Chris Mitchell

I've been using the #salvo webserver for quite some time now and really like it. The way you can inject and share state along a route tree is very useful for access control. For orm I've been using #seaorm and I can't get used to it. #Tokio's #toasty seems to have a much better form factor. #rust

One of the really nice parts of COVID was how research was being disseminated. Lots of small bite sized research published in preprints/blogs/etc that was easy to review. It was a glimpse into another way and I found it far more engaging.

Given the lack of any traffic enforcement I'm pretty sure Boston doesn't give a shit any ticket based revenue unless it's street cleaning.

#Ai has been great for writing code when feeding a newborn/being nap-trapped. I can slowly type with one hand what to do and then review what comes out. #kiro has been amazing for this due to the way it drafts up requirements and tasks.

I'm upset this doesn't do my taxes

You also get the reverse of this where samples with different vasculature (tumor-normal) report whatever is enriched in blood as biomarkers. Mir21 is a popular one.

I think pure scale is often misguided as well simply from a waste stream/consumable point of view. Cool, your small lab now has to deal with an Olympic sized pool of biological waste every day. Do more pooled stuff, then deconvolute in arrays. Which also doesn't require extensive automation.

I've been using AI more for some frontend experimentation and often times the code looks good enough. However, how often do you see caveats in the actual documentation about risks of using certain methods and security considerations. AI workflows will completely hide this from users. #ai

Keep it all pooled and separate winners in a pooled screen.

This is going to sound passive aggressive but I am delighted by the diversity of opinion in synbio today

People who think using fermentation to make food protein is an over-ambitious pipe dream

People who think genetic super-astronauts for mars is a near-term R&D priority

I'd actually love to see what your claude.md file looks like. I poked around that ai fix you published in minijinja but couldn't find a copy of your instruction file.

You can’t eat microchips. Social media can’t heal the sick. B2B SaaS will never love you

The final measure of any technology is the value it creates, directly or indirectly, for living things. In this sense, biology is everyone’s mission

Okay hear me out, what if building your protein model was a first person horror survival experience??

#MolecularNodes #b3d #GeometryNodes 🧪

I think when someone comes out with an orchestration engine where the local and production experience are similar it'll quickly replace k8s. That was the initial great part of docker + compose but now we're at a place where an entire separate stack exists that only runs in some cloud.

a cartoon of homer simpson holding a card ALT: a cartoon of homer simpson holding a card

I am not going to call this a Bluetorial, because I do not want to sully that term (which, at least I, find joyful).

Let’s call it a Orwellial.

I have also loaded human samples (HG00096 from 1000G) into this viewer, which while wildly impractical is nice to know it can handle it.

There's a huge need for more visualization options for #pangenomes. My coworker recently made this terminal graph view for gen (github.com/ginkgobiowor...) that I'm loving. Here's an example of a simple import of a genbank file and exploring the changes in the terminal #ratatui #rust

SAM/BAM is probably the best exception to the viable text based format I can think of (even fasta has a litany of broken readers/writers). Maybe it's because its complex enough nobody wants to make their own writer, but harmless enough to awk/etc. the output.

I've been working more with sqlite these days and keep thinking about how many problems may have been avoided if GFF/GTF or GFA started as a sqlite schema. That way the community could actually carry out schema/data migrations of changes instead of having 10s of versions with slight differences.

I've been working a lot with the GFA format and I find it so odd how many bioinformatic formats are text based. I think a large part of it is because it's the simplest thing to parse, but this led to so many broken parsers that work for the application that person happened to care about.

Encoding genomes to midi files is this year's lofi beats.

If you have 20 minutes to spare and want to get up to speed, watch this video. Very helpful to understand how rigorous and fair the indirect cost process actually is.

www.aau.edu/key-issues/i...

It’s been a tough few weeks. My 10yo daughter was diagnosed with a very rare, aggressive cancer called interdigitating dendritic cell sarcoma (IDCS). I’m reaching out to identify clinicians/patients who have encountered pediatric IDCS or other (non-LCH) dendritic or histiocytic sarcomas cases.

🐉

How population stratification makes environments look like genes. A short 🧵:

I can see that. I personally would lean towards parabricks since I can run that locally and incorporate it into unit tests/CI.

De novo designed proteins neutralize lethal snake venom toxins - Nature Deep learning methods have been used to design proteins that can neutralize the effects of three-finger toxins found in snake venom, which could lead to the development of safer and more accessible an...

This pretty much settles it: we're entering the era of low-cost, on-demand binders for everything

Excited to see what people build now that "inject a llama with some junk and bleed it a month later" is no longer the cutting-edge tech
www.nature.com/articles/s41...

One place I know of that does lean into closed source solutions is proteomics, where mascot is often used if it can be afforded. Conveniently the algorithm nobody can see is also the one that gives the most identified peptides.

I'd like to know who is really using dragen. I never really got the appeal of proprietary algorithms especially when the NGS world has been so blessed with open standards and tools.