Research Paper Volume 15, Issue 8 pp 3120—3140
An artificial neural network model to diagnose non-obstructive azoospermia based on RNA-binding protein-related genes
Figure 4. DDX20 and NCBP2 were con-determined via feature selection methods and PPI network analysis. (A) 5 genes, including NCBP2, DDX20, TSN, SRPK2, and CARHSP1, were identified as significant features to NOA via LASSO regression. (B) The coefficients of the 5 selected genes in the LASSO regression model. (C) 30 genes were determined as important features via the Boruta algorithm. (D) 6 genes, including NCBP2, DDX20, CCDC86, TSN, CARHSP1, and TDRD7, were selected by the SVM-RFE. (E) DDX20 and NCBP2 were con-determined by the machine learning algorithms and PPI network analysis.