Pinned Tweet
Today I'm announcing a protein compute web service: subseq.bio
It's a simple web tool to submit compute-intensive tasks related to protein design, and biotechnology jobs.
You can select from state-of-the-art, open source protein ML models, pre-configured and can begin running within minutes. Currently supported are RFdiffusion, and Alphafold (v2), with more coming.
There are no restrictions on your output data. Just select the program, enter program arguments, and an optional upload folder, and click submit. Download the output data when finished.
It is pay-per-use: Jobs are charged from your account credits when submitted. You can top up credits easily, any time. Currently jobs cost between 5¢ - $1.
Feedback is important! Let me know ways it can be improved. More features are coming. For detailed updates please follow @subseqbio
ALT synthetic protein 51
Here's a chance to try serious biomolecular design work. This is enough credits for multiple full design campains, or hundreds of biomolecular design jobs!
BioEmu2 available on @subseqbio! I'll be playing around with some big complexes
BioEmu2 is now available on web, API, and MCP!
Equilibrium ensemble sampling for proteins, nucleic acids & ligands — up to ~1,000,000× faster than traditional md simulation
github.com/microsoft/bioemu2
Emin Bursa retweeted
Got a molecular design idea? Send it over and have all work done for you, from concept to ranked candidates.
Use the new form, or DM here or @0xCF88
Emin Bursa retweeted
Have you used local AI? (open weight LLMs on your own hardware) If so how much
22%nope
57%tried it
22%use it often
0%more than claude/gpt/etc
23 votes • Final resultsHave you used local AI? (open weight LLMs on your own hardware) If so how much
22%nope
57%tried it
22%use it often
0%more than claude/gpt/etc
23 votes • Final results
Emin Bursa retweeted
I'm building a biomolecular design platform @subseqbio It's awesome because it provides latest bio tools exposed to both humans and agents. The interface is geared for de novo protein design and easy monitoring of bio agents. (It's also fast and fairly cheap)
OpenAI has decided to fully shut down a protein design project I'm working on for a client.
Needless to say, open weight models are the only way forward.
Interestingly, bioemu takes msa's and clearly improves results. The AI agent decided to experiment with bioemu settings: msa and physical steering for villin (1VII), via glm-5.2
Emin Bursa retweeted
🧬 Protenix-v2 is now available on subseq
A high-accuracy all-atom cofolder for challenging antibody–antigen complexes, #3 on CoFold Arena, with optional physics-guided protein–ligand sampling.
github.com/bytedance/Proteni…
Available over web, API & MCP
I built a detailed nutrition logger android app, and open sourcing it! Take pictures and/or notes, auto-log 20+ macros/micros, fully customizable with your own chat completions AI endpoint.
github.com/oxpsi/omnitracker
1/n
The accuracy is going to heavily depend on what AI model you choose to use, even smaller open weight models can be quite accurate, but the main weakness is total quantity estimation, so you'll get much better results if you give any sort of quantity reference per log. 4/n
