Modeling SARS-CoV-2 substitution processes: predicting the next variant

We build statistical models to describe the substitution process 6 in the SARS-CoV-2 as a function of explanatory factors describing 7 the sequence, its function, and more. These models serve two differ- 8 ent purposes: first, to gain knowledge about the evolutionary biology 9 of the virus; and second, to predict future mutations in the virus, 10 in particular, non-synonymous amino acid substitutions creating new 11 variants. We use tens of thousands of publicly available SARS-CoV-2 12 sequences and consider tens of thousands of candidate models. 13 Through a careful validation process, we confirm that our chosen 14 models are indeed able to predict new amino acid substitutions: can- 15 didates ranked high by our model are eight times more likely to occur 16 than random amino acid changes. We also show that named variants 17 of interest were highly ranked by our models before their appearance, 18 emphasizing the value of our models for identifying likely variants of 19 interest and potentially utilizing this knowledge in vaccine design and 20 other aspects of the ongoing battle against COVID-19. 21 The intense community effort of SARS-CoV-2 sequencing has yielded a 22 wealth of information about the mutations that have occurred in the virus 23 it first in humans.