erf generated by deepseek (left) and gpt-4o (right) are nearly identical. even have same comments and formula reference. only difference I see is parens in sign definition:

This is very cool work and I'm happy to see it published. Vcfexpress by @brent_p and @aaronquinlan allows building (essentially) arbitrary VCF filters expressed in lua code with parsing & eval powered by rust! https://t.co/Lbki2dLHOl

A present, emotionally healthy father is one of the most underrated forms of mental healthcare.

new release of vcfexpress with some performance improvements for common uses: https://t.co/DBUkErzfHI also adding to list of examples, for example filtering on posterior genotype prob:

vcfexpress applies simple user expressions variants in a VCF. it can replace one-off python scripts to manipulate VCFs, likely with better performance. we'd like to collect use-cases here: https://t.co/D4o6h0zrGM if you have a use-case and want some pointers, open an issue

already got a PR accepted to fix this. and then additional help to update my code. 👍

In addition to my deficiencies in rust, I often hit stuff like this in noodles: https://t.co/v2w3sFqZhN so noodles seems to have gaps, *and* it requires greater facility with rust than rust-htslib, which I tend to use.

We reproduced DeepSeek R1-Zero in the CountDown game, and it just works Through RL, the 3B base LM develops self-verification and search abilities all on its own You can experience the Ahah moment yourself for < $30 Code: https://t.co/B2IsN1PrXV Here's what we learned 🧵

vcfexpress is a command-line tool built in rust that lets users apply lua expressions to modify or filter a vcf from the command-line https://t.co/jOeg9dDAp8 new release with better docs https://t.co/L1sa9fBsIL and examples

I am looking for a mentor for the rust programming language. My latest issue is with trait bounds ( https://t.co/wUIjeFOfJ8) but I have a few things I generally hit. I can compensate with $$ or interesting problems. :) Please share with relevant people and feel free to DM.

with @JKunisaki from @aaronquinlan lab, have been dusting off fraguracy, which evaluates sequencing error rates using the portion of bases from paired end reads that overlap. new release adds, among other niceties, tracking for distance to homopolymer. https://t.co/FDDSQOoyzC

Here's a small package that integrates htslib with your Zig build, making it easier to work on genomics projects using #ZigLang. Check it out 👉 https://t.co/t0bj8VIOkt #Genomics #Bioinformatics

echtvar still (echt waar) annotates variants extremely rapidly with large population databases where most variants are encoded into 32 bits (and are then compressed with stream v-byte). latest release fixes a small bug https://t.co/3SlD8ndJWu

Rust execution engine + Lua expressions, God what a great combination! This looks awesome. Congrats @brent_p and @aaronquinlan !

Vcfexpress: flexible, rapid user-expressions to filter and format VCFs https://t.co/Acg20PMZ41 work with @aaronquinlan lua expressions and a sandbox feature so it can be run safely.

@brent_p Thanks @brent_p this is the paper but there are other extensions done by @GenomeInABottle https://t.co/wiXG2gVVYS

don't use hg38.fa as-is. checkout the references 😜 here: https://t.co/UBhkn6zjDx rendered the ipynb (not mine) here: https://t.co/Zv6J1d6XAy in short, use: GRCh38_GIABv3_no_alt_analysis_set_maskedGRC_decoys_MAP2K3_KMT2C_KCNJ18.fasta.gz other updates on the best hg38 reference?

This looks like a nicely done comparison of pathogenicity prediction tools. Even stratifies by date the (truth-set) variant was created to hint at leakage/training on truth. BayesDel metamethod does quite well. Scores recalibrated by conservation ++

vcfexpress is fast in benchmarks (re-run from vembrane paper). it's very customizable; simple user lua code to parse CSQ fields (to find HIGH impact variants, for example) is still quite fast. https://t.co/IDUi14nKyc

new tool: vcfexpress let's user apply lua expressions to filter/modify a VCF and optionally to write templated output: https://t.co/jOeg9dDAp8 it's fast. feedback appreciated. Work with @aaronquinlan