Human indels as predictors of antibody responses to COVID-19 vaccines

Abstract

Vaccine efficacy varies significantly among adults. This variability underlies the limitation of a one-size-fits-all vaccination strategy and the need for more personalized approaches. We investigated factors influencing inter-individual variability in antibody responses to COVID-19 mRNA vaccine among adults. Neutralizing antibody (nAb) levels after the first vaccine dose were associated with infection outcomes within 1 year after vaccination, suggesting their potential as a correlate of protection. Age, sex, and Chinese ethnicity were associated with nAb and anti-spike protein antibody levels. Two indels located at chr1:31433042 and chr15:76311269 showed significant association with antibody responses. Leveraging these host factors, we developed a Random Forest model that predicted vaccine-induced antibody responses with 72.7% accuracy for mRNA vaccine and 76.9% for the Sinopharm COVID-19 inactivated virus vaccine. These findings support predictive modeling as a tool to identify individuals at risk of low vaccine responses, enabling more targeted and effective vaccination strategies.

Keywords: Public health; Virology.

Graphical abstract

Figure 1

Variation in vaccine-induced antibody responses among individuals (A) Illustration of the study design: 328 individuals received two doses of BNT162b2 vaccines on days 0 and 21. Blood samples were collected on days 0, 21, 90, and 180 to measure their nAb or anti-spike protein IgG levels. (B) Comparison of variance in nAb levels on days 0, 21, 90, and 180 (% inhibition). (C) Comparison of variance in anti-spike protein IgG levels on days 0, 21, 90, and 180 (% binding). (D) Associations between infection outcomes within 1 year of vaccination and nAb levels on days 21, 90, and 180 (logistic regression, p value = 0.0013, 0.0038, and 0.0030 for days 21, 90, and 180, respectively). (E) Associations between infection outcomes within 1 year of vaccination and anti-spike protein IgG levels on days 21, 90, and 180 (logistic regression, p value = 6.7 × 10⁻⁵, 0.013, and 0.69 for days 21, 90, and 180, respectively). For (D) and (E), the boxplots show the median and interquartile range of the data. Ab, antibody; WES, whole-exome sequencing; nAb, neutralizing antibody; S IgG, anti-SARS-CoV-2 spike protein IgG.

Figure 2

Effects of age, sex, and ethnicity on neutralizing antibody and anti-spike protein antibody responses (A) Associations between age and nAb or anti-spike protein IgG levels on day 21 (linear regression, p value <2.2 × 10⁻¹⁶ for both, R² = 0.28 and 0.24 for nAb and anti-spike protein IgG levels, respectively, n = 328). Linear regression lines are plotted in green. (B) Associations between sex and nAb or anti-spike protein IgG levels on day 21 (linear regression, p value = 6.0 × 10⁻⁹ and 7.3 × 10⁻⁸, R² = 0.10 and 0.08, respectively, n = 328). (C) Associations between Chinese ethnicity with nAb or anti-spike protein IgG levels on day 21 (linear regression, p value = 3.2 × 10⁻⁸ and 5.2 × 10⁻⁶ for nAb and anti-spike protein IgG levels, respectively, n = 328). The boxplots show the median and interquartile range of the data. (D) Left: distribution of nAb levels on day 21 with a multiple linear regression plane: nAb levels on day 21 ∼ age + sex, in a 3D plot (adjusted R² = 0.30). Right: distribution of anti-spike protein IgG levels on day 21 with a multiple linear regression plane: anti-spike protein IgG levels on day 21 ∼ age + sex, in a 3D plot (adjusted R² = 0.26). n = 328. nAb, neutralizing antibody; S IgG, anti-spike protein IgG.

Figure 3

Effects of indels on antibody response to COVID-19 mRNA vaccines (A) Associations of nAb levels on day 21 with indels 1 and 2. Indel 1 REF and ALT alleles are A and ACAG; indel 2 REF and ALT alleles are T and TC. (B) Association of indel 1 with memory B cell counts on day 360 (as determined with an additive model, p value = 0.026). (C) Association between anti-spike protein IgG levels on day 21 and indel 3. Indel 3 REF and ALT alleles are CA and C. (D) Associations of indel 3 with pseudovirus-neutralization activities for Delta or Omicron strains on day 90 (p value = 0.026 or 0.009, respectively). (E) Pie charts of allele frequencies of indels 1–3 in the Singapore population in our study and different global populations in the All of Us program (REF allele in blue, ALT allele in red). For (A), (B), (C), and (D), the boxplots show the median and interquartile range of the data. REF, reference; ALT, alternative.

Figure 4

Prediction of vaccine-induced antibody responses (A) Comparison of nAb level distributions on days 0, 21, 90, and 180 in the Sinopharm inactivated virus and BNT162b2 mRNA vaccine cohorts. (B) Violin plot representation of the data shown in (A). The boxplots show the median and interquartile range of the data. (C) Effect of indel 1–3 on nAb level on day 90 in individuals in the Sinopharm cohort. No significant associations between indels and nAb level on day 90, as determined with an additive model incorporating age, sex, and Chinese ethnicity as covariates (p value = 0.730 for indel 1, p value = 0.288 for indel 2, and p value = 0.218 for indel 3). The boxplots show the median and interquartile range of the data. (D) Performance of Random Forest models for predicting vaccine-induced antibody responses. This panel presents receiver operating characteristic (ROC) curves and corresponding area under the curve (AUC) values for six Random Forest models. These models were trained on the BNT162b2 vaccine cohort’s training set and evaluated on its test set. Each model incorporates a different combination of variables to demonstrate the impact of additional host factors on prediction performance. The variables included are M1: age groups; M2: age groups + sex; M3: age groups + sex + Chinese ethnicity; M4: age groups + sex + Chinese ethnicity + indel 1; M5: age groups + sex + Chinese ethnicity + indel 3; M6: age groups + sex + Chinese ethnicity + indel 1 + indel 3 (Age, age groups; Chi, Chinese ethnicity; In1, indel 1; In3, indel 3; ASC, age groups + sex + Chinese ethnicity).