Fig. 6
From: Identifying glycan motifs using a novel subtree mining approach
Predicted carbohydrate-binding motifs of two haemagglutinins from a human and an avian strain of influenza from CFG glycan microarray data. a Distribution of RFUs and classification of non-binding (blue), intermediate binding (orange), and binding glycans (red) for A/Puerto Rico/8/34 (H1N1) HA. b ROC curves for the test (n=138) and training (n=412) sets for A/Puerto Rico/8/34 (H1N1) HA. The ratio of negative to positive samples was 26.5. c Logistic regression coefficients for identified motifs for A/Puerto Rico/8/34 (H1N1) HA. d Distribution of RFUs and classification of non-binding (blue), intermediate binding (orange), and binding glycans (red) for A/harbor seal/Massachusetts/1/2011 (H3N8) HA. e ROC curves for the test (n=145) and training (n=433) sets for A/harbor seal/Massachusetts/1/2011 (H3N8) HA. The ratio of negative to positive samples was 11.4. f Logistic regression coefficients for identified motifs for A/harbor seal/Massachusetts/1/2011 (H3N8) HA. See Additional file 1 for a detailed notation key