Skip to main content
Fig. 6 | BMC Bioinformatics

Fig. 6

From: Prediction and analysis of multiple protein lysine modified sites based on conditional wasserstein generative adversarial networks

Fig. 6

a Distance comparisons of 6 categories between CGAN and CWGAN. The Euclidean distance was calculated between the mean value of the simulated data and the real data. Distances using CGAN are all above 0.1, while distances with CWGAN of 6 categories are below 0.03, indicating that simulation data aided by CWGAN are more similar to the original real data. P value was calculated by two-sided Mann–Whitney U test. b Loss vs iterations graph of CGAN and CWGAN. The loss iteration graph reflects the stability of the iterative process of different algorithms, which proves that the CWGAN-training process is more even. (Python 3.8)

Back to article page