Thread by @trvrb

An update on genomic surveillance in the US and spread of the Delta variant (PANGO lineage B.1.617.2, Nextstrain clade 21A). At this point, 95% of viruses circulating in the US are "variant" viruses that have been designated as "Variant of Concern" or "Variant of Interest". 1/12

This update mirrors how I was looking at the rise of P.1 across the US in May. 2/12

Here, we can look at frequencies of different variant lineages through time and across states where it's clear that variant viruses and in particular B.1.617.2 viruses are continuing to increase in frequency. 3/12

I'm plotting frequencies on a logit axis as it makes logistic growth appear as a straight line in logit space. This makes it easier to visually compare growth from say 1% to 2% vs 10% to 20%, where both are doublings of frequency. 4/12

We can partition cases reported by @CDCgov based on variant frequency from @GISAID data. Doing so gives the following picture across the US, where it's clear that although B.1.1.7 and P.1 have begun decline, B.1.617.2 is increasing in absolute numbers. 5/12

We see a similar picture across states, again with B.1.617.2 at lower incidence but rising. 6/12

There will be a necessary lag as genomic data comes in (I'm comfortable with estimates 18 days back for US states shown here), but we generally want to know current estimate of variant-specific Rt to assess whether variant sub-epidemics are growing or subsiding. 7/12

Here, @marlinfiggins has come up with a clever Bayesian approach that models the process of observing daily case counts alongside the observation process describing how sequenced genomes are distributed across variants. 8/12

This gives a quantitative estimate of daily Rt across variants and across states. The main assumption is that genomes deposited into GISAID are a decent surveillance sample of the viral population. Due to CDC NS3 surveillance efforts, I believe this is reasonable for the US. 9/12

Here, we see daily Rt estimates for each major variant across states, where it's clear that Rt for non-variant viruses has been consistently below 1, while Rt for B.1.1.7, P.1 and B.1.526 has been drifting downwards since March and is now generally below 1 across states. 10/12

However, Rt for B.1.617.2 is significantly higher and well above 1 in most of the 16 states analyzed. This figure shows Rt at the final timepoint estimated (May 31). 11/12

We know from @PHE_uk that B.1.617.2 is primarily transmitting through the unvaccinated (or partially vaccinated) population. It's difficult to predict the size of a B.1.617.2 epidemic we'll see, but I would expect at least some correlation with regional vaccine coverage. 12/12