Genetic mismatch explains sizable variation of COVID-19 vaccine efficacy in clinical trials

ABSTRACT

Timely evaluation of the protective effects of COVID-19 vaccines is challenging but urgently needed to inform the pandemic control planning. Based on vaccine efficacy/effectiveness (VE) data of 11 vaccine products and 297,055 SARS-CoV-2 sequences collected in 20 regions, we analyzed the relationship between genetic mismatch of circulating viruses against the vaccine strain and VE. Variations from technology platforms are controlled by a mixed-effects model. We found that the genetic mismatch measured on the RBD is highly predictive for vaccine protection and accounted for 72.0% (p-value < 0.01) of the VE change. The NTD and S protein also demonstrate significant but weaker per amino acid substitution association with VE (p-values < 0.01). The model is applied to predict vaccine protection of existing vaccines against new genetic variants and is validated by independent cohort studies. The estimated VE against the delta variant is 79.3% (95% prediction interval 67.0 - 92.1) using the mRNA platform, and an independent survey reported a close match of 83.0%; against the beta variant (B.1.351) the predicted VE is 53.8% (95% prediction interval 39.9 - 67.4) using the viral-vector vaccines, and an observational study reported a close match of 48.0%. Genetic mismatch provides an accurate prediction for vaccine protection and offers a rapid evaluation method against novel variants to facilitate vaccine deployment and public health responses.