Skip to main content
. 2022 Dec 8;12:21251. doi: 10.1038/s41598-022-25758-2

Figure 2.

Figure 2

Development of a prediction model using a DBN algorithm. (a) Schematic of the molecular subtype in vitro exercise prediction model. Nine publicly available datasets (n = 108) were used to determine the predictive value of the model based on the H2O deep-learning method. (b) Prediction results obtained using the molecular subtype prediction model. Among 60 sedentary samples, 50 were predicted accurately as control samples. Among 48 exercise samples, 32 were predicted correctly as EPS samples.