Fig. 4

Identification using machine learning algorithms. A Lasso regression was used to screen the key genes of GSE73680, with the optimal gene count (n = 7) at the curve’s lowest point. B-C Core genes were screened by Boruta algorithm, and a total of 2 genes were screened. D Lasso regression was used to screen the key genes of GSE56815, with the optimal gene count (n = 10) at the curve’s lowest point. E–F Core genes were screened by Boruta algorithm, and a total of 9 genes were screened. G Common genes identified by the two algorithms