-
Notifications
You must be signed in to change notification settings - Fork 31
BioStatistics
Creating SEIR, infection, deducing the impact of policies (or lack thereof) on infection data, prediction of future states, statistical inference from published data...
There is a #biostats channel on the Slack group (check out the [email protected] group for the invitation link). During the BioHackathon, we'll update this section.
We've setup a dedicated GitHub organization here. For a detailed list of all tasks, code and resources, please go there. This page will be updated for the main points only.
- 1st e-meeting Sunday, April 5th @ 16:00 CEST, using zoom. Registration link here - in order to get the connection details. Feel free to ping Fotis or Janne on the slack channel if there are any issues connecting. For live notes/minutes, check the HackMD link here.
- 2nd e-meeting Monday, April 6th @ 15:00 CEST, using zoom. Registration link here. This will be our second, brief (~30') BioStats call. The agenda will be to review the tasks and subtasks listed here.
Please check out the Datasets and Tools page.
Any new resources you might have in mind, please add them there directly.
- (new, to be added) Contagiousness of COVID-19 Part I: Improvements of Mathematical Fitting
- (new, to be added)Assessment for the seasonality of Covid-19 should focus on ultraviolet radiation and not 'warmer days'
- Fotis Psomopoulos (co-coordinating)
- Janne Solanpää (co-coordinating)
- JJ Merelo
- Bonface Munyoki
- Felizitas Eichner
- Noushin Nabavi
- Maciej Bak
- Ceci Valenzuela
- Arvon Clemons II
- Saeed Omidi
- Paul Lassmann-Klee
- Sara Vilella
- Gonzalo Colmenarejo
- Thanasis Vergoulis
- Kostis Zagganas
- Laura Zaragoza Infante
- Festus Nyasimi
- Nikolaos Pechlivanis
- Maria Tsagiopoulou
- Güney Işık Tombak
- İlkay Civelek
Left here for reference - refer to the covid19-bh-biostats GitHub repo for details.
Implement a SEIR model (similar to here) as a function, and tested against the various data points we have (e.g. vs country or other geographical split). This can be consequently tested against the different policies in place (of which we know at least), trying to identify "interesting" cases. An interesting case could be characterized as the case of two geographical locations that implemented similar measures/policies across time, but have significantly different outcome (i.e. the SEIR model parameters do not align). For these cases, the corresponding entries of the COVID-19 phylogeny can be tested for molecular markers, as they could be of potential value. Also try to correlate the SEIR parameters to environmental data (to confirm the "evidence that a 1C increase in local temperature reduces transmission by 13%" - article here )
Since public life can only stand still for a certain time, one could model different scenarios such as suggested by Uri Allon. Scenarios could include various combinations of work-week vs. lock-down (2-3, 3-2 etc.), which is either homogenous (all follow the same rhythm) or heterogeneous (rhythm is shifted) in a society. Another potential scenario is a regular switch such as a one-week-work - one-week-lock-down scenario.
Having online SEIR models at hand is one thing, but "feeding" them with the right assumptions is another. One could establish a repository with an overview of the parameters needed combined with all identified literature sources (if possible stratified for countries). This could serve as a solid starting point e.g. for epidemiologists currently building models on regional level.
Create reference libraries that read the Johns Hopkins repository and put data in the format of a particular language.
- There is now an R package for the data colated by Johns Hopkins