Skip to main content
. 2021 Jun 3;11:11698. doi: 10.1038/s41598-021-91278-0

Figure 2.

Figure 2

The diagnostic microRNA signatures designed on cohort from one recruitment site are not well performing on cohorts from a different recruitment site. (A) Flowchart of the Random Forest procedure: Belgian-recruited subjects are used to design the models, and validation is performed either on Belgian or Rwandan-recruited subjects. (B) Performances of the 13 signatures on two validation datasets containing subjects from Rwanda [(MATCHED + REST)-RW] or Belgium [REST-BE]. (C) Flowchart of the Random Forest procedure: Rwandan-recruited subjects are used to design the models, and validation is performed either on Belgian or Rwandan-recruited subjects. (D) Performances of the 13 signatures on two validation datasets containing subjects from Rwanda [REST-RW] or Belgium [(MATCHED + REST)-BE].