Fig. 6

Gene co-expression data. Applying the proposed method to gene co-expression data from the plant Arabidopsis thaliana shows that the code length savings in the test network has a maximum at threshold \(\tau ^* \approx 0.75\) (left). This is the threshold value giving the most parsimonious model of the data. The right figure shows that the code length savings at the optimal threshold do not increase linearly with the number of samples, indicating in accordance with the results based on synthetic data that the reward for sampling effort decreases as the number of samples increases