Computational Structural Biologist |
Joined April 2017
- Tweets473
- Following391
- Followers715
- Likes3.5K
sk retweeted
Today we are excited to release Artios, a fully open source system for designing functional de novo enzymes.
Enzymes present a new frontier for de novo protein design models due to the challenge of coordinating precise chemistry, rather than binding a target.
Enzymes are life’s biological machines and by designing them we can truly start to engineer biology.
We believe these fundamental capabilities should be built in the open and shared with the community to progress science faster.
🧵 1/6
sk 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. 🧵⬇️
Nice! Others are;
-LenioBio
-GeneScript
-SinoBiological
Today we are launching RootPath BinderVerse, a cloud lab for AI-designed binders.
Sequence in, binding data out in as few as 3 days, gene synthesis time included! Antibodies and Fab fragments made with CHO cells (TAT from 6 days) and CAR-Ts made with primary T cells (TAT from 10 days) can also be created from your binder sequences and tested.
Trustable data at lightning speed -- Learn more at binderverse.rootpath.com
I also remember MP3/MP4 from @310ai__ doing this two years ago
ipSAEmin - i see it everywhere, each new paper, preprint or tool in binder design. I haven't been able to find a single study demonstrating that it correlates with affinity
sk retweeted
Introducing AlphaProtein Novo, a pipeline for de novo enzyme design! In collaboration with @francesarnold's group, we developed AP Novo and used it to get SoTA "novel-scaffold" activity on 2 benchmark reactions and create enzymes to synthesize the pharmaceutical motif piperidine and degrade the environmental toxin DEHP.
Work led by @ZvxyWu @ZiqiLi0513 with teammates @alexechu_ @kelly_ruijie @schulluc @jacobjinkelly @rosaliags @davidla @solserpens @j_reisenbauer, @pengliu_group, Josh, Thomas, Amy, Tristan, Harshnira, and others. It was a joy and privilege to work with you all!
Preprints and code:
biorxiv.org/content/10.64898…
biorxiv.org/content/10.64898…
github.com/google-deepmind/a…
sk retweeted
Biosecurity is too important for security theater.
In a quick test with Raygun, we found that DeepMind's SynthIDBio watermark signal could be washed out while preserving the predicted structure.
We don’t think protein sequence watermarks buy much biosecurity.
sk retweeted
This is the worst idea since NFTs
Very happy to announce that our team @GoogleDeepmind has pushed the boundaries of generative biology, achieving the successful synthesis of AI-designed proteins that are both functional and watermarked.
This proof-of-concept watermarking of the building blocks of life is enabled by SynthID Bio, our new protein watermarking method. It is designed to safeguard the new era of AI-powered generative biology and strengthen global biosecurity.
You can read my thoughts here on why watermarking AI-designed proteins is an important research breakthrough: xitter.cf/pushmeet/status/210531…
Assume --- three of your group members
- Designed binders
- Using different techniques/tools
- For folding they used Boltz → Protenix → ColabFold
- For interface scoring/ranking somewhat similar metrics.
You/PI: "can you guys also rank and check with i_pDAE, in BindCraft2. I just want to see how our binders rank/look"
-One of them simply clones the repo👇& sets it up in 3 min using uv.
-They paste the folded binders from Boltz → Protenix → AF2/ColabFold into one folder and run one command: "python ipdae.py designs/"
-By the time they finish their coffee☕️, the results are ready.
-One ranked CSV file, best binders(0-1 i_pDAE scores) first. For every seed/structures user get (i_pDAE, i_pDAE_median, Contact_Pairs, interface_Binder_Residues/interface_Target_Residues).
-Plain-text report.
-PyMOL views with F1–F3: ipdae_view.pml stores three scenes on the F1, F2 & F3 keys.
i_pDAE values/colors/residues/surfaces in pymol/clip👇:
🔴 Deep Red (0.8–1.0): High confidence (PAE <0.5Å) 🌸 Light Red (0.6–0.8): Confident
⚪ White (0.4–0.6): Moderate / mixed
🩵 Light Blue (0.2–0.4): Uncertain
🔵 Deep Blue (0–0.2): Interface contact, unknown placement
🩶 Grey (-1): Off-interface (unscored)
Dropping it tomorrow.
github.com/ullahsamee/i_pDAE
BindCraft2 - LICENSE
-Selling binders "you designed" with BindCraft2: allowed
-Selling access to BindCraft2 as a "hosted designer service": not allowed
github.com/PacesaLab/BindCra…
ʙɪɴᴅᴄʀᴀꜰᴛ2 is out, and we're not waiting for the paper. The full code drops today, free for academic and industry use.
We're releasing it early so you can start designing right now, and bring its full power to the current Adaptyv competition.
github.com/PacesaLab/BindCra…
Using cms, we screened cyclic peptides (1st image) & binder (last image) in single & double-digit nM affinities. In both cases we dropped sc and preferred deltaG. Both showed promising in vitro and in vivo activity.
GPU-accelerated (CUDA) contact molecular surface (CMS), Lawrence & Colman shape complementarity (SC) for protein design.
Originally developed by Cao, code opensource by Brian. The original NumPy code is unchanged and utilize CPU. The GPU/CUDA version gives the same results while 36-165× faster per complex. cms-cuda is built on bcov77/py_contact_ms(github.com/ullahsamee/cms-cu…)
GPU-accelerated (CUDA) contact molecular surface (CMS), Lawrence & Colman shape complementarity (SC) for protein design.
Originally developed by Cao, code opensource by Brian. The original NumPy code is unchanged and utilize CPU. The GPU/CUDA version gives the same results while 36-165× faster per complex. cms-cuda is built on bcov77/py_contact_ms(github.com/ullahsamee/cms-cu…)
Visualization(cms/sc) experimental features in PyMOL
target surface = CMS per atom, white 0 → red ≥15 Ų
target surface = local fit: purple good · gray neutral · orange poor · white not in interface
Use it for binders screening if you're participating in @AnthropicAI protein design competition @amirshanehsaz
esmfold2 - i wanted to capture diffusion sampling coordinates for trajectory animation similar to @sokrypton at localfold.org/ so using Opus5.5 in 2-3 prompts, its not bad.
i_pDAE is Dunbrack’s ipSAE recipe with one ingredient swapped.
i_pDAE in BindCraft2(contact-masked TM-of-PAE), copies ipSAE’s three structural ideas - restrict the pairs, shrink d0d_0d_0 ,take the best residue in both directions - then replaces Dunbrack’s PAE cutoff with an 8 Å Cα contact cutoff.
So i_pDAE is ipSAE’s recipe with the gate changed from “does AF trust this pair?” to “are these two Cα atoms touching?
github.com/PacesaLab/BindCra…
ʙɪɴᴅᴄʀᴀꜰᴛ2 is out, and we're not waiting for the paper. The full code drops today, free for academic and industry use.
We're releasing it early so you can start designing right now, and bring its full power to the current Adaptyv competition.
github.com/PacesaLab/BindCra…
sk retweeted
ColabFold2 Sneak Peak - integrates: alphafold3, openfold3, intellifold2, protenix2, boltz2, opendde, rosettafold3, chai1, esmfold2, openbind0 (thx Julia Buhmann), fully in Jax!
🗒️ Colab Notebook colab.research.google.com/gi…
🐍 pip install gist.github.com/sokrypton/75…
(2/3)
