Fig. 7
From: A machine learning approach to predict metabolic pathway dynamics from time-series multiomics data
Prediction errors decrease markedly with increasing training set size. As the number of available proteomics and metabolomics times-series data sets (strains) for training increases, the prediction error (RMSE, Eq. (6)) decreases conspicuously. Moreover, the standard deviation of the predictions error (vertical bar) decreases notably as well. The change from 2 to 10 strains is more pronounced that the change from 10 to 100. This fact indicates that it is more productive to do ten rounds of metabolic engineering collecting ten time-series data sets, than a single round collecting 100 time series
