Fig. 2: SWING predicts pMHC. | Nature Methods

Fig. 2: SWING predicts pMHC.

From: Sliding Window Interaction Grammar (SWING): a generalized interaction language model for peptide and protein interactions

Fig. 2

a, Schematic of the pMHC prediction task adaptation of the SWING framework. b, Representation of the SCV evaluation metric. c, Depiction of the cross-prediction evaluation metric. d, Class I allele functional clustering as defined by MHCCluster v.2.0. Orange, alleles in the training set; magenta, alleles in the validation set and blue, distant allele validation set. e, SWING class I model performance plotted across ten replicates of tenfold cross-validation with permutation testing defined by the AUROC. Blue, validation curve; red, permuted mean; green, perfect classifier and gray, random classifier. f, Class I model performance on three unseen functionally close alleles in the validation set as defined by the AUROC. Blue, HLA-A02:02; orange, HLA-B40:02; magenta, HLA-C05:01; green, perfect classifier and gray, random classifier. g, Class I model cross-prediction performance on three unseen functionally distinct alleles in the distant validation set as defined by the AUROC. Blue, HLA-A32:01; orange, HLA-B38:01; magenta, HLA-C03:03; green, perfect classifier and gray, random classifier. h, Class II allele functional clustering as defined by MHCCluster v.2.0. Orange, alleles in the training set and blue, alleles in the validation set. i, Class II model performance plotted across ten replicates of tenfold cross-validation with permutation testing defined by the AUROC. Blue, validation curve; red permuted mean; green, perfect classifier and gray, random classifier. j, Class II model cross-prediction performance on two unseen alleles in the validation set as defined by the AUROC. Blue, DRB1_0102; orange, DRB1_0404; green, perfect classifier and gray, random classifier. Mean AUROC ± 2 × standard deviation. Panels a–c created using BioRender.com.

Back to article page