Hua Bai

New publication from the lab @pnas . With the collaboration of Dr. Ping Kang @pingkang0, we uncovered how developmental NF-kB signaling links developmental timing (time to maturity) to lifespan😊😊pnas.org/doi/10.1073/pn…

@GuonHuang @XiaoxinChen2024 @PNASNews Congratulations 🎉

Our new paper led by @XiaoxinChen2024 is out @PNASNews! bit.ly/4785aCD We built a tissue atlas of ~2K transcription regulators ➔ identified Tox3 as a long-sought Atoh1 amplifer in cerebellar development, tumorigenesis & evolution. 1/10 📢 We're recruiting postdocs!

📢 Preprint: we present a whole-mouse-brain in vivo Perturb-seq atlas, 7.7 million cells, 1947 disease-associated perturbations, moving toward direct readout of how human genetics rewires cell states & circuits in vivo. Grateful for the Team! @NVIDIAHealth biorxiv.org/content/10.648…

At @GoogleDeepMind, we believe AI is the ultimate catalyst for science. 🧬 The best example of this has been the AlphaFold database (AFDB) of protein structure predictions which has been used free of cost by more than 3.3 millions researchers across the world! Today, in collaboration with @emblebi, @Nvidia and @SeoulNatlUni, we are expanding the database by adding millions of AI-predicted protein complex structures to the AlphaFold Database. To maximise global health impact, we’ve prioritised proteins that are important for understanding human health and disease, including homodimers from 20 of the most studied organisms, including humans, as well as the @WHO’S bacterial priority pathogens list. Read more here: embl.org/news/science-t…

Lipid metabolism drives dietary effects on T cell ferroptosis and immunity @Nature 🇦🇺 nature.com/articles/s4158…

Today in @Nature, the @ChristophThaiss & @MaayanLevyPhD labs trace age-related memory loss to the gastrointestinal tract & map the complete gut-to-brain pathway driving it. Targeting this pathway can reverse cognitive aging in mice. nature.com/articles/d4158…

Early-activated extracellular matrix proteins shape the metabolic and spatial dynamics of the kidney fibrotic microenvironment dlvr.it/TRGR8q

A signaling pathway that inhibits DNA virus replication is also protective against RNA viruses. STING–NF-κB signaling builds an influenza spillover barrier | Science science.org/doi/10.1126/sc…

"the most dangerous form of blindness is believing that your perspective is the only reality"

#AI models have their own internal representations of knowledge or concepts that are often difficult to discern, even as they are critical to the models’ output. For instance, knowing more about a model’s representation of a concept would help explain why an AI model might “hallucinate” information, or why certain prompts can trick it into responses that dodge its built-in safeguards. Researchers in Science now introduce a robust method to extract these representations of concepts, which works across several large-scale language, reasoning, and vision AI models. Learn more: scim.ag/4cybcQk

New! Online now: Modeling lipid homeostasis using stable isotope tracing and flux analysis dlvr.it/TR4Jj0

Single-cell epigenome atlas identifies kidney epithelial cells as a driver of epigenetic aging. Lake et al. (2026) build a cross-species atlas showing that aging is driven by collapse of 3D chromatin structure in tubular epithelia. Beautiful multi-omic work. biorxiv.org/content/10.648…

Honoring the complex legacy of James D. Watson, whose discovery of the DNA double helix helped found modern molecular biology and influenced @CSHL, the Human Genome Project, and generations of scientific education. Read the PNAS Retrospective: ow.ly/CAvZ50YiZUZ

Generative deep learning enables rapid and accurate prediction of disordered protein ensembles from sequence Intrinsically disordered proteins and regions (IDRs) make up roughly 30% of eukaryotic proteomes and are central to processes like transcription, signaling, and phase separation. Unlike folded proteins, IDRs don't adopt a single structure — they exist as broad ensembles of interconverting conformations. This structural plasticity is not noise; it encodes function. But characterizing these ensembles computationally has been expensive and technically demanding, and tools like AlphaFold, optimized for predicting single structures, are poorly suited to the task. Novak, Lotthammer, Emenecker, and Holehouse now present STARLING, a generative deep learning framework that predicts full coarse-grained conformational ensembles of IDRs directly from amino acid sequence — in seconds, on commodity hardware. The key insight is that IDR ensemble generation is analogous to text-to-image generation: a single prompt (the sequence) should produce many distinct, uncorrelated outputs (conformers), each consistent with that prompt. STARLING combines a variational autoencoder that compresses inter-residue distance maps into a compact latent space with a denoising diffusion model conditioned on sequence and ionic strength. Trained on nearly 12 million distance maps from coarse-grained simulations of ~50,000 natural and synthetic IDR sequences, STARLING generates 400 independent conformers in ~12 seconds on a GPU or ~20 seconds on an Apple CPU. Predictions show excellent agreement with both simulations and experimental data from SAXS and single-molecule FRET across diverse sequence chemistries and lengths up to 384 residues. Beyond ensemble prediction, STARLING enables ensemble-aware sequence search — identifying "biophysical look-alikes" across ~35 million IDRs in UniRef50 — and rapid inverse design of sequences with prescribed conformational properties, reducing design time from hours to seconds. It also supports Bayesian maximum-entropy reweighting to integrate experimental restraints. Applied to systems including the Myc transcription factor, RNA polymerase II CTD, and disordered protein complexes, STARLING generates testable hypotheses about how sequence encodes conformational behavior and function. By making ensemble prediction fast, accurate, and accessible via a simple pip-installable tool, STARLING lowers a major barrier to studying the large fraction of the proteome that doesn't fold — but still functions. Paper: nature.com/articles/s4158…

A Fourfold Male-Specific Lifespan Extension via Canonical Insulin/IGF-1 Signaling biorxiv.org/content/10.648…

What if histology moved beyond thin sections to full 3D, whole-organ, molecularly resolved maps? In this recent Nature Methods Perspective, I outline how tissue clearing, spatial-omics, and AI converge to create deep 3D histology, enabling unbiased, organism-scale biology and next-generation digital pathology. We have a bit of way to go, but the path is clear. nature.com/articles/s4159… #DISCO #Clearing #SpatialBiology #LightSheetMicroscopy #3DHistology #DigitalPathology #AIinBiology #Omics #Deeplearning #Transformers

As I mentioned, I started a really crazy project with the new OpenAI GPT-5.2 Deep Research; I created a 110-chapter textbook focused only on T cells, the immune cells I’ve studied for 35 years, & it’s over 1,000 pages long! Sharing the first 15 chapters here: tcell-textbook.netlify.app I’ll assemble and share the rest soon. Next, I plan to create similar, even more ambitious textbooks on the entire immune system, cancer, aging, and ME/CFS. Because now you can just do things!

📢New Research Highlight Cardiovascular–kidney–metabolic (CKM) syndrome is increasingly recognized as a major driver of age-related disease. This large study examines how CKM syndrome is associated with the risk of 14 different age-related health outcomes in an older population without prior cardiovascular events. 📄 Association between cardiovascular-kidney-metabolic syndrome and risks of 14 age-related health outcomes in primary prevention older population✍️ Zhen Zhou & Joanne Ryan et al. 🔗 Read the article here: rdcu.be/e2Ylu These findings underscore the importance of integrated cardiometabolic and kidney health for reducing multiple age-related risks.

How to extract data from papers for literature review in seconds? 1. Go to link.researchcollab.ai/p-u9 2. Upload 2-3 papers you already know are relevant 3. Start with clicking on one paper 4. @ResearchCollab_ do the following for the paper ✦ Provide a short overview of the paper ✦ Identify weaknesses in the paper ✦ Share papers that contrast the existing paper ✦ Extract meta data about the paper ✦ Evaluate each part (e.g., methodology) of the paper ✦ Provide AI chat to extract any other data 5. Now click on the Related Papers button at top right 6. You will get papers related to the seed paper 7. You can click on each paper and read its abstract 8. If you find it relevant, just add it to your list of papers 9. Do the same for the other 2 seed papers 10. This way you will collect all relevant papers 11. And extract data from those papers After this, analyse the data and report the findings. Try ResearchCollab today: link.researchcollab.ai/p-u9