Fig. 4: Comparison of DeepTMP and the stat-of-the-art methods. | Nature Communications

Fig. 4: Comparison of DeepTMP and the stat-of-the-art methods.

From: Deep transfer learning for inter-chain contact predictions of transmembrane protein complexes

Fig. 4

a The violin plots of top L precisions by DeepTMP, CDPred, DeepHomo2.0, GLINTER, DeepHomo and DNCON2_Inter on n = 52 cases in the TMP test set. The box plot inside the violin displays the 25th and 75th percentiles of the data, and the dot indicates the median. The whiskers extend to the highest and lowest values within 1.5 times the interquartile range of the data. b The top L precisions of DeepTMP versus those of CDPred, DeepHomo2.0, GLINTER, DeepHomo, and DNCON2_Inter on the test set. c The 3D structures of 5TIN with predicted inter-chain contacts by DeepTMP, CDPred, DeepHomo2.0, and GLINTER. The correct (incorrect) predicted contacts are colored in green (red). The parts between two dashed lines are the transmembrane regions.

Back to article page