... and second is to have a map from the figures to where they are made in the associated "experiments" code repository (github.com/matsengrp/dn...):

I forgot to post two things I liked doing in this paper that I hope catch on. First is to have links in the methods section to the model fitting code (in a tagged version github.com/matsengrp/ne... as the code continues to evolve):

Oh, and here is a picture of a cyborg-Darwin (cooked up by Gemini), after he realized how useful transformers are. For some reason MBE didn't want it as a cover image!

And because natural selection is predicted for individual sequences, we can also investigate changes in selection strength as a sequence evolves down a tree:

Because this model isn't constrained to work with a fixed-width multiple sequence alignment we can do things like look at per-site selection factors on sequences with varying CDR3 length:

The final version of our transformer-based model of natural selection has come out in MBE. I hope some molecular evolution researchers find this interesting & useful as a way to express richer models of natural selection. doi.org/10.1093/mol... (short 🧵)

Solution: first embed 3-mers and then the number of parameters goes up only linearly with the context width.

Here are some useful subagents we've developed for Claude Code.
github.com/matsengrp/c...
(description in README if you don't know what I'm talking about)

Example uses:

Working with Claude Code has been transformative. Yes, it does great on new code and refactoring. VSCode integration is 👍.

But the big surprise: it can seamlessly edit notebooks and then view the plots to plan next steps. Time to update my blog post matsen.group/general/202...

Paper 3: doi.org/10.1101/2024... (in revision for NAR.) Clear evidence for the effects of microhomology in VDJ recombination: if there is a little homology between the two trimmed ends, then the sequence is more likely to ligate. This was a very challenging project; Maggie crushed it!

Paper 2: dx.doi.org/10.7554/eLif... Sequence determinants of VDJ recombination trimming. VDJ trimming distributions have all these crazy shapes yet are consistent between individuals. Why? Maggie modeled sequence dependence with a handful of parameters and got a model that fits beautifully (orange)

Paper 1: dx.doi.org/10.7554/eLif... . GWAS finding first examples of trans genetic variants associated with VDJ recombination outcome. Look, there's Artemis (DCLRE1C), providing an independent line of evidence for Artemis being the agent of VDJ recombination trimming.

Congratulations to Maggie Russell @magdalenarussell.bsky.social for graduating from @mcbseattle.bsky.social with a phenomenal thesis. I will retell the story of her thesis through images from her papers. 🧵

(Here we are with Phil Bradley, who co-advised Maggie.)