Twitter | Search | |
Ryan Seamus McGee
PhD student at UW. How does selection shape information in genomes and networks? Evo theory, info theory, comp sci, dynamics. Sports analytics. Feathered dinos.
69
Tweets
426
Following
4,578
Followers
Tweets
Ryan Seamus McGee retweeted
Lior Pachter Aug 11
Seemingly chaotic and in some cases incoherent recently announced college reopening statements prompted , , , & myself to analyze testing plans for the fall term. 1/10
Reply Retweet Like
Ryan Seamus McGee retweeted
Carl T. Bergstrom Aug 11
Replying to @RS_McGee
I'll be talking about a simple analytic approximation we've developed for the value of proactive testing in college settings and similar, and comparing that to the results from highly detailed stochastic SEIR network models that I've done with .
Reply Retweet Like
Ryan Seamus McGee Aug 9
Replying to @dshihlai @CT_Bergstrom
I think it would be reasonable to represent mask wearing as changes to the susceptibility and/or transmissibility parameter according to data on these. In the network models you can vary these on an individual basis and explore impact of different levels of mask compliance, etc.
Reply Retweet Like
Ryan Seamus McGee Aug 6
I can assist you with using our model to test your scenario. More documentation and tutorials for these kinds of scenarios will be posted shortly, and I can help you with your use case directly. We are also looking to run our model for TTI in university populations
Reply Retweet Like
Ryan Seamus McGee Aug 6
For testing at set intervals it seems to matter relatively little if you test all at once or stagger the intervals. But there may be benefits for surveillance and 2nd order interventions such as tracing and preemptive isolation to not have large spans where no tests are done.
Reply Retweet Like
Ryan Seamus McGee Aug 6
As an example, if most cases contribute few new infections and a small portion of cases contribute many (i.e. superspreading, gray distn), TTI relies on isolating the key spreaders, but if variation in transmissibility is low (red distn) all isolated cases help to reduce spread.
Reply Retweet Like
Ryan Seamus McGee Aug 6
The model allows for heterogeneity in all parameters, including transmissibility. Different distributions of individual transmissibility can have effects on spread and efficacy of testing, although these effects tend to be marginal relative to testing frequency and other factors.
Reply Retweet Like
Ryan Seamus McGee May 17
Here's one pre-COVID paper that looks at the effects of individual variation in R0. There are others, but this one includes figures for previous epidemics, including SARS1. SARS1 had a lot of variation, so SARS-Cov-2 may also, but the jury is out.
Reply Retweet Like
Ryan Seamus McGee May 17
But I don't think it's clear yet how much superspreading is attributable to large variation in transmissibility versus circumstances where individuals of typical transmissibility make contact with a lot of people at large gatherings, etc.
Reply Retweet Like
Ryan Seamus McGee May 17
Unfortunately, I haven't seen much data on this. Here's one study that assessed support of ranges for R0 and individual variation using data from early on in China. They and others argue that superspreading is important for capturing overall dynamics.
Reply Retweet Like
Ryan Seamus McGee May 11
Replying to @bansallab
We are thinking a lot about contact heterogeneity and global mixing beyond one’s local contacts in our network modeling. Would you be able to point me to references describing characteristics of these for respiratory infections as compared to others?
Reply Retweet Like
Ryan Seamus McGee retweeted
Shweta Bansal May 11
Replying to @bansallab
Infection spread is indeed not random, and heterogeneity should be on our mind. But let's not take our eyes off the prize of reducing overall connectivity and transmission with social distancing, handwashing and masks.
Reply Retweet Like
Ryan Seamus McGee retweeted
Carl T. Bergstrom May 8
Replying to @CT_Bergstrom
Context from one of the authors of one of the heterogeneity papers: thread.
Reply Retweet Like
Ryan Seamus McGee retweeted
Natalie E. Dean, PhD May 9
Seeing papers make the rounds that the herd immunity threshold may be much lower than the rough approximation 1-1/R0. Maybe, but let's slow down a minute. #1. There is still way too much uncertainty. #2. This does not qualitatively change our strategy. My comments. 1/
Reply Retweet Like
Ryan Seamus McGee retweeted
Carl T. Bergstrom May 7
From math professor , a nice explanation for why we want to keep the basic reproductive number as far below unity as possible.
Reply Retweet Like
Ryan Seamus McGee May 7
If we put in place extensive raptor detection and tracking systems, then we can reopen sooner than later.
Reply Retweet Like
Ryan Seamus McGee May 7
Verity et al. have good numbers for onset to discharge and onset to death (among other things): Also Zhou et al.:
Reply Retweet Like
Ryan Seamus McGee May 7
Figures for onset to hospitalization seem to be pretty variable (many human factors). Several studies have medians in the 7-11 day range (e.g., , ), but Lauer et al (linked above) has a median of 1.2 days with a huge 0-29 day range.
Reply Retweet Like
Ryan Seamus McGee retweeted
Muge Cevik May 4
A lot of discussion recently about transmission dynamics, most of which are extrapolated from viral loads & estimates. What does contact tracing/community testing data tell us about actual probability of transmission(infection rate), high risk environments/age? [thread]
Reply Retweet Like
Ryan Seamus McGee May 4
As I have pointed out in our previous communication, we use precisely this matrix as one of several sources for our calibration of contact networks in our modeling work. We also incorporate information from contact surveys in the setting of COVID, demographic census data, etc.
Reply Retweet Like