Paper on the Spread of COVID-19 in the Boston Area

[T_aquaticus]

Congrats to @glipsnort for his recent publication in Science on the spread of SARS-CoV-2 in the Boston area. There are some really fascinating findings on how the virus has spread:

Genome data reveal that the impact of the conference was far larger than the approximately 100 cases directly associated with the event. Using state-reported case counts, we estimate that by the end of the study period, approximately 50,000 diagnosed cases [44,000 - 56,000] in the US resulted from conference-associated viruses; of these, 46% [40.4 - 51.8%] were in Massachusetts.
https://science.sciencemag.org/content/early/2020/12/09/science.abe3261

Note: Dr. Schaffner has very limited time to spend on these threads, so please don’t expect quick replies, or any at all. We will of course appreciate any comments he has time for.

[Michael_Okoko]

Any reason for this near-fixation of the “D614G” amino acid polymorphism?

Consistent with a larger global trend ( 13 , 14 ), we observed a rise in frequency of viruses harboring the D614G amino acid polymorphism in the Spike protein, conferred by a SNP at nucleotide 23,403 in the Wuhan reference strain, which rose to near-fixation in our dataset by the end of the study period (Fig. 3C) and is present in all of the dominant lineages.

[glipsnort (Steve Schaffner)]

@Michael_Okoko

Any reason for this near-fixation of the “D614G” amino acid polymorphism?

It’s very likely under positive selection: it makes the virus more transmissible (based on a bunch of lab studies).

[Michael_Okoko]

Yeah based on those lab studies cited in the article, its likely the increased infectivity due to the polymorphism has placed it under positive selection.

An aside. Any idea whether the Pfizer or Moderna mRNA vaccines encode for that particular polymorphism, since it has persisted within SARS-CoV-2 lineages and would probably be a good vaccine target?

[glipsnort (Steve Schaffner)]

@Michael_Okoko

An aside. Any idea whether the Pfizer or Moderna mRNA vaccines encode for that particular polymorphism, since it has persisted within SARS-CoV-2 lineages and would probably be a good vaccine target?

I know the Moderna vaccine was designed against the unmutated form, and I’d bet the Pfizer one was as well. There’s good evidence that the mutation won’t affect vaccine efficacy, though.

[Michael_Okoko]

@glipsnort

There’s good evidence that the mutation won’t affect vaccine efficacy, though.

I guess one can quite safely draw that conclusion based on the impressive results from the Moderna and Pfizer interim vaccine data analyses.

In general this was excellent research conducted by excellent scientists. Kudos to you all.

[chris_doesdna2018 (Chris Hughes)]

Any evidence for spread along T routes?

Maybe a Novembre-style PCA if the location data are granular enough.

[glipsnort (Steve Schaffner)]

We generally can’t analyze data at a fine geographic scale, even if the information has been collected, since cases can become identifiable if you specify too precisely where they were, and that’s a no-no. Dealing with all the IRB issues is a major hassle in this kind of work.

Having seen the data, my impression is that the initial spread from the conference was primarily in two directions: into poorer parts of the immediate Boston area, and into some of the wealthier suburbs west of the city. It’s easy to imagine that this reflected the homes of conference/hotel staff and of attendees, respectively – but it’s also quite possible it was meaningless noise. There was a great deal of rapid mixing of different lineages, after all.

ETA: I’ve thought some in the past about the use of PCA with non-recombining viruses, and I don’t think it really does the job here. PCA works with humans and such because recombination turns variants into multiple independent variables/dimensions. Without recombination, though, the data is inherently hierarchical, and the right approach to apply is not obvious.