Very glad to see our CherryML method published online in Nature Methods. The published version has updated and expanded results, so please check it out. doi.org/10.1038/s41592… (1/3)

After lengthy anticipation, we finally got to compare PrimateAI-3D with our Cross-Protein Transfer (CPT) model doi.org/10.1186/s13059… Below is a summary of our findings, including a comparison of ESM-1b and ESM-1v protein language models. 1/8

We recently posted a preprint describing GPN-MSA, a DNA language model that leverages whole-genome alignments across multiple species while taking only a few hours to train. This thread summarizes its performance on the human genome. doi.org/10.1101/2023.1… 1/12

GPN-MSA scores for ~500M gnomAD variants are now available. Scores for a specific region can be queried using a simple tabix command, as detailed here: huggingface.co/datasets/songl… Preprint: doi.org/10.1101/2023.1… Previous Post: x.com/yun_s_song/sta…

After 2 rejections & 4 rounds of review (all by the same journal), our first GPN paper is finally published: pnas.org/doi/10.1073/pn… We feel jaded, but it was worth the effort and we are grateful to the reviewers for their time & helpful feedback. Some lessons learned: (1/2)

I’ll be presenting GPN-MSA at #ASHG, Reviewers’ Choice Poster PB4125 Fri 3-5PM. Come chat about DNA language models! x.com/yun_s_song/sta…

Lately I am into making potential cover images for journal articles from my lab. None has been selected so far, but that's alright; I enjoy the process. Here is an image I made for our recent GPN article: pnas.org/doi/10.1073/pn…

A long-context language model for the generation of bacteriophage genomes biorxiv.org/cgi/content/sh… #biorxiv_bioinfo

I’m excited to share our work on the somatic evolution of the mitochondrial genome: nature.com/articles/s4155… . Mt-genomes are a dynamic subcellular population. But, how does the interaction between nuclear and mitochondrial ancestry impact mutation within individuals’ lifespans?

We are delighted to share our work on a robust, accurate method for sampling genome-wide genealogies (aka ARGs) from the full posterior distribution. This is a product of years of hard work by Yun Deng 邓赟, who is jointly advised by Rasmus Nielsen and me. doi.org/10.1101/2024.0… 1/n

Highlighting some work I've enjoyed reading recently: "DNA language models are powerful predictors of genome-wide variant effects" by Gonzalo Benegas, Sanjit Batra, and Yun S. Song: pnas.org/doi/10.1073/pn… It's simply incredible unsupervised DNA language models can learn so much!

An independent benchmarking study comparing 84 variant effect predictors ranked our method, CPT, at the top for DMS data, beating AlphaMissense. Many thanks to Joe Marsh for conducting this study! Marsh lab paper: doi.org/10.1101/2024.0… Our CPT paper: doi.org/10.1186/s13059…

We’re excited to release gReLU (github.com/Genentech/gReLU), a Python package for deep learning on DNA! gReLU covers all major steps in DNA sequence modeling: data processing, training, evaluation, interpretation, variant effect prediction, and ML-guided DNA sequence design. 1/13

We were invited to write a review article on DNA/genomic language models (gLMs). We took this occasion to gather our thoughts on promising applications, and major considerations for developing and evaluating gLMs. Pls share with your colleagues: Preprint: arxiv.org/abs/2407.11435

Had so much fun writing this small review with Nicolas Lou and Peter Sudmant about structural variation in primates in the era of T2T genomes and pangenomics :) sciencedirect.com/science/articl…

My main points at the #ICML2024 AI for Science panel (general guidelines, exceptions exist): -- Good projects are structured to have intermediate milestones that are measurable, thus creating constructive feedback loops. (I knew this one would get to Max Welling 😀) --

We are excited to share the preprint of the main project I worked on during my PhD: understanding the color polymorphism of the strawberry 🍓 🐸 poison frog in Bocas del Toro, Panama biorxiv.org/cgi/content/sh… Rasmus Nielsen

Biologists taking machine learning techniques from computer vision and nlp

Our paper on the evolution of structural variation at the amylase locus is out today in Nature! nature.com/articles/s4158… congrats to co-first authors Joana L. Rocha Nicolas Lou Davide Bolognini Alessandro Raveane & Alma Halgren, co-corresponding author Erik Garrison