Fig. 4
From: MicroRNA target prediction using thermodynamic and sequence curves
Performance of model on V-CLIP data by progressively varying the fraction of mRNAs used for training and prediction. In the first case, data for 1 % of the total mRNAs (around 1200) was used for training and data for the rest of the mRNAs was used to evaluate the model. The percentage of mRNAs used for training was then progressively increased to 20 %, 40 %, and finally 60 %. These three curves are overlapping, indicating that there is no performance improvement by increasing the training size beyond 20 %