Share this page!

Enter URL or ID to Unroll

You can paste full URL like: https://x.com/threadreaderapp/status/1644127596119195649
or just the ID like: 1644127596119195649

How to get URL link on X (Twitter) App

  1. On the Twitter thread, click on or icon on the bottom
  2. Click again on or Share Via icon
  3. Click on Copy Link to Tweet
  4. Paste it above and click "Unroll Thread"!
  5. More info at Twitter Help
Tom Wenseleers now @twenseleers.bsky.social Profile picture
Tom Wenseleers now @twenseleers.bsky.social
@TWenseleers
Views my own & tweeting in my own name Also at @twenseleers.bsky.social
Mar 22, 2023 • 7 tweets • 3 min read
So there was a bit of talk if the high SARS-Cov2 positivity found in the SW corner of the Huanan Market could be an artefact of a sampling bias. It is not. OK, a basic density plot of SARS-Cov2 positive samples is confounded by sampling intensity...🧵 But before the authors also calculate the relative risk to encounter SARS-CoV2 positive samples across the market & that gave the same conclusion...
Mar 22, 2023 • 4 tweets • 2 min read
Ha the good old "Western scientists re-analysing data from LMIC (or in this case even of an UMIC) is neocolonialist". Bollocks. It took over a year for Gao et al to share their fastq files... That's a reasonable referee request... Especially since the original preprint included statements like a sample from a feather removal machine only containing human DNA and stuff like that.
Mar 22, 2023 • 7 tweets • 4 min read
"The reviewer, who asked not to be named, noted that the paper also includes an analysis of a sample from a defeathering machine at the market that found only human DNA. ‘Either only humans were defeathered or the analysis was wrong—pick your favorite’" science.org/content/articl… Yes, this @sciencecohen article is from the 18th of August 2022, and is discussing the Gao et al. / Liu et al. preprint, researchsquare.com/article/rs-137…. When received by Nature, the referees said this analysis could not be published without providing the raw data.
Mar 21, 2023 • 6 tweets • 4 min read
And here the much awaited report of @flodebarre, @MichaelWorobey & @K_G_Andersen, based on a re-analysis of the Chinese CDC GISAID data (that recently were taken down again)! Seems pretty compelling to me! zenodo.org/record/7754299… Image The evidence should even satisfy eternal contrarian @BallouxFrancois: there is a positive association between samples with high Sars-Cov2 read content & the ones containing raccoon dog reads. What more could one wish? Image
Feb 10, 2023 • 13 tweets • 6 min read
Amazing how SARS-CoV2 keeps on evolving. Since end of Jan Belgium started its 10th wave now - the latest one presumably driven by Omicron subvariants CH.1.1+XBB.1.5+XBB.1.9.1. Slower initial rate of increase in hosps for Covid than for BQ.1 Dec wave predicts smaller wave though. When I say 10 I'm not counting 2 smaller waves in fact. So some might put it at 12. Two weeks ago this fit already detected a changepoint (ie growth in infections starting again).
Feb 6, 2023 • 5 tweets • 2 min read
In Indonesia, Omicron XBB.1.9.1 (yellow) now has a >50% share - one country at least that will skip an Omicron XBB.1.5 wave (cyan) - not because it didn't have an advantage there, but because another even fitter variant emerged. Second is Singapore, where XBB.1.9.1 has 15% share. Also in England XBB.1.9.1 is taking off fast...
Feb 2, 2023 • 9 tweets • 4 min read
Covid hospitalisations on the rise again in England. Not entirely unexpected, given that share of Omicron variants CH.1.1+XBB.1.5 just reached a share >50%. Quite predictable pattern now, with small Covid wavelets every 3-4 months, luckily with decreasing amplitude. Likewise, decline in Covid hospitalisations has also stopped again in the Netherlands & Germany...
Feb 1, 2023 • 5 tweets • 2 min read
Just tried the OpenAI AI Text Classifier platform.openai.com/ai-text-classi… on ChatGPT output, and I can confidently say that it's totally unreliable: for my more academic oriented prompts & ChatGPT outputs it concluded in 8/10 cases that the ChatGPT text was "unlikely AI-generated"... And in 2/10 cases the result was "unclear if it is AI-generated". I guess my prompts are sufficiently nerdy that it invariably concludes that a scientist must have written the ChatGPT created pieces. Just forget about these AI detectors ever being reliable...
Jan 12, 2023 • 7 tweets • 2 min read
New #ChatGPT co-authored manuscript up on medRXiv. Not particularly impartial to ask ChatGPT to evaluate itself though. :-) medrxiv.org/content/10.110… And now also a ChatGPT coauthored paper out on Pubmed... pubmed.ncbi.nlm.nih.gov/36549229/
Jan 9, 2023 • 11 tweets • 5 min read
Haven't seen many formal growth rate advantages of US variant XBB.1.5*, besides that of @MoritzGerstung below, so here my attempt using a multinomial spline model. Clear that advantage is at least as large as the one that BQ.1* had, both in US & Europe. Here is the current daily growth rate advantage (in % per day) of different variants relative to the dominant type BQ.1*. XBB.1.5* is the clear winner & this initial advantage is at least as large than the one that BQ.1* had at an equivalent share in the population.
Jan 7, 2023 • 5 tweets • 3 min read
For aspiring corona centrists: just take the average of the CDC nowcast of last week (XBB.1.5 at 40%) & that of @BallouxFrancois, claiming that XBB.1.5 was at <10% in the US, and you'll be about right... The latest SGTF data show XBB.1.5 is at ca 30% the first week of Jan 2023, which is in fact in line with this week's CDC nowcast (but the SGTF data is much more accurate, as it's actual data, not a nowcast, and based on a much larger sample size). observablehq.com/@andy-bloch/wa…
Jan 4, 2023 • 8 tweets • 3 min read
Just discovered the gpuMagic R package, which provides an easy way to compile R functions (using a number of basic functions) to a much faster OpenCL GPU version. Here a quick example.🧵 #rstats github.com/Jiefei-Wang/gp… Here I define a pure R Mandelbrot fractal function (en.wikipedia.org/wiki/Mandelbro…) that could be called for all indices i of pixels from 1 to width*height.
Jan 4, 2023 • 4 tweets • 2 min read
Interesting thread of @c_althaus concluding that by looking at the asymptotic behaviour of the relative cumulative excess mortality it would appear that Covid could keep on increasing overall mortality by 5-10% in Western European countries in the near term. In Belgium baseline expected mortality (in absence of Covid) was now ca. 110K deaths/y. Covid being projecting to add ca. 5 500-11 000 deaths/year is a tad high I think - but possible that death toll of ca 5K Covid deaths we had this year will not massively improve in the future.
Jan 4, 2023 • 6 tweets • 3 min read
"EU agency says COVID checks on China travelers 'unjustified'": I agree - the epidemiological situation in China, tragic as it is, will not make a material difference to the epidemic in Western countries. p.dw.com/p/4LWgK?maca=e… The top circulating lineages in China - BA.5.2 & BF.7 - are ones that in the rest of the world are already long in decline.
Jan 3, 2023 • 12 tweets • 5 min read
Balloux claims the CDC's projection that Omicron XBB.1.5 would already make up 40% of all sequenced genomes in the US is not correct. I believe he's mistaken though, as he is not using the Nextclade Pangolin XBB.1.5* classification but rather the default GISAID pangolin one,... ...which misclassifies most XBB.1.5* as XBB.1 or BA.2. The actual GISAID data, using Nextclade Pangolin XBB.1.5% show it is indeed at 40% already in the US & at 56% in New York state. It has a logistic growth rate of 0.14 per day, which is very fast. cov-spectrum.org/explore/United…
Dec 29, 2022 • 12 tweets • 5 min read
"Why did China relax its Covid policy – and should we be worried?" - new article by @NicolaKSDavis in The Guardian with a bit of input from myself. 🧵theguardian.com/world/2022/dec… For the record - while the situation in China is a bit tragic - the main mistake was for many elderly not to get vaccinated. But ultimately that was their own choice. And a zero Covid policy was never going to be sustainable in the long run... Vaccination was the only way out.
Dec 2, 2022 • 6 tweets • 2 min read
Update to post below (which had a bit of a downward bias because of a strike in one of the testing labs): so in the meantime BQ.1.1 did cause cases in France to start rising again & same pattern seen in Belgium & Denmark. Hospitalisation figures have also been on the rise again in France for a little while now.
Dec 1, 2022 • 4 tweets • 2 min read
Is er nu werkelijk iemand die de huidige boerentoren vindt passen in de skyline van het historische centrum van Antwerpen? Image Toen hij gebouwd werd was zo wat iedereen er tegen. In de jaren '70 werd hij door KBC ei zo na gesloopt & het art deco interieur gestript. Als men hem als erfgoed had willen beschermen had men dat duidelijk toen moeten doen. gva.be/cnt/dmf2022112… Image
Nov 19, 2022 • 8 tweets • 3 min read
Als je het mij vraagt zaten predicties hospitalisaties met Covid er onder de gerealiseerde vaccinatiegraad (77% uptake 2de booster bij 65+ tov 1ste booster, tussen 2 gemodelleerde scenario's in dus) niet zo ver naast: oranje lijn: voorspelde amplitude: iets lager dan de vorige.🧵 Eigenlijke data: golf iets kleiner dan de vorige. Wat is het probleem eigenlijk? Dat België een succesvolle vaccinatiecampagne heeft opgezet die heel wat hospitalisaties & sterfte heeft kunnen voorkomen?
Oct 31, 2022 • 5 tweets • 2 min read
Impact of new variants always hard to predict but encouraged by hospitalisation waves now showing an exponential decline (linear decrease on a log scale). Unless another Omicron-level variant would appear, I'd be tempted to think we should be able to extrapolate this pattern.🧵 Same pattern for confirmed Covid deaths: over the long term an exponential decline.
Oct 23, 2022 • 6 tweets • 3 min read
New update from @yunlong_cao: confirms extreme immune evasion by Omicron BQ.1.1 and XBB: near zero neutralizing antibody immunity 7.5 months after 3xSinovac vaccination+BA.1 breakthrough infection. Higher neutralizing titers expected after mRNA vaccination & bivalent booster. Note: this does not mean XBB & BQ.1.1 will cause a massive wave again. Rather, based on observed growth rate advantage (ca. 0.11/day) & expected impact on R value I would - in terms of infections - expect a wave of similar magnitude than BA.5. (R code at github.com/tomwenseleers/…)