Figure 3
From: OutPredict: multiple datasets can improve prediction of expression and inference of causality
Arabidopsis in Shoot Tissue (time series only dataset) (A) Predicted gene expression using OutPredict (grey dots) compared to actual expression (red line) at the left-out time point. (B) Comparison of time series forecasting: the accuracy of forecasting, measured by Mean Squared Error, has higher values in this case than for other species, because the data is RNAseq and read counts have a broad dynamic range. Table 2 describes which method and data were used for each model in the x axis. OutPredict (OP) performs 34.2% better than Penultimate Value (P < 0.05, non-parametric paired test), and 61.5% better than Dynamic Genie3 (P < 0.05, non-parametric paired test). The incorporation of priors from TARGET (OP-Priors) improves the performance of OutPredict compared to the time series alone (9% improvement with P = 0.12, non-parametric paired test). The ODE-log model is better than Time-Step based on the out-of-bag score. The Neural Network model doesn’t converge because the dataset is small.
