Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
[Preprint]. 2023 Dec 14:2023.11.29.569310.
doi: 10.1101/2023.11.29.569310.

MHC-Fine: Fine-tuned AlphaFold for Precise MHC-Peptide Complex Prediction

Ernest Glukhov  1   2 , Dmytro Kalitin  1   2 , Darya Stepanenko  1   2 , Yimin Zhu  3   2 , Thu Nguyen  3   2 , George Jones  1   2 , Carlos Simmerling  4   2 , Julie C Mitchell  5 , Sandor Vajda  6 , Ken A Dill  4   7   2 , Dzmitry Padhorny  2 , Dima Kozakov  1   2
Affiliations

MHC-Fine: Fine-tuned AlphaFold for Precise MHC-Peptide Complex Prediction

Ernest Glukhov et al. bioRxiv. .

Update in

Abstract

The precise prediction of Major Histocompatibility Complex (MHC)-peptide complex structures is pivotal for understanding cellular immune responses and advancing vaccine design. In this study, we enhanced AlphaFold's capabilities by fine-tuning it with a specialized dataset comprised by exclusively high-resolution MHC-peptide crystal structures. This tailored approach aimed to address the generalist nature of AlphaFold's original training, which, while broad-ranging, lacked the granularity necessary for the high-precision demands of MHC-peptide interaction prediction. A comparative analysis was conducted against the homology-modeling-based method Pandora [13], as well as the AlphaFold multimer model [8]. Our results demonstrate that our fine-tuned model outperforms both in terms of RMSD (median value is 0.65 Å) but also provides enhanced predicted lDDT scores, offering a more reliable assessment of the predicted structures. These advances have substantial implications for computational immunology, potentially accelerating the development of novel therapeutics and vaccines by providing a more precise computational lens through which to view MHC-peptide interactions.

Keywords: AlphaFold; Cellular immune system; Deep learning; Fine-tuning; MHC Class I.

PubMed Disclaimer

Figures

Fig. 1.
Fig. 1.
Comparative Analysis of Prediction Accuracy for MHC-Peptide Complexes
Fig. 2.
Fig. 2.
Comparative Visualization of pMHC Complex Prediction Accuracy. True peptide structure in red, Pandora model in orange, MHC-Fine (fine-tuned AlphaFold) in blue. (A) High precision of MHC-Fine (RMSD: 0.25 Å) versus Pandora (RMSD: 1.44 Å) for PDB ID: 6vb3; (B) Moderate accuracy for both models (MHC-Fine RMSD: 1.23 Å; Pandora RMSD: 1.19 Å) for PDB ID: 7n2o; (C) Significant deviations in predictions (MHC-Fine RMSD: 3.73 Å; Pandora RMSD: 4.30 Å), indicating areas for enhancement, for PDB ID: 7mj7.
Fig. 3.
Fig. 3.
Error Distribution for Predicted lDDT Values. The red distribution represents the original AlphaFold model with a standard deviation of 9.2, while the blue distribution showcases our fine-tuned AlphaFold model with a reduced standard deviation of 4.6.
Fig. 4.
Fig. 4.
Confidence Prediction: All samples are grouped by their predicted Local Distance Difference Test (pLDDT) interval. The chart displays the percentage of samples falling within each specific pLDDT range.
See this image and copyright information in PMC

References

    1. Abella Jayvee, Antunes Dinler, Clementi Cecilia, and Kavraki Lydia. Ape-gen: A fast method for generating ensembles of bound peptide-mhc conformations. Molecules, 24(5):881, Mar 2019. - PMC - PubMed
    1. Ahdritz Gustaf, Bouatta Nazim, Kadyan Sachin, Xia Qinghui, Gerecke William, Tim O’Donnell Daniel Berenberg, Fisk Ian, Zanichelli Niccoló, Zhang Bo, Nowaczynski Arkadiusz, Wang Bei, Stepniewska-Dziubinska Marta M., Zhang Shang, Ojewole Adegoke A., Guney Murat Efe, Biderman Stella, Watkins Andrew M., Ra Stephen, Lorenzo Pablo Ribalta, Nivon Lucas, Weitzner Brian D., Andrew Ban Yih-En, Sorger Peter K., Mostaque Emad, Zhang Zhao, Bonneau Richard, and Alquraishi Mohammed. Openfold: Retraining alphafold2 yields new insights into its learning mechanisms and capacity for generalization. bioRxiv, 2023. - PMC - PubMed
    1. Bradbury James, Frostig Roy, Hawkins Peter, Johnson Matthew James, Leary Chris, Maclaurin Dougal, Necula George, Paszke Adam, VanderPlas Jake, Wanderman-Milne Skye, and Zhang Qiao. JAX: composable transformations of Python+NumPy programs, 2018.
    1. Bradley Philip. Structure-based prediction of t cell receptor:peptide-mhc interactions. eLife, 12:e82813, jan 2023. - PMC - PubMed
    1. Deng Lu and Mariuzza Roy A.. Recognition of self-peptide–mhc complexes by autoimmune t-cell receptors. Trends in Biochemical Sciences, 32(11):500–508, 2007. - PMC - PubMed

Publication types