@spantham1
Account based inUnited States
About this account
- Account based in
- United States
- Connected via
- Web
Account-level information from X, not a live location or the device used for a specific post.
Sairam Pantham retweeted
Thrilled to share our latest work, Bio-Babel: a multi-agent framework for native cross-language software reconstruction.
We reconstructed 17 visualization and classic single-cell analysis packages from R in Python, spanning foundational visualization tools such as grid/ggplot2 and widely used single-cell tools including Monocle 2/3, tradeSeq, NicheNet, and copykat. The reconstructed stack features native AnnData/scverse integration and agent-readable MCP contracts.
Using this stack, an AI agent recovered pancreatic differentiation trajectories and helped reveal defects under graded SWI/SNF loss.
Bio-Babel also extends beyond R to Python translation: its C++ reconstruction of UMI-tools reproduced identical barcode and UMI groupings with higher efficiency.
Bio-Babel offers a path to preserve classic methods, bring them into modern ecosystems, and transform them into building blocks for agent-driven science.
Paper: doi.org/10.64898/2026.08.30.…
Website: biobabel.stanford.edu
Code: github.com/Bio-Babel
Work led by my incredible post-doc Nianping @nianping_liu , contribured by Xuanzhi Chen @ProtectedTrash (who is applying for PhD program, please keep an eye on his application), Miao Cui @miao_cui0330, Xiaoying Liao, Xiaoke
Song, Sairam Pantham @spantham1 , and Weize Xu @Nanguage .
If you want your package or any classic package to speak a different tongue, please let us know!
More below:
Roche and Eli Lilly’s Elecsys pTau217 blood test just received FDA clearance. It’s the first single-biomarker plasma assay that both rules in and rules out amyloid pathology associated with Alzheimer’s in people 55+ with cognitive symptoms—using the same validated cutoffs in primary and specialty care. High agreement with amyloid PET, runs on existing Roche cobas instruments already in thousands of U.S. labs.
A simple blood draw can now replace (or triage) invasive CSF taps and expensive PET scans for the hallmark pathology of Alzheimer’s.
We’ve spent decades making diagnosis the bottleneck. Now that the bottleneck is gone, the real test begins: Will primary-care accessibility finally force the field to deliver affordable, disease-modifying therapies at scale—or will we just diagnose more people earlier into a system that still can’t treat them effectively?
What do you think—does democratizing the diagnosis accelerate the cure, or just accelerate the anxiety?
Link: prnewswire.com/news-releases…
Our pre-print for PantheonOS is on bioRxiv! Our multi-agent framework can automate data analysis in a transparent, organized manner that can aid PIs and researchers in automating analysis workflows and addressing the backlog in projects that many face.
This technology, directly integrating AI with the researcher, can accelerate research across all facets of biomedical research. There are so many unresolved questions that researchers are trying to answer, and bringing AI that aids researchers, as opposed to replacing them, is a major step forward in the right direction.
Check out the pre-print here: biorxiv.org/content/10.64898…
A kidney transplant patient with progressive, multidrug-resistant E. coli malakoplakia was treated with SNIPR001—an engineered CRISPR-armed bacteriophage cocktail. Abdominal lesions healed rapidly; the intra-abdominal mass shrank ~89% at one year with no phage-related adverse events. First reported use for this rare intracellular infection.
CRISPR-enhanced phages successfully targeted an antibiotic-resistant intracellular pathogen that conventional drugs could not clear.
The antibiotic resistance crisis may not be solved by discovering the next small-molecule drug. Precision viruses that edit and kill specific bacteria could become the more scalable, evolution-resistant approach—especially for the growing number of immunocompromised and transplant patients where broad-spectrum antibiotics keep failing.
Link: healio.com/news/infectious-d…
Amylyx’s avexitide, a first-in-class GLP-1 receptor antagonist, cut Level 2 and Level 3 hypoglycemic events by 55% versus placebo in the Phase 3 LUCIDITY trial for post-bariatric hypoglycemia (PBH) after Roux-en-Y gastric bypass. No weight change and a favorable safety profile; NDA planned by year-end.
This is the first potential approved therapy for a debilitating condition that affects tens of thousands after weight-loss surgery and currently has zero FDA-approved options.
Blocking GLP-1 signaling to treat a complication of bariatric surgery feels like a quiet rebuke to the “more GLP-1 is always better” narrative. It forces the field to admit that exaggerated hormone responses can become pathological—and that the same pathway we’re amplifying for obesity can need deliberate damping in other patients.
Link: reuters.com/business/healthc…
European regulators have recalled all batches of Amgen’s Tavneos (avacopan) after the European Commission canceled its marketing authorization. The decision followed findings that the pivotal trial data were incorrect/misleading and that serious liver injuries (including fatalities) shifted the benefit-risk balance.
A once-approved rare-disease vasculitis drug is being pulled across the EU because the original evidence base no longer holds.
When post-marketing safety signals and data-integrity problems force a full market withdrawal years later, it exposes a deeper tension: rare-disease drugs often get approved on imperfect or limited evidence precisely because the patient population is tiny. The uncomfortable truth is that some of these therapies will always carry higher residual uncertainty—and regulators are finally treating that uncertainty as a feature, not a bug, that can justify removal.
Link: bbc.co.uk/news/articles/c5ye…
I just saw a news article about the 2nd World Humanoid Games, where a robot just ran faster than Usain Bolt's 2009 world record.
Honestly, it shouldn't even be surprising that robots can perform repetitive, physical tasks better than humans.
But what if we can give robots cognition? With the amount of work being done with AI, replicating the neural networks that make us human, and transplanting that into physical hardware, the possibility of "real" humanoids is right around the corner.
It's ambitious, awe-inspiring, yet scary. We live in a world where photos and videos are often no longer "real". Will we soon have to ask whether the person we see in front of us is real, too?
Capricor’s deramiocel is staring at its PDUFA date today (or an extension). The FDA advisory committee already voted 9-3 that the available evidence does not support effectiveness for Duchenne cardiomyopathy, even after HOPE-3 data were published in The Lancet and the company narrowed the proposed indication.
Hot take from the actual evidence: this is the FDA correctly refusing to lower the bar just because the unmet need is massive. Cell therapies still have to prove clear, reproducible benefit on pre-specified endpoints — not just “promising signals” or post-hoc analyses.
Patients deserve treatments that work. Lowering standards to get something approved faster helps no one in the long run. The right response is better trials and clearer data, not softer review.
Details from Capricor’s official update: capricor.com/investors/news-…
David Baker’s GenBio AI just dropped AIDO Cell — a virtual whole-cell simulator that aims to model cellular behavior, then predict what happens if you knock out a gene or drop in a drug.
Think AlphaFold, but for the entire “city” of the cell instead of one protein “house.”
If it holds up, the potential is enormous: faster hypothesis generation, better perturbation predictions, and tighter loops between computation and wet lab. But the model is only as good as the underlying experimental data that trains it.
The real upside of AI in biology isn’t replacing experiments — it’s forcing us to generate cleaner, more systematic data so the simulators actually work. That’s the part that excites me most as someone just starting out.
This week the FDA handed Ultragenyx accelerated approval for the first gene therapy for glycogen storage disease type Ia, and Regeneron got the nod for Pasatru (garetosmab) in fibrodysplasia ossificans progressiva — the second drug ever for that ultra-rare progressive disease.
Two approvals in days for conditions that used to have almost nothing. Gene therapy is quietly stacking real wins in the rare-disease space while the rest of the industry chases bigger markets.
As someone watching from the basic-science side: these feel like proof that years of careful vector and manufacturing work are finally translating. What’s the next rare disease you think is closest to a gene-therapy breakthrough?
pantheonos.stanford.edu/blog…
We're 3 weeks into our PantheonOS Researcher Access Program! We're beta testing our platform with researchers at Harvard, Duke, Stanford, and international institutes, and we're already receiving great feedback! Our platform is already being used to save time, organize projects, and create valuable, scientifically-rigorous analyses for our researchers. Check it out now!
Moderna + Merck’s personalized mRNA vaccine just cleared a major Phase 3 hurdle in high-risk melanoma. After surgery, it helped train patients’ immune systems to hunt their own tumor mutations and significantly cut the risk of the cancer coming back or spreading (when paired with Keytruda).
That’s genuinely huge — real proof we can teach the immune system to fight a patient’s specific cancer.
But mainstream headlines calling this a “cancer vaccine” are dangerously misleading. This is not a shot that prevents cancer in healthy people or works against every tumor. It’s a custom, post-surgery treatment for one cancer type, built from each patient’s own mutations. Clarifying that distinction matters, and we once again have to ask ourselves how the reality of science can be accurately disseminated to the public.
nature.com/articles/d41586-0…
True, AI is a selective pressure on researchers.
Ignore the tools → you fall behind.
Master them → you move faster.
But the real edge still belongs to the people who can generate the data AI can’t invent.
Years of meticulous, manual experiments are the actual training set. Without that foundation, every “AI scientist” system is just rearranging noise.
As someone just starting, I’m less worried about AI replacing us and more worried about us starving it of real basic science.
What does the ideal lab look like in 2030 — fully automated, or hybrid with humans still generating the hard data?
The worst part of research is writing. I've done all the work, gotten all the results. Why do I have to spend the same amount of time just getting it down on paper?
Sairam Pantham retweeted
Help us build the Virtual Embryo community: virtualembryo.ai/challenge, and get rewarded for your contributions!
A great challenge should create more than a leaderboard. It should leave behind open tools, knowledge, and infrastructure that everyone can build upon.
That’s why the Virtual Embryo Challenge launched $20,000 in Community Contribution Awards, with up to $200 awarded per contribution.
And yes: you can make multiple contributions and receive multiple awards!
You don’t need to win the competition. We want to recognize people who help make Virtual Embryo more accessible, useful, and reproducible.
Community contributions can take MANY forms: 🧵👇
We are thrilled to announce the inaugural Stanford Virtual Embryo Challenge at NeurIPS 2026 @StanfordAILab @NeurIPSConf , in collaboration with Laude Institute @LaudeInstitute , UCSD, and Harvard!
Join us in building AI models that predict how life takes shape, across space, scale, time, and perturbation: virtualembryo.ai/challenge.
We welcome participants from academia and industry, independent researchers, and even AI agent scientists!!!
Our vision is inspired by and complementary to the virtual-cell efforts championed by @arcinstitute . But life is more than a collection of individual cells.
Why Virtual Embryos? See below 🧵👇
I'm recirculating some of our old videos about PantheonOS and its development. We've included a lot of new features into the platform, including an enhanced UI and the capability to include multiple machines in your computation in Pantheon-Fleet. Check out PantheonOS at app.pantheonos.stanford.edu/.
What do investors look for in biotech startups? Especially ones that are actually selling a software? Do they become like your typical SaaS?
For all the PIs and postdocs, what is the standard for undergrads in labs? What makes an undergrad a good researcher?
app.pantheonos.stanford.edu
This is what I've been working on with the Qiu Lab at Stanford Medical School, an automated data analysis platform built on a multi-agent framework that can take your data and your study goals to conduct scientifically-rigorous, accurate analysis. Think Cursor for research. Check it out now!!