Extended Data Fig. 1: SWING’s performance on unseen experimental datasets. | Nature Methods

Extended Data Fig. 1: SWING’s performance on unseen experimental datasets.

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

Extended Data Fig. 1

a. Recall scores for H-2-IEk interacting peptides of different lengths for different SWING models, NetMHCIIpan 4.2 with variable length implementation, and MixMHC2pred 2.0. Blue, SWING Class I model; orange, SWING Class II model; magenta, SWING Mixed class model; green, MixMHC2pred 2.0; gray, NetMHCIIpan 4.2. b. AU-PR for H-2-IEk (left) and H-2-IAg7 (right) interacting and non-interacting peptides for different SWING models, NetMHCIIpan 4.2, and MixMHC2pred 2.0. Blue, SWING Class I model; orange, SWING Class II model; magenta, SWING Mixed model; green, MixMHC2pred 2.0; gray, NetMHCIIpan 4.2. c. Analysis of length bias across the 3 SWING model as reported by linear least-squares regression comparing the AUC for each peptide length using for H-2-IEk (left) and H-2-IAg7 (right). Significance was corrected using Bonferroni multiple comparisons correction. Additionally, an effect size threshold using the confidence interval was applied. P < 0.05. Blue, SWING Class I model; orange, SWING Class II model; magenta, SWING Mixed model. Created using BioRender.com.

Back to article page