This is a compelling argument @froggleston.carpentries.org That blindly trusting the random person who wrote a package is not the same as blindly trusting an llm

This sentiment is not so different than wanting to understand generally what a package or library does, without worrying about the internal implementation.

I dunno, yes you can dig for it, but does anyone actually? Like actually click through *all* old versions of a preprint to check for comments???

Biorxiv reviews just feel like reviewing into the void- public doesn't notice them, authors can ignore/bury them, now we learn even authors don't see it?

Missing from this discussion is that if authors want to bury a comment they don't like on their biorxiv preprint, all they have to do is just upload a new version. It happens even if they wanted it left in the public discussion, but want to address the comments with a new version!

I was discouraged from giving biorxiv reviews when I realized they are buried as soon as a new version is posted.

I think you should apply ;p After all it will fit "seamlessly into your existing commitments", and you've probably got an inside vote on the hiring committee

Pocket casts?

It is really cool that V4-V5 captures organelles so well... but I LOLed a little when I realized JEDI was just new branding for good old v4-v5 16S primers

Parallelle 😉

Another option in addition to sandpiper is IMNGS. Sandpiper searches NCBI metagenomes for markers, IMNGS searches NCBI amplicon datasets (and since you know the 16S... there are a lot more of those datasets out there)

Ha cool. My first thought was a breast pump flange...ouch 😂

I had to change pediatricians to do this.

I was met with derision when I asked about early vaccination. After seeing how parents are treated when requesting a "nonstandard vaccination schedule" I feel like I understand why parents become antivax.

Congrats!! Well-deserved!

Yes but just not the type of coding the original post was using ggplot downloads as a proxy for.

Also lots of people code in python and then use ggplot to make their figures. And that's not what you mean when you say "coding in R" right?

For the large fraction being lower, do you think that is because proportionally more euk reads in those metaGs?

Ahh thanks! Ok so yeah, def higher than 5% but I guess even with very carefully made mags I should not expect to get close to what I'm used to with water!

(And I do like your combined marker+mag approach for dealing with this too)

Yes I did not see the mag coverage listed explicitly since they used marker genes for the diversity part. Suspect that number was less impressive?

I realized I had no baseline for a sanity check! Does make you think about interpreting mag data differently by system though.