Figure 3.

Relationships between the canonical discriminant functions for GBM, AO and the group centroids. The two distinctive clusters formed by GBM and AO suggest three LDA-based models have excellent discriminant ability for GBM and AO. GBM, glioblastoma; AO, anaplastic oligodendroglioma; LDA, linear discriminant analysis; LASSO, least absolute shrinkage and selection operator; GBDT, gradient boosting decision tree; SVM, support vector machine. (A) Canonical discriminant functions for Distance Correlation + LDA. (B) Canonical discriminant functions for LASSO + LDA. (C) Canonical discriminant functions for GBDT + LDA.