I wanted to address the hypothesis put forward in Korber et al (biorxiv.org/content/10.110…) that the mutation in spike protein D614G causes an increase in transmissibility of SARS-CoV-2 virus. I find this hypothesis to be plausible, but far from proven. 1/16
I've been watching D614G closely as mutations in spike protein deserve added attention due to spike's role in binding to the human ACE2 receptor. 2/16
This D614G mutation occurred in the transmission chain that initially seeded the European outbreak in ~Jan 2020. Almost all viruses possessing this mutation descend from this initial introduction into Europe. 3/16 
European viruses are enriched for D614G because of a founder effect in which the initial introduction included this mutation. If we look at the geographic distribution of D vs G in sequenced viruses we see Europe with G, Asia largely D and a mix in the US and Australia. 4/16
The primary finding of Korber et al is that D614G appears to be increasing in frequency over time in sequenced SARS-CoV-2 genomes. I strongly caution against interpretation of selective effects in the global frequency of D614G. 5/16
Its global frequency is heavily confounded with epidemiological circumstance, ie perhaps G is prevalent because it got lucky in the European introduction. However, regional frequencies should be more robust to this confounding (although not perfectly so). 6/16
This figure recapitulates the Korber et al findings using @nextstrain. Here, I've shown states in the US and Australia with more than 70 sequences available and our estimate of frequency of D (green) vs G (yellow) from March 1 to April 15. 7/16
You can see that in every case the frequency of G increases in the course of these 45 days. 8/16
It's still very possible that this pattern could emerge from repeated introductions from Europe / NYC in the US spreading the G variant and multiple introductions from Europe spreading G in Australia. 9/16
The alternative explanation put forth by Korber et al is that this pattern is due to the G variant being more transmissible. I think this is possible, but it's difficult to distinguish between these hypotheses with this frequency data alone. 10/16
Additional evidence for the hypothesis of a functional effect of D614G comes from Korber et al's observation that the G variant has lower cycle threshold (Ct) value in clinical specimens from Sheffield. This indicates a possible higher viral load in these individuals. 11/16
Thanks to work by @wcassias and @pavitrarc, we see this difference in Ct between D and G replicated in @UWVirology specimens. Preliminary analysis here: github.com/blab/ncov-D614G. 12/16
There are confounders to worry about here as well (primarily time from symptom onset to specimen collection), but I believe replication in two study populations is suggestive of an effect of D vs G on Ct value and possibly viral load. 13/16
Both Korber et al and our analysis show no measurable effect on patient outcome. Hence, the hypothesis at this point is entirely in terms of transmissibility rather than severity. 14/16
Overall, I would refer everyone to @edyong209's piece on handling uncertainty during the pandemic (theatlantic.com/health/archive…). I don't agree with takes that there is "no evidence" that G is more transmissible. There is some evidence, but it's far from conclusive. 15/16
We have to live with this uncertainty over the functional impact of D614G while more data is gathered. We need:
1. Cell culture studies to demonstrate effect in vitro
2. Further clinical comparisons between patients with D vs G
3. More careful epidemiological analysis
16/16
1. Cell culture studies to demonstrate effect in vitro
2. Further clinical comparisons between patients with D vs G
3. More careful epidemiological analysis
16/16
