Feature Genes in Neuroblastoma Distinguishing High-Risk and Non-high-Risk Neuroblastoma Patients: Development and Validation Combining Random Forest With Artificial Neural Network

Abstract

There is a significant difference in prognosis among different risk groups. Therefore, it is of great significance to correctly identify the risk grouping of children. Using the genomic data of neuroblastoma samples in public databases, we used GSE49710 as the training set data to calculate the feature genes of the high-risk group and non-high-risk group samples based on the random forest (RF) algorithm and artificial neural network (ANN) algorithm. The screening results of RF showed that EPS8L1, PLCD4, CHD5, NTRK1, and SLC22A4 were the feature differentially expressed genes (DEGs) of high-risk neuroblastoma. The prediction model based on gene expression data in this study showed high overall accuracy and precision in both the training set and the test set (AUC = 0.998 in GSE49710 and AUC = 0.858 in GSE73517). Kaplan–Meier plotter showed that the overall survival and progression-free survival of patients in the low-risk subgroup were significantly better than those in the high-risk subgroup [HR: 3.86 (95% CI: 2.44–6.10) and HR: 3.03 (95% CI: 2.03–4.52), respectively]. Our ANN-based model has better classification performance than the SVM-based model and XGboost-based model. Nevertheless, more convincing data sets and machine learning algorithms will be needed to build diagnostic models for individual organization types in the future.

Predicting Neuroblastoma Patient Risk Groups, Outcomes, and Treatment Response Using Machine Learning Methods: A Review

“…In the survival section, ML algorithms used for constructing models to predict patient survival were addressed [ 13 , 49 , 51 ]. Another study attempted to combine RF- and ANN-based models to link NB patient genomic data with patient survival [ 13 ]. Accordingly, the authors selected the GSE49710 and GSE73517 datasets for training and testing, respectively.…”

“…Accordingly, the authors selected the GSE49710 and GSE73517 datasets for training and testing, respectively. GSE49710 featured 176 and 322 high- and low-risk cases, respectively, while GSE73517 featured 56 and 49 high- and low-risk cases, respectively [ 13 ]. Initially, the differentially expressed genes in the datasets were determined, and 94 differentially expressed genes were fed to RF.…”

“…Initially, the differentially expressed genes in the datasets were determined, and 94 differentially expressed genes were fed to RF. Of the 500 DTs constructed, 290 with the least error were selected, and 32 differentially expressed genes (variables) with an importance of >2 important values were obtained (including, PLCD4 , NTRK1 , and EPS8L1 ) [ 13 ]. Finally, this subset of 32 genes (variables) was shown to represent the NB high-risk feature genes by the K-means supervised clustering [ 13 ].…”

“…Of the 500 DTs constructed, 290 with the least error were selected, and 32 differentially expressed genes (variables) with an importance of >2 important values were obtained (including, PLCD4 , NTRK1 , and EPS8L1 ) [ 13 ]. Finally, this subset of 32 genes (variables) was shown to represent the NB high-risk feature genes by the K-means supervised clustering [ 13 ]. ANN then constructed a model that showed a high AUC when tested for both the training (0.998) and testing (0.858) datasets.…”

“…These technologies have also led to the proposition and improvement of predictive models of patient prognosis and survival [ 12 ]. Many studies have used machine learning (ML) and statistical methods to develop multi-gene models or other predictor models to anticipate patient subgrouping, risk, outcome (including survival, relapse, and recurrence), and treatment response [ 13 ]. It is important to note that patient outcomes refer to the destination of the clinical intervention used, which will provide feedback on its efficacy, and therefore patient survival and mortality rate would be viewed as a subset of the patient outcomes [ 14 ].…”