Michael Swift

@arjunrajlab the lack of clarity of thought here is astounding. I can't believe we send our best work there. thank goodness for bioarxiv

The most blatantly self-serving article I've seen in recent memory. The reason for loss of public trust in science is… that we don't treat editors with enough respect? Yes, editors should be respected. No, that has nothing to do with loss of trust. 1/2 science.org/doi/10.1126/sc…

@whstancil I haven't seen any nazi stuff in the past 6 months on this site. it's you

I mean, it’s not just me, right? The Nazis have taken over this site in the last two weeks? Or maybe it is just me?

@GeoffStanley9 @5gum no because why no TW?

@carlo_cervia @ScienceMagazine doi.org/10.17632/dvf6y…

@carlo_cervia @ScienceMagazine the doi link to the data (ref 101) does not exist

Our work on #LongCovid 🩸 is now out in @ScienceMagazine! We found increased complement activation and markers of thromboinflammation in patients with active Long Covid. 1/ doi.org/10.1126/scienc…

@GeoffStanley9 found a job for u

@tangming2005 @bcorrie @bottlefeeder1 @ireceptorgw I found celltypist was usually quite good for immune cells

@bcorrie @bottlefeeder1 @ireceptorgw I agree with this approach. Automatic annotation to get first impression of the data and then manually dive deeper

single-cell folks, what's the cell annotation tool do you recommend for immune cells and tumor cells? singleR? Seurat based label transfer? or sc-best-practices.org/cellular_struc… I still find myself using marker genes manually quite often

@WarrenCManuel @ChatGPTapp capitalization is for boomers

@ChatGPTapp Can we start using Capital letters at the beginning of a sentence please, good starting point

training chat models is not a clean industrial process. different training runs even using the same datasets can produce models that are noticeably different in personality, writing style, refusal behavior, evaluation performance, and even political bias

@ChatGPTapp x.com/stetsblake/sta…

@ChatGPTapp well there's also just a (presumably) lawyer based degradation in performance. for example when I ask it to continue generating lyrics that I've written it refuses on copyright grounds. also where did the "code interpreter" feature go. this was much better at coding

we've heard all your feedback about GPT4 getting lazier! we haven't updated the model since Nov 11th, and this certainly isn't intentional. model behavior can be unpredictable, and we're looking into fixing it 🫡

@obrzts nice paper!

What I find surprising, is that naive B cells still join the response and eventually generate better, stronger binding antibodies. In other words, the early antibody feedback in this case is too weak for #originalantigenicsin to happen. More details: biorxiv.org/content/10.110…

In new preprint we delve into early memory B cells which are the frontline in generating antibodies against a familiar threat. We found that after first COVID vaccination a lot of pre-existing memory rapidly makes antibody-secreting cells despite weak binding to the viral protein

@DTarockoff I found this pretty informative: youtube.com/watch?v=SzzQkb…

please somebody tell me what shen yun is

@metapredict @davisidarta @PNASNews exactly, wasn't a lost decade for illumina or txg that's for sure

@davisidarta @PNASNews Penny should also have dropped about the overall scSEQ method by looking at raw data. New cell types based on part random 10% of the transcriptome coupled with the greatest expressed 15% of genes. Very silly. Sequencing per se part to blame + guys behind the methods ($$$)

@GeoffStanley9 no because why are you using gpt-3 still you clown

Excited to share a preprint describing the relationships of B cells resident in a range of human immune organs, with lessons about the pace of immune memory formation! Massive, really fun effort with @mswift_science and @StephenQuake biorxiv.org/content/10.110…