Would this work in the context of quarto and Rmarkdown too? We've been using simple Google forms for internal training quizzes but there would be a huge benefit to switch to something more programmatic. This seems promising!

Even though it's maybe not an authentic output of a neural network, I'm partial to the Hallmark Christmas movie script:
www.reddit.com/r/RomanceBoo...

Similarly, so many empty poster boards 😢

Agree- but on the positive side, I've been impressed by whatever software they're using for closed captioning! Today I saw it pick up PI names, complicated disease names, gene names, etc- all despite speaker's volume/intonation. Maybe something AI has actually been good for..?

Yep! Will be there starting tomorrow - happy to meet up or go for a coffee!

For out of towners it should also be noted that although it's "just" a bus, the MBTA Silver Line runs mostly (if not completely) underground (traffic free) and connects Seaport with South Station for access to the Red Line and the rest of the city. And it's a simple tap-to-pay system now too

The stuff adjacent to MCEC in Seaport is a little sterile, but within a 20min walk or so there's:
Row34, Chickadee, Yume Ga Arukara, Somenya, Shojo, Woods Hill Pier 4, Nautilus, High Street Place food hall, Trillium, Little Wolf, Flour, James Hook and co, Yankee Lobster

Please hire Nur! She's brilliant, hard working, AND she bakes... What more could you want?

Years ago when I was at UNC (now at DFCI) we were looking at environmental chemicals and came across this class of fungicides that induced a lot of transcriptional changes in neurons:
www.nature.com/articles/nco...
Curious if there could be something additive here..?

Super interesting results! Do you think this proximity could lead to secondary problems, in that if these mitochondria are also dysfunctional and producing ROS, their proximity to the nucleus increases the probability of DNA damage?

Got it, thanks! It's less clear to me whether there's any benefit to this cohort-level approach if you're dealing with patient samples but curious if you or others have thoughts on that

but maybe it doesn't add much in this context? e.g. I have n=6 treated and control cell lines and I'm looking to predict antigens enriched in the treated cells. This workflow may find antigens enriched in any one replicate, but could/should we merge splice/transcript assemblies and quantify across?

This looks great! One question I have is whether each sample in a cohort gets processed completely independently, or whether there is any cohort-level discovery and quantification? I've always thought a 2-pass approach to assemble and quantify across samples was preferred if the data were available

Pretty sure original authors say attack-seek

"Healthcare access" is not healthcare

This is a test run, they'll employ this same model to polling places in NC for the next election

I also will often work backwards for an entire analysis: figure out what question I'm trying to answer, what the key conclusion is, sketch an idealized figure that would support that conclusion, then figure out the steps to make that figure. Then cry and repeat when it doesn't match my sketch