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
. 2021 Apr 23;24(4):102311.
doi: 10.1016/j.isci.2021.102311. Epub 2021 Mar 15.

Learning from HIV-1 to predict the immunogenicity of T cell epitopes in SARS-CoV-2

Ang Gao  1   2 , Zhilin Chen  3 , Assaf Amitai  1 , Julia Doelger  1 , Vamsee Mallajosyula  4 , Emily Sundquist  3 , Florencia Pereyra Segal  5 , Mary Carrington  3   6 , Mark M Davis  4   7   8 , Hendrik Streeck  9 , Arup K Chakraborty  1   2   3   10   11 , Boris Julg  3
Affiliations

Learning from HIV-1 to predict the immunogenicity of T cell epitopes in SARS-CoV-2

Ang Gao et al. iScience. .

Abstract

We describe a physics-based learning model for predicting the immunogenicity of cytotoxic T lymphocyte (CTL) epitopes derived from diverse pathogens including SARS-CoV-2. The model was trained and optimized on the relative immunodominance of CTL epitopes in human immunodeficiency virus infection. Its accuracy was tested against experimental data from patients with COVID-19. Our model predicts that only some SARS-CoV-2 epitopes predicted to bind to HLA molecules are immunogenic. The immunogenic CTL epitopes across all SARS-CoV-2 proteins are predicted to provide broad population coverage, but those from the SARS-CoV-2 spike protein alone are unlikely to do so. Our model also predicts that several immunogenic SARS-CoV-2 CTL epitopes are identical to seasonal coronaviruses circulating in the population and such cross-reactive CD8+ T cells can indeed be detected in prepandemic blood donors, suggesting that some level of CTL immunity against COVID-19 may be present in some individuals before SARS-CoV-2 infection.

Keywords: Artificial Intelligence; Immune Respons; Immunology; In Silico Biology.

PubMed Disclaimer

Conflict of interest statement

The authors declare no competing interests.

Figures

None
Graphical abstract
Figure 1
Figure 1
Performance of the A(s,M)-based classifier (A and B) The ROC curve of the binary classifier based on our model (red), compared with the immunogenicity prediction model developed by Calis et al. (2013) (green). (A) The ROC curves for the acute HIV infection group. The AUC of the red curve is 0.71. The AUC of the green curve is 0.57. (B) The ROC curves for the chronic HIV infection group. The AUC of the red curve is 0.66. The AUC of the green curve is 0.34. (C and D) The Spearman correlation coefficient between A(s,M) and p(s,M) on the test sets. The full model is compared with partial models where one or two of the three terms of A(s,M) are missing. (C) Corresponds to the acute group. (D) Corresponds to the chronic group. (E and F) The AUC of the binary classifier based on our model. The full model is compared with partial models where one or two of the three terms of A(s,M) are missing. (E) Corresponds to the acute group. (F) Corresponds to the chronic group. (G and H) The ROC curves for acute (G) and chronic infection (H) comparing our model with netMHCpan4.0, also summarized in figures (C–F) under “binding.” The error bars represent ± SD.
Figure 2
Figure 2
CTL responses elicited by predicted immunodominant SARS-CoV-2 optimal peptides IFN-γ T cell responses in 28 patients with COVID-19 were tested against 108 single 9-mer peptides (4–20 individual peptides per allele), of decreasing predicted immunodominance for the respective HLA alleles that were most frequently expressed in this cohort. (A) The mean amplitude of predicted immunogenic peptides (1–20), as determined by our model, across all tested alleles. Responses to N and S overlapping peptides pools were also included. (B) Frequency (%) of patients with the respective HLA allele with a response to a given peptide (1–20). (C) Mean magnitude of T cell responses (in SFC/1M PBMCs) against a given peptide (1–20) across all responding patients with the respective HLA allele. (D) Breadth of targeted peptides (each dot represents one patient) per HLA. The red highlighted dots show one patient who expressed four of the tested alleles and had a response to at least one peptide in each allele. (E) Distribution of targeted viral proteins for which T cell response was detected. 50% of the response was focused on orf1ab. (F) Most frequently (>50% of patients) targeted optimal peptides.
Figure 3
Figure 3
Comparison between model and ELISpot results for optimal peptides (A) Scatterplot of the fraction of positively responding patients graphed against the model prediction for the amplitude for each of the 108 experimentally tested peptides. The weighted Pearson correlation coefficient is 0.43. The number of patients per tested peptide is the weight for each peptide. (B) Peptides were grouped according to HLA types. Scatterplot of the fraction of positively responding patients for each HLA type graphed against the predicted mean amplitude for all peptides in this group. Shown is the weight Pearson correlation coefficient (0.82). The number of patients tested per HLA group is the weight. Error bars reflect the standard deviations of the fraction of positively responding patients within each group (each HLA type). Statistical significance was computed as described in methods.
Figure 4
Figure 4
Ex vivo detection of SARS-CoV-2 reactive CD8+ T cells in healthy, unexposed individuals (A) Representative FACS plots of A∗02:01/M1 and A∗02:01/Orf1ab tetramer staining. (B) Quantification of antigen-specific CD8+ T cells. Each donor (n = 8) data is represented by a dot, and the mean is indicated by the solid line. Statistical significance was determined by Wilcoxon matched-pairs signed rank test. (C) The mean percentages for each of the memory subsets defined based on the expression of CD45RA and CCR7 markers; naive (TN, CD45RA+CCR7+), central memory (TCM, CD45RA−CCR7+), effector memory (TEM, CD45RA−CCR7-), and effector memory expressing CD45RA (TEMRA, CD45RA+CCR7-) for Orf1ab+ CD8+ T cells in unexposed individuals.
See this image and copyright information in PMC

Update of

Similar articles

See all similar articles

Cited by

See all "Cited by" articles

References

    1. Abdul-Jawad S., Ondondo B., Van Hateren A., Gardner A., Elliott T., Korber B., Hanke T. Increased valency of conserved-mosaic vaccines enhances the breadth and depth of Epitope recognition. Mol. Ther. 2016;24:375–384. - PMC - PubMed
    1. Ahmed S.F., Quadeer A.A., Morales-Jimenez D., McKay M.R. Sub-dominant principal components inform new vaccine targets for HIV Gag. Bioinformatics. 2019;35:3884–3889. - PubMed
    1. Ahmed S.F., Quadeer A.A., McKay M.R. Preliminary identification of potential vaccine targets for the COVID-19 coronavirus (SARS-CoV-2) based on SARS-CoV immunological studies. Viruses. 2020;12:254. - PMC - PubMed
    1. Akst J. TheScientist; 2020. COVID-19 Vaccine Frontrunners.
    1. Barton J.P., Goonetilleke N., Butler T.C., Walker B.D., McMichael A.J., Chakraborty A.K. Relative rate and location of intra-host HIV evolution to evade cellular immunity are predictable. Nat. Commun. 2016;7:11660. - PMC - PubMed