NET-Related Gene as Potential Diagnostic Biomarkers for Diabetic Tubulointerstitial Injury

Figure 3

Machine learning algorithms applied to signature gene identification. (a) Penalty parameter selection in the LASSO model depicted with error bars representing standard errors. (b) LASSO coefficient profiles demonstrating shrinkage as the penalty parameter increases. (c) Confidence intervals for error rates in the random forest model. (d) Interplay between LASSO and random forest models. (e) Genes with a relative importance exceeding 0.15 in the random forest model.

(a)

(b)

(c)

(d)

(e)