Skip to main content
Communications Biology logoLink to Communications Biology
. 2025 Jul 5;8:1007. doi: 10.1038/s42003-025-08422-z

Author Correction: Geometric deep learning improves generalizability of MHC-bound peptide predictions

Dario F Marzella 1,#, Giulia Crocioni 2,#, Tadija Radusinović 3, Daniil Lepikhov 1, Heleen Severin 1, Dani L Bodor 2, Daniel T Rademaker 1, ChiaYu Lin 2, Sonja Georgievska 2, Nicolas Renaud 2, Amy L Kessler 4, Pablo Lopez-Tarifa 2, Sonja I Buschow 4, Erik Bekkers 3, Li C Xue 1,
PMCID: PMC12228751  PMID: 40617886

Correction to: Communications Biology 10.1038/s42003-024-07292-1, published online 19 December 2024

In the version of the article initially published, due to an inconsistency in the IDs assigned to the binding affinity cases in one post-processing step for the CNN outputs, the previously reported performances of the CNN in Fig. 2D were erroneous. The amended plot represents the correct per-allele performances for the CNN, which are more consistent with the performances of the other networks, especially for HLA-B*57:02 and HLA-C*07:02. The other networks’ performances were correct in the previous version of this plot and are therefore unchanged in this updated version, as shown in Fig. 1. Also, the order of the alleles on the x axis has been updated to more correctly represent the distance from the training set, using more decimal values to better separate alleles with similar distance.

Fig. 1 Original Fig. 2D graphic file with name 42003_2025_8422_Figa_HTML.jpg

Fig. 1 Corrected Fig. 2D graphic file with name 42003_2025_8422_Figb_HTML.jpg


Articles from Communications Biology are provided here courtesy of Nature Publishing Group

RESOURCES