Sign in

John Didion

@jdidion.bsky.social
1.6K followers 154 following 27 posts

Absolute binf. 🧬/🦀.

PostsRepliesMedia
John Didion @jdidion.bsky.social · 03/11/2024
Yes!
010
John Didion @jdidion.bsky.social · 03/11/2024
I'll be at #ASHG24 this week with @fulcrumgenomics.bsky.social and my #Rust -colored glasses...find (or DM) me if you want to talk bioinformatics, Rust, basketball (Go Warriors!) or pretty much anything except politics.
130
John Didion @jdidion.bsky.social · 23/02/2024
Guy complaining about how "every kid gets a trophy these days" while drinking out of his "Worlds's Best Dad" mug.
020
John Didion @jdidion.bsky.social · 12/02/2024
Kudos to Jakob (is he on here?) and to all the groups who do the important work of independently benchmarking tools in bioinformatics and other computational fields. It is too often underfunded and underappreciated, but I think it's just as critical as developing novel methods.
000
John Didion @jdidion.bsky.social · 12/02/2024
Peer review is supposed to mitigate such issues, but even then the biases can be so subtle that only the most diligent reviewer would catch them all. Our tendency is to just look at these headline, accept it at face value, and hit reskeet/tweet/toot.
000
John Didion @jdidion.bsky.social · 12/02/2024
The incentive to overlook one's own bias is proportional to the degree of financial incentive. It's difficult to get someone to understand their benchmarks are wrong when their salary depends on their not understanding it.
000
John Didion @jdidion.bsky.social · 12/02/2024
Even when someone is trying to be fair and unbiased in evaluating their own work, they are likely to fail: bias is often unconscious, and humans (especially smart ones) tend towards laziness, hubris, and impatience (to paraphrase Larry Wall).
000
John Didion @jdidion.bsky.social · 12/02/2024
Never, ever, ever trust benchmarks when the person doing them also has an interest in one of the things being benchmarked. It's the same logic as not wanting the person who implemented the code to be the (only) one testing it.
410
John Didion @jdidion.bsky.social · 07/02/2024
Ohhh...this is very cool
020
Reposted by John Didion
Carl Zimmer @carlzimmer.com · 01/02/2024
Does being XX lead to more autoimmune disorders than XY? Here's my story on a chromosomal mystery. Gift link: www.nytimes.com/2024/02/01/h... 🧪
12412
John Didion @jdidion.bsky.social · 03/02/2024
Sounds like there's still some exciting stuff that will actually be announced at the conference...let's wait to see.
000
John Didion @jdidion.bsky.social · 03/02/2024
Damn good title. Bravo. www.biorxiv.org/content/10.1...
000
Reposted by John Didion
Jonathan Eisen @phylogenomics.bsky.social · 03/02/2024
Dear 23 and Me -- I realize times at tough for you right now, but preventing people from downloading their own data is not OK
3298
John Didion @jdidion.bsky.social · 03/02/2024
IMO this is the most exciting news coming out of this year's AGBT. Anyone have anything better? (Please don't reply with "Ultima") seqwell.com/seqwell-and-...
100
John Didion @jdidion.bsky.social · 03/02/2024
I have no idea what the demographics are for people who listen to metal, but it's gotta be like 90% parents, right? I mean how else do you guys cope?
010
Reposted by John Didion
Tomalak @tomalak.bsky.social · 13/01/2024
The best Internet hack I've seen in a while. It actually works. 😯 "If you add “cooked.wiki/“ to beginning of ANY recipe URL (before the https:/) it strips away everything but the recipe and organizes the ingredients and method into separate columns." stolen from theblower.au/@Maxton/1117...
35817
John Didion @jdidion.bsky.social · 06/12/2023
I'm happy to announce the release of WDL 1.1.1! This patch release includes many improvements to the specification. Read all about it on WDL's new blog: dev.to/openwdl/anno...
001
John Didion @jdidion.bsky.social · 03/12/2023
Generally agree, but some poisons kill you while others can arguably, under the right circumstances, provide some benefit (e.g., alcohol).
100
Reposted by John Didion
Tim Triche, Jr.🔰 @timtriche.bsky.social · 03/12/2023
VC money is a poisoned chalice, avoid until unavoidable
101
John Didion @jdidion.bsky.social · 03/12/2023
A) Any. God please I just need enough to survive until the market improves. B) Only if the terms are really good. C) B, plus they’re a follow-on investor you won’t have to talk to much. D) None. Their advice and connections will be unhelpful at best, and likely detrimental.
000
John Didion @jdidion.bsky.social · 03/12/2023
Let’s say you’re a biotech founder raising money. You meet with someone from a VC firm who literally believes we have the ability to “uplift other species” (maybe someone gave him a David Brin novel and forget to tell him it’s Sci-Fi). Under what circumstances to you take their money?
310
Reposted by John Didion
Seth Stadick @ducktapeprogrammer.bsky.social · 02/12/2023
If you are looking to use Rust for Advent of Code, this repo is a nice template for the boilerplate setup used each day: github.com/sstadick/aoc... #rust #adventofcode
001
John Didion @jdidion.bsky.social · 02/12/2023
baltimorespiritsco.com They also make some nice Amaro derivatives. And gin. Basically everything they make is great.
110
John Didion @jdidion.bsky.social · 02/12/2023
I forgot how amazing this stuff is! Found a bottle with a finger left in it and I’m kicking myself or not buying more before I left Maryland.
100
John Didion @jdidion.bsky.social · 11/11/2023
Really excited about monaspace, but... Huge missed opportunity to name their new technique "Textural Healing"
022
Reposted by John Didion
Milot Mirdita @milot.bsky.social · 24/10/2023
The new Awk book second edition arrived today. Awk's enduring relevance is amazing. The first ed. of the book was published before I was born and despite (because?) its age I use Awk everyday and encourage all our students to learn it. Here are my Awk slides/colab: docs.google.com/presentation...
A picture of the new second edition Awk book.
0215
Reposted by John Didion
Nils Homer @nilshomer.com · 24/10/2023
A good test to see if you’re writing maintainable code is to re-read it six months later. Both pandas & tidyverse fail this test, unless I add copious comments about the shape/content/schema/types of the data frames after each transform. Comments should not replace data structures and typing…
1104
John Didion @jdidion.bsky.social · 24/10/2023
It's too bad, I really liked eLife. Will no longer consider reviewing or submitting there.
030
Reposted by John Didion
Jonathan Eisen @phylogenomics.bsky.social · 24/10/2023
So incredibly disappointed and disturbed by what elLife has done here x.com/mbeisen/stat... And yes this is my brother who they fired for posting a link and a comment about an Onion article. But I can say without any bias what they did was idiotic and wrong.
x.com
416637
John Didion @jdidion.bsky.social · 16/10/2023
Evil takes many forms
020
John Didion @jdidion.bsky.social · 14/10/2023
That's great, thanks! Never knew those were there.
010
John Didion @jdidion.bsky.social · 14/10/2023
I'm looking for an Evil BAM file. By Evil, I mean all the weird cases every bioinformatician has seen at some point and gone "Oh $*%!, I forgot to handle that case in my code!" If this doesn't exist then we should make it happen. Really needed for testing genomics tools.
5217
John Didion @jdidion.bsky.social · 13/09/2023
Thanks, thats what I figured, just hoping I’d missed a chapter in the Rustonomicon. Const generics are really nice for libraries, but eventually you have to set the parameter value. I guess the solution is a common API with different impls for static and dynamic use cases.
010
John Didion @jdidion.bsky.social · 13/09/2023
I guess I could create an enum with variants for all possible K, or build separate binaries each with a fixed K. /sarcasm
000
John Didion @jdidion.bsky.social · 13/09/2023
Rust folks - how are you using const generics in the context of command line tools? For example, I’m looking at using the Kmer struct in github.com/COMBINE-lab/..., which is parameterized by K (kmer size). In my CLI, K is an option set by the user, so it’s not known at compile time. What do I do?
220