
Computational scientist @MDAnderson | Building bioinfor-claw — local-first AI agent for daily bioinformatics analysis
Houston TX
Joined September 2017
- Tweets81
- Following81
- Followers123
- Likes168
Bioinforbricker retweeted
Over the past week, I’ve been overwhelmed by the love and support of friends, industry colleagues, and people I've never met. Thank you to everyone who has shown kindness to my family and me. It means the world to me.
I came to the United States 53 years ago, and becoming an American citizen is one of the greatest gifts and proudest moments of my life. NVIDIA has been my life’s work. Building it into the company it is today alongside my colleagues has been a privilege and an honor. Nowhere but in America could a company like NVIDIA be possible. I love our country, and I love NVIDIA.
I did not live up to the standards expected of everyone at NVIDIA, and I have made serious mistakes in my personal life. I created problems for the Board that I deeply regret. I own my mistakes completely, and I am committed to cooperating with the Board’s independent investigation.
Since my departure, I’ve seen an outpouring of calls from NVIDIA’s employees, investors, business partners, and customers for me to return as CEO. Many continue to reach out to me directly as the company navigates complex technology, market, competitive, and geopolitical dynamics. NVIDIA is a once-in-a-generation company. If the Board of Directors decides that my return is in the best interest of our shareholders, our employees, the company, and America, I would be honored to answer these calls and return as CEO.
Thank you again to everyone who has shared their good wishes. I am profoundly humbled by your support. To my family: I love you, and I will spend the rest of my life being worthy of yours.
Jensen
Big resources
Single-cell profiling of nearly 2,000 people brings scientists closer to understanding the link between genetic variation, gene expression and health
go.nature.com/4kn8ReX
18
Bioinforbricker retweeted
this might be my last normal tweet before biotech goes vertical
i'm not a whistleblower. I don't work at Lilly, Novo or any of the big labs
but there is a massive shift happening in human biology right now.
it's sitting in press releases and almost nobody reads them
the last 6 months alone:
- a once-daily pill roughly doubled median survival in metastatic pancreatic cancer (13.2 vs 6.6 months)
- a one-time gene editing infusion cut hereditary angioedema attacks 87%, and 62% of patients stayed attack-free and off all other prevention for 6 months
- a one-time gene edit for cholesterol dropped LDL up to 62%, and it held for over a year
- a personalized mrna vaccine beat keytruda alone in a 1,137 patient melanoma phase 3
- retatrutide hit ~28% weight loss in a phase 3
- an ai-designed drug entered phase 3
- a pill that regrows hair passed its second phase 3
- an immunotherapy that could let some rectal cancer patients skip surgery, chemo and radiation just got fda approved
- a man lived 271 days off dialysis on a pig kidney
- human cells were caught passing chromosome-sized chunks of DNA to their neighbors, and it changes how the neighbors behave
we are on the verge of an era where the human body is a software problem and the update cycle gets shorter everyday
stick around, it's going to get weird
bio/acc
A powerful gene-editing technique has been used to edit DNA directly in a person’s body — the first known instance of such a treatment.
go.nature.com/4iC2gMn
Bioinforbricker retweeted
When you knock out a gene in a tissue, the effect starts in the perturbed cells and extends into their neighborhood. The same knockout can produce one response in a tumor core and another at a tissue border
CONCERT is a generative model that predicts how perturbation effects propagate across tissue
biorxiv.org/content/10.1101/…
Cells in a tissue act in concert, and CONCERT predicts how the whole ensemble responds when one player changes its part 🧵
Bioinforbricker retweeted
Delighted to share a landmark preprint from our lab, using genome language models and AI agents for biological discovery and genome mining!
The stretches of DNA between protein-coding genes are often called the genome’s “dark matter,” but they are far from empty. They encode RNA molecules that sense metabolites, regulate genes, and control some of molecular biology's most important tools (such as programmable genome editors e.g. CRISPR/bridge)
Much of this biology is unexplored. One reason is that RNAs can retain similar structures and functions even as their DNA sequences change dramatically, making their relationships difficult to recognize with conventional computational tools
When I wanted to learn something new today, I asked Claude to make a movie. What a time to be curious!
The history of bioinformatics, from sequences to AI, in four minutes. An AI-assisted animated overview.
#Bioinformatics #AIinBiology #Claude
1
36
Bioinforbricker retweeted
AI-designed proteins can now expand the CRISPR toolbox beyond what evolution has produced.
In a recent Science study, researchers report the design of synthetic RNA-guided nucleases that match or exceed the activity of natural enzymes, revealing how structure-guided design can generate functional genome-editing proteins with substantially different sequences.
Learn more on #WorldCRISPRDay: scim.ag/4yXBW5c
We’ve acquired Orchestra Bio, a platform that helps biopharma teams coordinate the complex process of drug development.
Together, we’ll build on Rosalind, our AI models and tools for life sciences research, to streamline those operations and help new treatments reach patients sooner.
Welcome @ashokaraj_ and team!
openai.com/index/openai-to-a…
Bioinforbricker retweeted
Public preview of BioEmu2! @MSFTResearch AI for Science’s new foundation model for biomolecular structure, dynamics and function. Open weights, MIT license.
github.com/microsoft/bioemu2
Movie: BioEmu2-simulated binding of intrinsically disordered peptide PUMA to MCL-1 receptor. 🧵⬇️
Bioinforbricker retweeted
Programming biological behavior with AI-guided de novo protein design. www.cell.com/molecular-cell/fulltext/S1097-2765(26)00615-5
Bioinforbricker retweeted
An excellent overview/primer of the current state of de novo computational protein design with AI. Written by a panel of illustrious experts including Nobel Laureate David Baker, who is perhaps the most prolific and impactful protein engineer in history. nature.com/articles/s41586-0…
Bioinforbricker retweeted
✨Introducing Prompteina: all-atom protein design directly from natural language✨
- Control diverse protein properties through a single text prompt
- Higher predicted binding confidence than baselines
- First experimentally validated de novo binders reported for GM2A
🧵
Bioinforbricker retweeted
Base editing and prime editing clinical trial landscape, including today’s GSD type I clinical readout from @BeamTx.
Bioinforbricker retweeted
Today, Isomorphic Labs joins the Virtual Biology Initiative (VBI) as a founding member.
We are helping build an open resource for the global research community to accelerate how we understand and treat disease.
Read more here: isomorphiclabs.com/articles/…
Sequencing a new genome is becoming easier.Understanding where its genes are is still a bottleneck.That is why Carbon-A, released by Hugging Face Bio, caught my attention. It does not predict structures or design proteins. It returns to one of the oldest and most practical bioinformatics problems: gene annotation.Carbon-A is a 1.2B-parameter DNA model. It tokenizes DNA into non-overlapping 6-mers, reads a 98,304-bp context window, and predicts coding-sequence probabilities at nucleotide resolution on both strands. A decoder turns those probabilities into genes, mRNAs and CDS features, with GFF, nucleotide/protein FASTA and a confidence score per gene.One model is used across mammals, other vertebrates, invertebrates, plants, fungi and protists. Across 42 benchmark genomes, the authors report a macro-averaged nucleotide F1 of 0.944.Two stress tests are especially interesting.The model was trained on an older 2017 pig assembly but agreed better with the improved 2026 annotation, suggesting it learned coding regularities rather than simply memorizing old RefSeq labels. On Tetrahymena thermophila, which uses a non-standard genetic code, it still reached 0.960 nucleotide F1.The team applied Carbon-A to 48,167 GenBank assemblies from 22,617 species, producing 566 million protein-coding locus candidates. I would not read that as “566 million genes discovered.” I see it as a huge collection of structured, confidence-scored hypotheses that can be searched, compared and tested.Initial PacBio Iso-Seq experiments already support some coding regions predicted by Carbon-A but missing from RefSeq. Annotation disagreements can become places to investigate, not simply errors to hide.Genome sequencing put thousands of organisms into databases. Models like Carbon-A may help turn more of those assemblies into biology we can actually study.huggingface.co/blog/HuggingF… #Genomics #AIforBiology
51
We use RNA to infer a lot of biology that is actually carried out by proteins.That shortcut is often useful—but it becomes risky when “this transcript is present” turns into “this receptor is on the cell surface” or “these two cell types are communicating.”A new preprint measured RNA and protein in the same tissue and the same cells across two cancer cohorts: ~1.77 million cells, 380 RNA targets, 27 protein targets, and 21 directly matched RNA–protein pairs.The result is not “RNA is unreliable.”Canonical lineage markers including CD3, CD8A, CD20, CD45, CD68 and CD163 were often concordant. But PD-1, PD-L1 and LAG-3 were not. The two modalities could identify different positive cells and different spatial patterns.The most important point is that aggregation can hide this disagreement.For PD-L1, ranking samples by mean expression gave an RNA–protein Kendall’s τ of 0.276. Ranking the same samples by Euler characteristic—a measure of spatial connectivity—gave τ = 0.007.So RNA and protein may agree on “how much” while disagreeing on “where” and “in which cells.”Why? Some reasons are biological: translation and degradation occur on different timescales; receptors can be internalized; proteins can persist after transcription stops; secreted or transferred proteins may appear away from the cell that made them.Other reasons are technical: antibody background, marker-specific sensitivity, segmentation errors, and membrane or extracellular signal crossing computational cell boundaries.The authors are appropriately cautious. This is a preprint, only 21 matched markers were tested, the two cohorts differ substantially, and neither RNA nor protein provides an independent ground truth.My bioinformatics takeaway: RNA is powerful for discovering cell states and pathways. But if a claim depends on where a receptor, ligand, immune checkpoint, or cell-cell interaction exists, orthogonal protein validation should be part of the claim—not an optional final panel.RNA and protein are related maps. They are not always interchangeable maps.doi.org/10.64898/2026.10.05.… #SingleCell #Bioinformatics
Last week, I spent several days learning and writing about virtual cells. Today, a $1.8B announcement made one point much clearer to me:
AI biology may be reaching the stage where better data matters more than another model. Biohub, DOE, NIH, Google DeepMind, Isomorphic Labs, and Meta are expanding the Virtual Biology Initiative to generate the measurements needed for predictive models of cells.
The scale is striking. Existing datasets contain hundreds of millions of cells; Biohub argues that reliable models may eventually need billions or even trillions of measurements.
But cell count is only one dimension.
A useful virtual cell needs perturbations across cell types, donors, doses, and time points. It needs RNA, proteins, spatial context, morphology, imaging, and cell-cell interactions. It also needs provenance and shared standards. Otherwise, a larger dataset may simply amplify batch effects and platform-specific shortcuts.
That is why the less glamorous parts of this initiative may matter most: coordinated experiments, common identifiers, interoperable metadata, and a unified access layer.
There is also a tension worth watching. The project is presented as an open resource, but Reuters reports that commercial funders will receive an embargo period before some data become public. That may attract private investment, but “eventually open” is not the same as “equally usable.”
I do not think $1.8B means a faithful virtual cell will suddenly appear. Biology still contains hidden variables, noisy measurements, and major gaps in causal coverage. But this initiative acknowledges something important: AI biology cannot keep extracting progress from the same public datasets forever.
The most expensive part of the next biological foundation model may not be the training run. It may be the experiments that make the model worth training.
biohub.org/news/virtual-biol…
#AIforBiology #VirtualCells #Bioinformatics
1
93
We would expect tons of cell data for better modeling of cell response to drugs and perturbations
🧵 Biohub, @ENERGY, @NIH, and new funding partners today announced a major expansion of an international effort to generate and make accessible the data enabling predictive AI models of biology. Together, the organizations are investing $1.8 billion in funding, data, computation and new measurement technology - largest coordinated commitment to map the biology of the cell to power development of digital models of biology. bit.ly/3UhjoxD
40



















