al, 2019d), (2) 1NN (1 Nearest Neighbor) (Lei et al, 2013;Chen et al, 2016;Wang et al, 2017a), (3) 3NN (3 Nearest Neighbors), (4) 5NN (5 Nearest Neighbors), (5) Decision Tree (DT) (Huang et al, 2008;Huang et al, 2011;Chen et al, 2015), (6) Neural Network (NN) (Liu et al, 2017;Pan et al, 2018;Chen et al, 2019e). The function svm from R package e1071, function knn from R package class, function rpart from R package rpart, function nnet from R package nnet were used to apply these classification algorithms.…”

“…where TP, TN, FP, and FN stand for the number of true positive samples, true negative samples, false positive samples, and false negative samples, respectively. Since the sizes of KRAS mutation + samples and KRAS mutation -samples were imbalance and MCC can trade-off sensitivity and specificity (Chen et al, 2018a;Pan et al, 2018;Pan et al, 2019a;Pan et al, 2019b), MCC was used as the main performance metric.…”

Tissue differences revealed by gene expression profiles of various cell lines

“…12 To date, it has been successfully applied to tackle several such data sets. [18][19][20] Thus, in this study, we also used the MCFS method to analyze gene expression features. The MCFS method can rank investigated features by using the bootstrap technique and a decision tree algorithm.…”