Identifying Macrophage-Related Genes in Ulcerative Colitis Using Weighted Coexpression Network Analysis and Machine Learning

Figure 5

Identification of potential biomarkers for UC based on machine learning algorithms. (a and b) LASSO regression screening variables. (c and d) Ranger provided variable importance scores (VIS) for five genes only in patients with UC. Out-of-bag (OOB) error rate of these genes in the model together with genes included individually based on their VIS ranks. (e and f) Optimal error rate and accuracy of SVM model based on 29 characteristic genes. (g) Venn diagram shows overlapping genes in LASSO, SVM, and SWSFS. (h) ROC curves of SLC6A14, HMGCS2, and ENPP1 in tissues from GSE87473 and GSE38713.

(a)

(b)

(c)

(d)

(e)

(f)

(g)

(h)