Geometric deep learning improves generalizability of MHC-bound peptide predictions

Abstract The interaction between peptides and major histocompatibility complex (MHC) molecules is pivotal in autoimmunity, pathogen recognition and tumor immunity. Recent advances in cancer immunotherapies demand for more accurate computational prediction of MHC-bound peptides. We address the genera...

Full description

Saved in:
Bibliographic Details
Main Authors: Dario F. Marzella, Giulia Crocioni, Tadija Radusinović, Daniil Lepikhov, Heleen Severin, Dani L. Bodor, Daniel T. Rademaker, ChiaYu Lin, Sonja Georgievska, Nicolas Renaud, Amy L. Kessler, Pablo Lopez-Tarifa, Sonja I. Buschow, Erik Bekkers, Li C. Xue
Format: Article
Language:English
Published: Nature Portfolio 2024-12-01
Series:Communications Biology
Online Access:https://doi.org/10.1038/s42003-024-07292-1
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1846112266962862080
author Dario F. Marzella
Giulia Crocioni
Tadija Radusinović
Daniil Lepikhov
Heleen Severin
Dani L. Bodor
Daniel T. Rademaker
ChiaYu Lin
Sonja Georgievska
Nicolas Renaud
Amy L. Kessler
Pablo Lopez-Tarifa
Sonja I. Buschow
Erik Bekkers
Li C. Xue
author_facet Dario F. Marzella
Giulia Crocioni
Tadija Radusinović
Daniil Lepikhov
Heleen Severin
Dani L. Bodor
Daniel T. Rademaker
ChiaYu Lin
Sonja Georgievska
Nicolas Renaud
Amy L. Kessler
Pablo Lopez-Tarifa
Sonja I. Buschow
Erik Bekkers
Li C. Xue
author_sort Dario F. Marzella
collection DOAJ
description Abstract The interaction between peptides and major histocompatibility complex (MHC) molecules is pivotal in autoimmunity, pathogen recognition and tumor immunity. Recent advances in cancer immunotherapies demand for more accurate computational prediction of MHC-bound peptides. We address the generalizability challenge of MHC-bound peptide predictions, revealing limitations in current sequence-based approaches. Our structure-based methods leveraging geometric deep learning (GDL) demonstrate promising improvement in generalizability across unseen MHC alleles. Further, we tackle data efficiency by introducing a self-supervised learning approach on structures (3D-SSL). Without being exposed to any binding affinity data, our 3D-SSL outperforms sequence-based methods trained on ~90 times more data points. Finally, we demonstrate the resilience of structure-based GDL methods to biases in binding data on an Hepatitis B virus vaccine immunopeptidomics case study. This proof-of-concept study highlights structure-based methods’ potential to enhance generalizability and data efficiency, with possible implications for data-intensive fields like T-cell receptor specificity predictions.
format Article
id doaj-art-74a496029f184ad6b454f6b29b4a4566
institution Kabale University
issn 2399-3642
language English
publishDate 2024-12-01
publisher Nature Portfolio
record_format Article
series Communications Biology
spelling doaj-art-74a496029f184ad6b454f6b29b4a45662024-12-22T12:41:59ZengNature PortfolioCommunications Biology2399-36422024-12-017111210.1038/s42003-024-07292-1Geometric deep learning improves generalizability of MHC-bound peptide predictionsDario F. Marzella0Giulia Crocioni1Tadija Radusinović2Daniil Lepikhov3Heleen Severin4Dani L. Bodor5Daniel T. Rademaker6ChiaYu Lin7Sonja Georgievska8Nicolas Renaud9Amy L. Kessler10Pablo Lopez-Tarifa11Sonja I. Buschow12Erik Bekkers13Li C. Xue14Medical BioSciences department, Radboudumc, Radboud University Medical CenterNetherlands eScience CenterUniversity of AmsterdamMedical BioSciences department, Radboudumc, Radboud University Medical CenterMedical BioSciences department, Radboudumc, Radboud University Medical CenterNetherlands eScience CenterMedical BioSciences department, Radboudumc, Radboud University Medical CenterNetherlands eScience CenterNetherlands eScience CenterNetherlands eScience CenterDepartment of Gastroenterology and Hepatology, Erasmus MC, University Medical Center RotterdamNetherlands eScience CenterDepartment of Gastroenterology and Hepatology, Erasmus MC, University Medical Center RotterdamUniversity of AmsterdamMedical BioSciences department, Radboudumc, Radboud University Medical CenterAbstract The interaction between peptides and major histocompatibility complex (MHC) molecules is pivotal in autoimmunity, pathogen recognition and tumor immunity. Recent advances in cancer immunotherapies demand for more accurate computational prediction of MHC-bound peptides. We address the generalizability challenge of MHC-bound peptide predictions, revealing limitations in current sequence-based approaches. Our structure-based methods leveraging geometric deep learning (GDL) demonstrate promising improvement in generalizability across unseen MHC alleles. Further, we tackle data efficiency by introducing a self-supervised learning approach on structures (3D-SSL). Without being exposed to any binding affinity data, our 3D-SSL outperforms sequence-based methods trained on ~90 times more data points. Finally, we demonstrate the resilience of structure-based GDL methods to biases in binding data on an Hepatitis B virus vaccine immunopeptidomics case study. This proof-of-concept study highlights structure-based methods’ potential to enhance generalizability and data efficiency, with possible implications for data-intensive fields like T-cell receptor specificity predictions.https://doi.org/10.1038/s42003-024-07292-1
spellingShingle Dario F. Marzella
Giulia Crocioni
Tadija Radusinović
Daniil Lepikhov
Heleen Severin
Dani L. Bodor
Daniel T. Rademaker
ChiaYu Lin
Sonja Georgievska
Nicolas Renaud
Amy L. Kessler
Pablo Lopez-Tarifa
Sonja I. Buschow
Erik Bekkers
Li C. Xue
Geometric deep learning improves generalizability of MHC-bound peptide predictions
Communications Biology
title Geometric deep learning improves generalizability of MHC-bound peptide predictions
title_full Geometric deep learning improves generalizability of MHC-bound peptide predictions
title_fullStr Geometric deep learning improves generalizability of MHC-bound peptide predictions
title_full_unstemmed Geometric deep learning improves generalizability of MHC-bound peptide predictions
title_short Geometric deep learning improves generalizability of MHC-bound peptide predictions
title_sort geometric deep learning improves generalizability of mhc bound peptide predictions
url https://doi.org/10.1038/s42003-024-07292-1
work_keys_str_mv AT dariofmarzella geometricdeeplearningimprovesgeneralizabilityofmhcboundpeptidepredictions
AT giuliacrocioni geometricdeeplearningimprovesgeneralizabilityofmhcboundpeptidepredictions
AT tadijaradusinovic geometricdeeplearningimprovesgeneralizabilityofmhcboundpeptidepredictions
AT daniillepikhov geometricdeeplearningimprovesgeneralizabilityofmhcboundpeptidepredictions
AT heleenseverin geometricdeeplearningimprovesgeneralizabilityofmhcboundpeptidepredictions
AT danilbodor geometricdeeplearningimprovesgeneralizabilityofmhcboundpeptidepredictions
AT danieltrademaker geometricdeeplearningimprovesgeneralizabilityofmhcboundpeptidepredictions
AT chiayulin geometricdeeplearningimprovesgeneralizabilityofmhcboundpeptidepredictions
AT sonjageorgievska geometricdeeplearningimprovesgeneralizabilityofmhcboundpeptidepredictions
AT nicolasrenaud geometricdeeplearningimprovesgeneralizabilityofmhcboundpeptidepredictions
AT amylkessler geometricdeeplearningimprovesgeneralizabilityofmhcboundpeptidepredictions
AT pablolopeztarifa geometricdeeplearningimprovesgeneralizabilityofmhcboundpeptidepredictions
AT sonjaibuschow geometricdeeplearningimprovesgeneralizabilityofmhcboundpeptidepredictions
AT erikbekkers geometricdeeplearningimprovesgeneralizabilityofmhcboundpeptidepredictions
AT licxue geometricdeeplearningimprovesgeneralizabilityofmhcboundpeptidepredictions