Fig. 6. GlyNet predictions across the 599 glycans for two protein samples with well understand recognition profiles. The false positives in GlyNet's learned outputs contain the same iconic motifs as the true positives. The glycans with the strongest predicted RFUs, but which are not binders (by the CFG data) are shown with pink dots (left side), and their structures are shown above the plots. Those with the strongest RFUs and are positive binders in the CFG data are red (in the upper right). Also shown is the mean CFG data (black line) and its 95% CI (grey band) from N = 6 replicates. The small numbers at the bottom of the glycan structures represent the spacer/linkers as indicated on the CFG website (https://www.functionalglycomics.org/static/consortium/resources/resourcecoreh8.shtml), and details can be found in ESI Table S6.† (a) Both true and false positives for an anti-blood group B anti-body (https://youtu.be/oHaFF4A22D8?t=374) contain blood group B trisaccharide (Gal(α1–3)[Fuc(α1–2)]Gal(β1–)). (b) A sample of influenza hemagglutinin shows several positive predictions containing Neu5Ac(α2–6)Gal(β1–) (https://youtu.be/oHaFF4A22D8?t=435).