Skip to main content
. 2016 Dec 18;6(1):67–78. doi: 10.1002/cam4.951

Figure 1.

Figure 1

Three cohorts of miRNA profiles of 10 control and 10 OSCC samples were created from TCGA dataset of oral tumors and used to train 3 distinct OSCC classifiers. (A) Receiver operating characteristic curve of the Bayesian Compound Covariate based classifier for OSCC versus normal tested using leave‐one‐out cross‐validation cohort 1. (B) Receiver operating characteristic curve of the Bayesian Compound Covariate based classifier for OSCC versus normal tested using leave‐one‐out cross‐validation cohort 2. (C) Receiver operating characteristic curve of the Bayesian Compound Covariate based classifier for OSCC versus normal tested using Leave‐one‐out cross‐validation cohort 3. (D) The classifier generated on cohort 1 produced the diagrammed receiver operating characteristic curve when tested on 10 controls plus 20 OSCCs. (E) Same as (D) except the cohort 2 trained classifier was tested. (F) Same as (D) except the cohort 3 trained classifier was tested.