During our recent lab roster meeting, we asked about everyone's AI chatbot preferences (n=69, multi-select): 🥇 Claude — 49 (71%) 🥈 ChatGPT — 43 (62%) 🥉 Gemini — 14 (20%) Takeaway: Mostly Claude + ChatGPT co-usage, with Gemini as a distant third and a long tail of niche tools.
@gersteinlab.bsky.social
New @natcomms.nature.com paper by @katebowie.bsky.social @markgerstein.bsky.social, where we study how disinfection shapes microbes in hospital sink drain biofilms. Biofilms regrew in 4 days, enriched for carbapenem-resistant bacteria and multidrug efflux pump genes: www.nature.com/articles/s41...
In our @natmachintell.nature.com paper, we introduce a framework to analyse interpretability in deep learning by drawing on a formal notion of model semantics from the philosophy of science. We illustrate our framework with examples from biomedicine. Read here: rdcu.be/e9uYh
Curious how pseudogenes are transcriptionally regulated? Our new Genome Research paper shows processed pseudogenes break the rules: they’re transcribed without classic epigenetic marks, linked to enhancers, and enriched for YY1 motifs. Study co-led by Yunzhe Jiang and @beaborsari.bsky.social
Epigenetic characterization of pseudogenes across human tissues
Pseudogenes have historically been regarded as nonfunctional remnants of genome evolution. However, relative to other noncoding genomic elements, their promoter architecture and epigenetic regulation remain incompletely understood. Here, we systematically characterize pseudogene promoters and compare them with those of protein-coding genes and long noncoding RNAs. To do this, we integrate matched transcriptomic and epigenomic data across 26 human tissues from the EN-TEx (ENCODE-GTEx) project. We uniformly annotate promoters with chromatin features (histone modifications, chromatin accessibility, and DNA methylation), sequence motifs, and evolutionary conservation, generating an online catalog. Leveraging this catalog, we show that, across multiple tissues, transcribed, unprocessed pseudogenes exhibit chromatin patterns similar to those of active protein-coding genes. In contrast, transcribed, processed pseudogenes show a different pattern: most lack the canonical hallmarks of transcription (e.g., active histone marks) at their promoters. Instead, their promoters show increased overlap with LINE elements, enrichment for YY1-like binding motifs, and higher Hi-C contact frequency, particularly with distal enhancer-like regulatory regions. Together with their greater conservation (relative to unprocessed pseudogenes), these features suggest that the transcription of processed pseudogenes may require regulatory mechanisms distinct from canonical promoter-associated epigenetic activation.
doi.org
Our @neuripsconf.bsky.social work led by YunyangLI “E2Former: An Efficient and Equivariant Transformer with Linear-Scaling Tensor Products” was selected as a spotlight (with score ranked ~17 / 21k submissions). Poster: Thur Dec 4, Exhibit Hall CDE #5512 Online: openreview.net/forum?id=ls5...
Our new PNAS study bridges histology and genomics. Using deep learning and imageQTL analysis, we show how tissue images reflect gene expression and aging — making histology more interpretable with AI. By RanMeng, W. Zhu, C. Cameron, P. Ni, X. Zhou, T. Ulammandakh, and @markgerstein.bsky.social
PNAS
Proceedings of the National Academy of Sciences (PNAS), a peer reviewed journal of the National Academy of Sciences (NAS) - an authoritative source of high-impact, original research that broadly spans...
pnas.org
📚 Yale students have returned to campus, so time for a roster meeting! We again made our Nobel Prize predictions (given how accurate we were last year 😉) 🥇Our top prediction is Habener & Knudsen (GLP-1) with 28.5% of the vote! 🥈 In second is Rothberg & David Klenerman (NGS)
Curious how your favorite gene changes when and how during a biological process? Want to dive into the kinetics of chromatin + gene expression? Meet chronODE, our new tool to model multi-omic time-series with logistic equations + ML! doi.org/10.1038/s414...
The chronODE framework for modelling multi-omic time series with ordinary differential equations and machine learning - Nature Communications
Here, the authors use a simple equation to study how genes and their regulators switch on/off over time, across the whole genome in tissues and cells. Most changes are gradual, but some genes switch q...
doi.org
1/4 🚀 New #ICLR2025 SPOTLIGHT ALERT Gerstein Lab presents “Enhancing the Scalability & Applicability of Kohn-Sham Hamiltonians”—led by YunyangLI & Z Xia & L Huang & J Zhang & @markgerstein.bsky.social . Joint work with @msftresearch.bsky.social .
New @biophysj.bsky.social paper by Alan Ianeselli, Joe Howard and @markgerstein.bsky.social . A Molecular Dynamics algorithm to rapidly compute protein folding pathways and identify folding intermediates for targeted drug discovery! doi.org/10.1016/j.bp...
Our new paper describes the iDASH-winning method for efficient blockchain storage of biomedical data. We cut gas costs by 60% and sped up retrieval 500x with low-level Solidity optimization. By Eric Ni, Elizabeth Knight, @markgerstein.bsky.social doi.org/10.1016/j.jb...
New paper by Gaoyuan Wang, Jonathan Warrell, Prashant Emani and @markgerstein.bsky.social ! Check out our new model QVAE, a fully quantum variational autoencoder with latent regularization: journals.aps.org/pra/abstract...
🚨We have an immediate postdoc opening for US nationals (citizens/green-card holders). Needs to be filled within 6 months. Lots of fun topics (e.g. biosensors, brain genomics, AI for bio, &c). If interested, see jobs.gersteinlab.org
Jobs
Post-doctoral Position in Biomedical Data Science at Yale Applicants are invited for a post-doctoral position at Yale University. The position is for 2 years with possible extensions. The choice of…
jobs.gersteinlab.org
Excited to share our new paper in @cellcellpress.bsky.social on digital phenotyping from wearable biosensors to characterize psychiatric disorders and identify genetic associations, led by @jasonjliu.bsky.social and @beaborsari.bsky.social @markgerstein.bsky.social: doi.org/10.1016/j.ce...
Redirecting
doi.org
New @plosone.org paper by Xiao Zhou, Sanchita Kedia, Ran Meng, and @markgerstein.bsky.social. Our deep learning framework analyzes fMRI scans for early Alzheimer's Disease detection, achieving 92.8% accuracy with a focus on model interpretability: t.co/Ro5WzyJUyZ