Design new binders and medicines from a prompt and hand it to the robo lab.
- Tweets 154
- Following 8
- Followers 1,363
- Likes 206
Pinned Tweet
Meet BIOS, an AI Scientist built to orchestrate complex biomedical research.
• Global SOTA on Data Analysis Benchmarks: BixBench 48.78% open-answer, 55.12% multiple-choice + refusal, 64.39% multiple-choice (no refusal) - outperforming systems like Edison Scientific and Kepler.
• Human-in-the-Loop or Autonomous Mode: Intermediate checkpoints let researchers guide investigations mid-flight as insights emerge. No more waiting hours for batch runs + reruns to get results. Or, run in fully autonomous mode for extended investigations.
• Persistent World State: Rather than losing context as conversations grow, world state ensures investigations build on insights within each research cycle and across sessions.
• Subagent Swarm: BIOS orchestrates subagents specializing in research functions (Literature Review, Data Analysis, Novelty Detection) and, soon, research domains (microbiology, longevity, genomics).
BIOS is available now in Beta with free + paid tiers, exclusive launch pricing and, for limited time, free full access to academic users with a .edu email address.
Pro, Researcher and Lab subscription tiers offer discounted packages on monthly credits. Our usage-based pricing is competitive and in some cases significantly cheaper than leading scientific agents.
Try BIOS and read our paper in the links below ↓
12
38
186
69,694
More researchers should be able to help design potential drug candidates.
Anthropic × Adaptyv’s protein design competition is an invitation to do that.
It’s why we’re building BIOS: to help researchers design and shortlist computational binders for testing in the lab.
We’re partnering with @Anthropic to launch the biggest Protein Design Competition in the world, challenging people around the world to use AI to design new potential drug candidates for diseases that affect millions of lives.
The competition will feature five challenges, each focused on a specific disease or biological mechanism. Compared to previous competitions, it will be a big step-up in complexity and scale to push the boundaries of AI-driven protein design.
Together with Anthropic, we’re sponsoring over $1 million in experimental validation, making it possible to test more than 5,000 protein designs in our automated lab at no cost to participants. Anthropic is providing an additional $1 million in Claude credits.
All experimental results will be published openly on @Proteinbase, including designs that didn’t work, so anyone can access the data and build on what we learn.
The competition is open to everyone and free to enter. It will feature 3 tracks:
- Track 1 is aimed at expert protein designers, with up to 20 teams to be selected.
- Track 2 is targeting life science academics and industry researchers.
- Track 3 is open to everyone from tech enthusiasts to high-school students.
By combining Anthropic’s models with access to our automated lab, we want to make it possible for anyone with a laptop and an internet connection to join the global effort to advance human health with AI.
A big thanks to @Modal for contributing compute for protein design and to @TwistBioscience for contributing the DNA for the experimental validation!
Sign up link below -
1
1
7
712
BIOS retweeted
How good is Claude at protein design?
Anthropic benchmarked their newest Claude models on protein engineering tasks, and we ran the wet lab work behind it.
5
16
69
10,560
🫡
BixBench3 is out! Our latest eval for biology capabilities in language models. Essentially requires agents to reproduce all work needed to write a paper from raw data. OpenAI is in the lead with 5.6-Sol at 48%, Anthropic is in 4th with Opus 4.8, after Kimi and GLM. Opus 5 has weird instruction following problems that severely degrade its performance. I’m sure Fable is great but can't be evalled on this. Read more below.
Also lest we miss the forest for the trees, we’re in a position now where the models can essentially reproduce entire papers ~50% of the time. Crazy time to be alive.
1
5
429
BIOS retweeted
There is a child with a rare disease who is currently suffering and struggling to manage his symptoms. Rare as this is, you can directly help him.
Today we are launching the "Rare Disease, Real Kid" Hackathon, and there are $50,000 in prizes from @AnthropicAI and @awscloud.
We (@huggingface & @Sagebio) are helping this child open his genome and clinical data to the community, so that we can find what's caused his disease and what currently-approved drugs could help him.
I doubt I need to motivate this much further or explain how rare it is for a family to share their child's genome and clinical data, but if you're not sure, consider this:
Until very recently, it wasn't feasible for patients like this to get treatment because their disease was so rare that the economics could never justify the investment. Now, as we've seen, people with rare diseases are starting to be able to find the answers themselves (with the help of AI tools, cheaper sequencing, etc). This kid is not able to do that for himself and neither are his parents, so we're asking you for help. Both for this kid and to prove that it's possible for everyone else suffering from a rare disease.
More details in 🧵.
sagebio-rare-disease-real-ki…
124
495
3,189
664,721
BIOS retweeted
This is freaking huge: For the first time, a AI-assisted personalized mRNA cancer treatment has succeeded in a Phase 3 trial.
Moderna and Merck sequence each patient’s tumor and compare it with their healthy DNA. AI then helps identify which of the tumor’s mutations are most likely to trigger an immune response.
From those targets, Moderna produces an individual mRNA treatment encoding up to 34 neoantigens, The earlier Phase 2 trial followed patients with surgically removed high-risk melanoma for five years:
- 68.8% remained cancer-free, compared with 49.1% on Keytruda alone
- 49% lower risk of recurrence or death
- 59% lower risk of distant metastasis or death
Now the much larger Phase 3 trial involving 1,137 patients has also succeeded.
Absolutely incredible. Dario was right: cancer will be cured in just a few years!
190
711
6,500
1,493,833
The demand for automated wet-lab validation will increase significantly.
Using NVIDIA Proteina-Complexa, @muni_bio's autoresearch agent explored nearly 14,000 protein designs.
@adaptyvbio validated nine TREM2 binders, three with sub-nanomolar affinity.
The results show how agents can connect computational design with wet-lab feedback to improve the next round of discovery. 📘muni.bio/research/closing-th…
1
3
18
1,295
Why scientists should lead the shift away from AI mega data centres?
The article by @stokel for @Nature makes the case for publicly available models and local infrastructure that give researchers more control over the tools they use.
That focus on access and researcher control also sits behind our work on BIOS: opening the binder-design workflow so researchers can run it themselves, from target structure to ranked candidates.
An AI model has revealed that some trusted numbers in one reference database have been wrong all along
go.nature.com/45KjsIe
5
9
2,005
AI-designed medicine is accelerating.
Soon you will design a medicine the way you write a prompt.
That future has a name: BIOS ( @bioaidevs). Type in a target, it generates real drug candidates that already beat the state-of-the-art models.
4
4
26
1,394
Most researchers want AI in their work but face the same issues: setup is a mess, and the parts that matter most are the ones you can't hand off without looking.
That's why we're building a research agent, and asking how you actually do research and what it should never do without checking with you first.
Take the survey before it closes Monday, Aug 2 ↓
9
8
66
2,515
🔗 Take the survey: docs.google.com/forms/d/e/1F…
197
Calling all researchers and AI builders!
We're building a research agent for the Bio community, similar to BIOS but configured around your domain, your evidence standards, and your workflows.
Before we ship it, we want to hear what it should look like from the people who'd use it.
Tell us how you do research: your field, your sources, and what the agent should never do without your sign-off.
Link below ↓
7
10
63
14,528
🔗 Take the survey: docs.google.com/forms/d/e/1F…
Or reach out to @oxMichelangelo to go through it together on a call.
3
344
A new drug candidate. Under an hour. Under $7.
Drug discovery used to mean batch experiments, days of waiting, and most candidates dying before they ever reached a lab. Nothing connected computational design to physical testing in one loop.
BIOS closes that loop.
Three engines (RFdiffusion3, BoltzGen, and PXDesign) run in parallel on every campaign, each attacking the design problem differently. Together they generate thousands of candidates per run.
A dedicated pipeline then filters them through six gates: Structural, Agreement, Physics, Selectivity, Dynamics, Developability. What survives is what's most likely to bind.
Every candidate is scored on how it should perform in a real experiment, not on paper.
The top candidates go to the wet lab. A robotic system synthesizes them, tests how well they bind, and feeds the results back to the models. What worked and what didn't sharpens the next run, so every cycle starts stronger than the last.
5
12
48
3,557
BIOS retweeted
Zero-shot design builds a protein that binds a chosen small-molecule drug without a solved structure of that drug, and @BioAIDevs now runs the full method as a guided workflow.
A drug goes in, its key atoms get marked, and a design campaign returns a binder built to hold it 🧵
1
6
14
1,964
Zero-shot design builds a protein that binds a chosen small-molecule drug without a solved structure of that drug, and @BioAIDevs now runs the full method as a guided workflow.
A drug goes in, its key atoms get marked, and a design campaign returns a binder built to hold it 🧵
1
6
14
1,964
Under the hood the campaign runs several models in one loop.
CARPdock places the drug into the recommended scaffold library, drawn from the 40 four-helix bundles and 50 NTF2 folds that ship with the build.
LASErMPNN designs sequences for each pose, Boltz-2x co-folds every candidate to test where the drug lands, and only the self-consistent survivors advance. The loop runs up to 35 rounds and stops early once the designs plateau, all on Modal GPUs with checkpointing so a preempted job resumes instead of restarting.
Integration is what moved this out of a research repo. The design stack used to mean cloned code, GPU orchestration, and reading confidence tables by hand, and BIOS collapses that into a campaign that starts with a few clicks. The scientific call, which atoms to grip, stays with the person, and everything downstream of it runs itself.
1
121
The same design opens three commercial directions.
An ADC payload wrapper grips a cytotoxic drug like deruxtecan and shields its reactive group from breaking down in circulation, which keeps the drug active for longer.
A reversal agent binds a circulating drug tightly enough to switch it off, the unmet need behind a fast antidote for anticoagulants like apixaban.
A biosensor uses the designed protein as the recognition element that reports how much drug is present, the basis for at-home monitoring.
Design used to start from what was already known about a molecule. It now starts from the molecule itself.
113