ASHG25 Poster 4106 Today at 2:30pm! Whole-genome long-read sequencing from 10 NANOGRAMS, not micrograms! ULI-HiFi delivers: ✅ 99.82% F1 for SNVs ✅ 98.9% TR accuracy ✅ Discovered LIMD1 TR expansion progressively increasing from normal→polyp→adenocarcinoma, with evidence of function
Graham Erwin
@grahamerwin.bsky.social
Asst. Professor, Baylor College of Medicine. Previously Stanford and UW–Madison. We use molecules and genomes to understand repetitive DNA sequences. #FirstGenAcademic ErwinLab.org
At ASHG25 RIGHT NOW: Deep-coverage PacBio HiFi sequencing reveals hotspots of tandem repeat variation in the male germline. 12 men, paired blood/sperm, 1.7M repeats analyzed. What we found about mutational bias will surprise you! Poster 6044W, 2:30-4:30pm today. Lucas Finot presenting.
Proud to announce that we received the NIH Director's New Innovator Award to develop a new class of genome editing tools!
Dr. Graham Erwin @grahamerwin.bsky.social receives NIH New Innovator Award! www.bcm.edu/news/dr-grah... @bcmhouston.bsky.social @bcmgenetics.bsky.social #BCMCancerCenter
A follow-up to my quote of the day this morning: A short Bluetorial about Vannevar Bush and his legacy
a cartoon says hey everybody an old man 's talking while bart simpson looks on
ALT: a cartoon says hey everybody an old man 's talking while bart simpson looks on
media.tenor.com
Interested in modern grant writing in the age of AI? Join us this summer through Stanford Continuing Studies! continuingstudies.stanford.edu/courses/prof...
Modern Grant Writing with AI
AI is transforming grant writing—turning a time-intensive task into a more strategic, efficient process that elevates both research and storytelling. This hands-on course explores how to integrate AI ...
continuingstudies.stanford.edu
Big: The final version of a randomized, controlled World Bank study finds using a GPT-4 tutor with teacher guidance in a six week afterschool program in Nigeria had "more than twice the effect of some of the most effective interventions in education" ("equating to 1.5 to 2 years" of standard school)
I had such a powerful experience at BCM SKY Retreat this weekend led by @grahamerwin.bsky.social and met so many amazing people! It was a thought-provoking and reflective weekend with lessons that I will carry with me! What a privilege to partake in such a transformative experience! 🥰
A first in human genomics: a 4-generation pedigree reference, now in Nature. Built with #PacBio HiFi, it maps de novo mutation rates, reveals paternal bias, and captures high mutation rates in tandem repeats—even in Y and repeat-rich regions. Paper here: go.nature.com/4lGMPlP #TheresHiFiForThat
Human de novo mutation rates from a four-generation pedigree reference - Nature
Analysis of more than 95% of each diploid human genome of a four-generation, twenty-eight-member family using five complementary short-read and long-read sequencing technologies provides a truth set t...
go.nature.com
Advanced genomic analysis of 4 generation family offers new knowledge about genetic mutations & their transmission, including inherited variants & those that arise anew @nature.com @uwgenome.bsky.social @eichlerlab.bsky.social @uwmedicine.bsky.social @utah.edu @pacbio.bsky.social
There are a number of tools (Elicit, Perplexity, Google's Deep Research, Consensus within ChatGPT) to find references. One of the complaints from academics is these tools don't find "the right papers". So far, ChatGPT Deep Research and their o3 model have the best "taste" that I have found. (1/3)
ChatGPT
A conversational AI system that listens, learns, and challenges
chatgpt.com
The LLM (AMIE in this case) performs higher (59.1% top-10 accuracy) than the clinician with the LLM (51.7%). Another study consistent with what @emollick.bsky.social has been sharing for a while now. www.nature.com/articles/s41...
Updated this chart with the newest Gemini. It shows the rapid progress in AI over less than two years: costs for GPT-4 class models has dropped 99.7% and even the most advanced models (beating PhDs at the GPQA) are still 82% cheaper than GPT-4. Probably not worth betting on this trend ending soon.
A new means of selectively imaging the inner mycolate of mycobacteria using a compound that covalently modifies an enzyme intercepts to create a new intermediate with the Kim lab, @thebrysonlab.bsky.social and the Bernhardt lab. #TB
First post here! To inaugurate this account, I'd like to share my recent work, now published in JACS. Intercepting a Mycobacterial Biosynthetic Pathway with Covalent Labeling | Journal of the American Chemical Society pubs.acs.org/doi/10.1021/...
The strchive.org paper is out!! The paper describes STRchive as a resource to improve the diagnosis of tandem repeat disorders, then goes beyond it to consider what can be learned about childhood onset and population prevalence of these diseases. 🖥️ 🧬 link.springer.com/article/10.1...
STRchive: a dynamic resource detailing population-level and locus-specific insights at tandem repeat disease loci - Genome Medicine
Approximately 8% of the human genome consists of repetitive elements called tandem repeats (TRs): short tandem repeats (STRs) of 1–6 bp motifs and variable number tandem repeats (VNTRs) of 7 + bp moti...
link.springer.com
A new article from Laurel Hiatt and @hdashnow.bsky.social describing STRchive is now available at Genome Medicine. genomemedicine.biomedcentral.com/articles/10.... Check out the database resource, as STRchive "streamlines TR variant interpretation at disease-associated loci." strchive.org
STRchive: a dynamic resource detailing population-level and locus-specific insights at tandem repeat disease loci - Genome Medicine
Approximately 8% of the human genome consists of repetitive elements called tandem repeats (TRs): short tandem repeats (STRs) of 1–6 bp motifs and variable number tandem repeats (VNTRs) of 7 + bp moti...
genomemedicine.biomedcentral.com
Evidence that a well prompted LLM can help learning from a (small, single-subject) randomized controlled trial at Harvard: “here we show that students learn more than twice as much in less time with an AI tutor compared to an active learning classroom, while also being more engaged and motivated.”
We don't spend a penny on marketing. It's all word of mouth for @costplusdrugs.com That's how we keep prices so low. Please spread the word !
The first bifunctional protein degrader to deliver Phase III data! It seems to work, but people were definitely expecting more:
A Bifunctional Degrader Reads Out in Phase 3
science.org
5 yr anniversary of the day I started my COVID ICU visit
Spectacular news this morning for CRISPR gene editing to treat genetic disease from Beam Therapeutics. This method was developed by David Liu's lab at the Broad. = "academic" science produces innovations that enables the biotech sector to build medicines investors.beamtx.com/news-release...
investors.beamtx.com
Excited to share the peer-reviewed version of our paper, which is out today!! Bispecific antibodies targeting the N-terminal and receptor binding domains potently neutralize SARS-CoV-2 variants of concern | Science Translational Medicine www.science.org/doi/10.1126/...
Bispecific antibodies targeting the N-terminal and receptor binding domains potently neutralize SARS-CoV-2 variants of concern
Bispecific antibodies incorporating a cross-reactive NTD-specific antibody are resilient to SARS-CoV-2 variants of concern.
science.org
Researchers from @pollyfordyce.bsky.social lab explore what makes some transcription factors more or less selective for their DNA ligands. This work could guide the design of more selective DNA-binding proteins for many bioengineering applications:
Mutations to transcription factor MAX allosterically increase DNA selectivity by altering folding and binding pathways - Nature Communications
In this work, high-throughput microfluidic assays reveal how allosteric substitutions that alter folding-and-binding pathways shape DNA specificity and selectivity landscapes in the human…
buff.ly
ChatGPT Code Interpreter is still my favorite implementation of the pattern where an LLM can write code, run it, see any errors and iterate to fix them It's been stuck on GPT-4o for ages - today it got a huge upgrade and now work with GPT4.5 and o3-mini! simonwillison.net/2025/Mar/5/c...
Demo of ChatGPT Code Interpreter running in o3-mini-high
OpenAI made GPT-4.5 available to Plus ($20/month) users today. I was [a little disappointed](https://simonwillison.net/2025/Feb/27/introducing-gpt-45/) with GPT-4.5 when I tried it through the API, bu...
simonwillison.net
As #AGBT approaches, discover how #PacBio SPRQ chemistry for Revio is advancing #HiFisequencing efficiency, reducing costs, and unlocking new research possibilities. bit.ly/4g89HXK #AGBT25
Lighting a SPRQ: How new Revio chemistry fuels innovative research - PacBio
New SPRQ chemistry for the Revio system is revolutionizing HiFi sequencing with increased efficiency, affordability, and reduced DNA input. Learn about its impact on research in rare diseases, cancer, and multiomics.
bit.ly
We are starting to see what "AI will accelerate science" actually looks like. This Google paper describes novel discoveries being made by AI working with human co-scientists (something I think we have all been waiting to see), along with an early version of an AI scientist for wider use.
At #BPS2025? Check out the posters by @sovannytaylor.bsky.social and @gtleung.bsky.social on Monday. Venom peptides as tools to study cell penetrance and designer condensates to probe cytoplasmic rheology across length scales.
Highly recommend the Unpaywall extension to access manuscripts — it's easier than Sci-Hub, and legal! unpaywall.org/products/ext...
Unpaywall
unpaywall.org
Here’s my post on the NIH indirect costs slash-and-burn. God, these people are exhausting:
The Continuing NIH/NSF Crisis, Part III: Indirect Costs
science.org
You will never guess the report that took Deep Research the longest to generate of any I made before. 30 pages, 10,600 words, and actually super interesting to read based off a single prompt. Full report (and prompt): chatgpt.com/share/67a152...
A long post about what’s happening to the science funding agencies in the US and why. As mentioned, this one just kept getting longer even as I kept stripping curse words from it. www.science.org/content/blog...
What's Happening Inside the NIH and NSF
science.org
Friends and colleagues, I’ve written a book on effective functional genomics study design, which will be available on Amazon in a couple of weeks. Sharing the TOC to spark interest. I hope students and those planning genomics experiments will find it useful! I’ll share updates soon