Alternate between predicting structure with AF3 and designing sequence with ProteinMPNN to generate proteins, including protein, small molecule and nucleic acid binders
@yehlincho.bsky.social @sokrypton.org
www.biorxiv.org/content/10.1...
@yehlincho.bsky.social @sokrypton.org
www.biorxiv.org/content/10.1...
October 15, 2025 at 10:01 PM
Alternate between predicting structure with AF3 and designing sequence with ProteinMPNN to generate proteins, including protein, small molecule and nucleic acid binders
@yehlincho.bsky.social @sokrypton.org
www.biorxiv.org/content/10.1...
@yehlincho.bsky.social @sokrypton.org
www.biorxiv.org/content/10.1...
Our most practical episode yet for people who want to dabble in biotech
As of 2022, you can hallucinate protein structures using AI similar to Midjourney (RFDiffusion) -> create amino acid strings for them (ProteinMPNN) -> validate with AlphaFold
Biology becomes engineering…
As of 2022, you can hallucinate protein structures using AI similar to Midjourney (RFDiffusion) -> create amino acid strings for them (ProteinMPNN) -> validate with AlphaFold
Biology becomes engineering…
New episode of Hard Drugs!
What if you could design a protein never seen in nature?
Scientists are using new AI tools like RFDiffusion, AlphaFold & ProteinMPNN to hallucinate novel proteins to solve problems nature hasn't.
@jacobtref.bsky.social & I talk about the art of protein design 🧑🎨
What if you could design a protein never seen in nature?
Scientists are using new AI tools like RFDiffusion, AlphaFold & ProteinMPNN to hallucinate novel proteins to solve problems nature hasn't.
@jacobtref.bsky.social & I talk about the art of protein design 🧑🎨
The art of protein design with AI
YouTube video by Works in Progress
www.youtube.com
October 15, 2025 at 5:02 PM
Our most practical episode yet for people who want to dabble in biotech
As of 2022, you can hallucinate protein structures using AI similar to Midjourney (RFDiffusion) -> create amino acid strings for them (ProteinMPNN) -> validate with AlphaFold
Biology becomes engineering…
As of 2022, you can hallucinate protein structures using AI similar to Midjourney (RFDiffusion) -> create amino acid strings for them (ProteinMPNN) -> validate with AlphaFold
Biology becomes engineering…
The problem:
AI tools like AlphaFold & ProteinMPNN accelerate design of stable protein folds by inverting the sequence-structure map.
But IDPs don't have 1 shape - they occupy a huge ensemble of shapes. Physics simulations are good models to generate ensembles but hard to design/invert over!
AI tools like AlphaFold & ProteinMPNN accelerate design of stable protein folds by inverting the sequence-structure map.
But IDPs don't have 1 shape - they occupy a huge ensemble of shapes. Physics simulations are good models to generate ensembles but hard to design/invert over!
October 10, 2025 at 6:16 PM
The problem:
AI tools like AlphaFold & ProteinMPNN accelerate design of stable protein folds by inverting the sequence-structure map.
But IDPs don't have 1 shape - they occupy a huge ensemble of shapes. Physics simulations are good models to generate ensembles but hard to design/invert over!
AI tools like AlphaFold & ProteinMPNN accelerate design of stable protein folds by inverting the sequence-structure map.
But IDPs don't have 1 shape - they occupy a huge ensemble of shapes. Physics simulations are good models to generate ensembles but hard to design/invert over!
We're building the complete research ecosystem, starting with:
• Structure prediction (Boltz-2, Chai-1, ProteinMPNN)
• Database integration (PDB, UniProt, ChEMBL)
• Digital notebooks & file management
Natural language for everything biology!
• Structure prediction (Boltz-2, Chai-1, ProteinMPNN)
• Database integration (PDB, UniProt, ChEMBL)
• Digital notebooks & file management
Natural language for everything biology!
October 6, 2025 at 11:55 AM
We're building the complete research ecosystem, starting with:
• Structure prediction (Boltz-2, Chai-1, ProteinMPNN)
• Database integration (PDB, UniProt, ChEMBL)
• Digital notebooks & file management
Natural language for everything biology!
• Structure prediction (Boltz-2, Chai-1, ProteinMPNN)
• Database integration (PDB, UniProt, ChEMBL)
• Digital notebooks & file management
Natural language for everything biology!
Every week brings groundbreaking AI advances in biology - Boltz-2, ProteinMPNN, and more.
Most biologists can't access these tools due to complex CLI, GPU requirements, and extensive setup.
This results in transformative technology being inaccessible to researchers.
Most biologists can't access these tools due to complex CLI, GPU requirements, and extensive setup.
This results in transformative technology being inaccessible to researchers.
October 6, 2025 at 11:55 AM
Every week brings groundbreaking AI advances in biology - Boltz-2, ProteinMPNN, and more.
Most biologists can't access these tools due to complex CLI, GPU requirements, and extensive setup.
This results in transformative technology being inaccessible to researchers.
Most biologists can't access these tools due to complex CLI, GPU requirements, and extensive setup.
This results in transformative technology being inaccessible to researchers.
Cool episode on how AI can help us design new proteins (e.g., for vaccines).
I’d heard of AlphaFold before, but not of ProteinMPNN. You feed a protein structure into PMPNN & generate possible sequences of amino acids. And then you feed those into AlphaFold to check whether they’d fold up that way!
I’d heard of AlphaFold before, but not of ProteinMPNN. You feed a protein structure into PMPNN & generate possible sequences of amino acids. And then you feed those into AlphaFold to check whether they’d fold up that way!
New episode of HARD DRUGS!
AlphaFold, ProteinMPNN & other AI tools are transforming biology and drug design.
But how do they work? What can’t they do? And can we use them to make a vaccine against Strep A for the very first time?
In this episode, Jacob and I talk about hacking proteins with AI.
AlphaFold, ProteinMPNN & other AI tools are transforming biology and drug design.
But how do they work? What can’t they do? And can we use them to make a vaccine against Strep A for the very first time?
In this episode, Jacob and I talk about hacking proteins with AI.
Hacking proteins with AI
open.spotify.com
October 2, 2025 at 6:28 AM
Cool episode on how AI can help us design new proteins (e.g., for vaccines).
I’d heard of AlphaFold before, but not of ProteinMPNN. You feed a protein structure into PMPNN & generate possible sequences of amino acids. And then you feed those into AlphaFold to check whether they’d fold up that way!
I’d heard of AlphaFold before, but not of ProteinMPNN. You feed a protein structure into PMPNN & generate possible sequences of amino acids. And then you feed those into AlphaFold to check whether they’d fold up that way!
New episode of HARD DRUGS!
AlphaFold, ProteinMPNN & other AI tools are transforming biology and drug design.
But how do they work? What can’t they do? And can we use them to make a vaccine against Strep A for the very first time?
In this episode, Jacob and I talk about hacking proteins with AI.
AlphaFold, ProteinMPNN & other AI tools are transforming biology and drug design.
But how do they work? What can’t they do? And can we use them to make a vaccine against Strep A for the very first time?
In this episode, Jacob and I talk about hacking proteins with AI.
Hacking proteins with AI
open.spotify.com
October 1, 2025 at 4:23 PM
New episode of HARD DRUGS!
AlphaFold, ProteinMPNN & other AI tools are transforming biology and drug design.
But how do they work? What can’t they do? And can we use them to make a vaccine against Strep A for the very first time?
In this episode, Jacob and I talk about hacking proteins with AI.
AlphaFold, ProteinMPNN & other AI tools are transforming biology and drug design.
But how do they work? What can’t they do? And can we use them to make a vaccine against Strep A for the very first time?
In this episode, Jacob and I talk about hacking proteins with AI.
RFD-VIM
If you work with RFdiffusion(1) and ProteinMPNN workflows, you might find my PyMOL extension useful for selecting residues in 3D to freeze in designs. I prefer adjusting settings interactively on the structure instead of typing contigs. It can also generate structures for RMSD calculations.
If you work with RFdiffusion(1) and ProteinMPNN workflows, you might find my PyMOL extension useful for selecting residues in 3D to freeze in designs. I prefer adjusting settings interactively on the structure instead of typing contigs. It can also generate structures for RMSD calculations.
September 29, 2025 at 4:57 PM
RFD-VIM
If you work with RFdiffusion(1) and ProteinMPNN workflows, you might find my PyMOL extension useful for selecting residues in 3D to freeze in designs. I prefer adjusting settings interactively on the structure instead of typing contigs. It can also generate structures for RMSD calculations.
If you work with RFdiffusion(1) and ProteinMPNN workflows, you might find my PyMOL extension useful for selecting residues in 3D to freeze in designs. I prefer adjusting settings interactively on the structure instead of typing contigs. It can also generate structures for RMSD calculations.
I am thrilled to release ProteinDJ: a high-performance and modular protein design pipeline. Our open-source workflow incorporates #RFdiffusion, #ProteinMPNN, #FAMPNN, #AlphaFold2 and #Boltz-2. It is a fast, free, and fun way to design proteins (1/5)
doi.org/10.1101/2025.09.24.678028 #proteindesign
doi.org/10.1101/2025.09.24.678028 #proteindesign
September 28, 2025 at 9:16 PM
I am thrilled to release ProteinDJ: a high-performance and modular protein design pipeline. Our open-source workflow incorporates #RFdiffusion, #ProteinMPNN, #FAMPNN, #AlphaFold2 and #Boltz-2. It is a fast, free, and fun way to design proteins (1/5)
doi.org/10.1101/2025.09.24.678028 #proteindesign
doi.org/10.1101/2025.09.24.678028 #proteindesign
cool new paper on designing proteins against immunogenicity by incorporating human k-mers and hidden k-mers (not presented on MHC-1) into proteinMPNN designs. i've been thinking about this for a while - cool to see an implementation!
pmc.ncbi.nlm.nih.gov/articles/PMC...
pmc.ncbi.nlm.nih.gov/articles/PMC...
A novel decoding strategy for ProteinMPNN to design with less visibility to cytotoxic T-lymphocytes
Due to their versatility and diverse production methods, proteins have attracted a lot of interest for industrial as well as therapeutic applications. Designing new therapeutics requires careful consi...
pmc.ncbi.nlm.nih.gov
September 28, 2025 at 7:33 PM
cool new paper on designing proteins against immunogenicity by incorporating human k-mers and hidden k-mers (not presented on MHC-1) into proteinMPNN designs. i've been thinking about this for a while - cool to see an implementation!
pmc.ncbi.nlm.nih.gov/articles/PMC...
pmc.ncbi.nlm.nih.gov/articles/PMC...
7/13 Then it uses ProteinMPNN to optimize for solubility, increasing the chances of experimental success. Finally, uses AF2 to predict the structure. To demonstrate Bindcraft’s utility, the authors carried out many wet-lab experiments, something not as common as I would like.
August 27, 2025 at 7:54 PM
7/13 Then it uses ProteinMPNN to optimize for solubility, increasing the chances of experimental success. Finally, uses AF2 to predict the structure. To demonstrate Bindcraft’s utility, the authors carried out many wet-lab experiments, something not as common as I would like.
I wrote a blogpost ( blog.matteoferla.com/2025/08/conf... ) about a potential aid to crystallography thought up when I worked w/ @xchem.bsky.social: generating a fake crystal lattice from a protein model in order to design optimal crystal contacts with ProteinMPNN. #ProteinDesign #StructuralBiology
Confabulating a crystal lattice
Matteo Ferla, a blog containing various tidbits of useful information, how-tos and what-ifs from biochemistry to Star Trek.
blog.matteoferla.com
August 17, 2025 at 6:02 PM
I wrote a blogpost ( blog.matteoferla.com/2025/08/conf... ) about a potential aid to crystallography thought up when I worked w/ @xchem.bsky.social: generating a fake crystal lattice from a protein model in order to design optimal crystal contacts with ProteinMPNN. #ProteinDesign #StructuralBiology
In the past, every BioML model had its own data pipeline, creating loads of overhead. With AtomWorks, >80% of code is shared between models like ProteinMPNN, RF3 or design models. 2/6
August 15, 2025 at 5:45 PM
In the past, every BioML model had its own data pipeline, creating loads of overhead. With AtomWorks, >80% of code is shared between models like ProteinMPNN, RF3 or design models. 2/6
(3/7)
We're also thrilled to release AtomWorks, which we used as the foundation for not only RF3, but also RF2AA, LigandMPNN, ProteinMPNN, and a design model — all by just swapping out a handful of modular components we call Transforms (just like Torchdata, for those familiar)
We're also thrilled to release AtomWorks, which we used as the foundation for not only RF3, but also RF2AA, LigandMPNN, ProteinMPNN, and a design model — all by just swapping out a handful of modular components we call Transforms (just like Torchdata, for those familiar)
August 15, 2025 at 5:17 PM
(3/7)
We're also thrilled to release AtomWorks, which we used as the foundation for not only RF3, but also RF2AA, LigandMPNN, ProteinMPNN, and a design model — all by just swapping out a handful of modular components we call Transforms (just like Torchdata, for those familiar)
We're also thrilled to release AtomWorks, which we used as the foundation for not only RF3, but also RF2AA, LigandMPNN, ProteinMPNN, and a design model — all by just swapping out a handful of modular components we call Transforms (just like Torchdata, for those familiar)
The @uwproteindesign.bsky.social's experimental pipeline behind models like RFdiffusion and ProteinMPNN:
- A rapid, scalable, pipeline for producing and characterizing proteins
- A demultiplexing protocol for converting oligopools to clonal constructs
Jason Qian @lfmilles.bsky.social Basile Wicky
- A rapid, scalable, pipeline for producing and characterizing proteins
- A demultiplexing protocol for converting oligopools to clonal constructs
Jason Qian @lfmilles.bsky.social Basile Wicky
August 12, 2025 at 2:34 PM
The @uwproteindesign.bsky.social's experimental pipeline behind models like RFdiffusion and ProteinMPNN:
- A rapid, scalable, pipeline for producing and characterizing proteins
- A demultiplexing protocol for converting oligopools to clonal constructs
Jason Qian @lfmilles.bsky.social Basile Wicky
- A rapid, scalable, pipeline for producing and characterizing proteins
- A demultiplexing protocol for converting oligopools to clonal constructs
Jason Qian @lfmilles.bsky.social Basile Wicky
Householder et al. used RFdiffusion and ProteinMPNN to de novo design an α-helical TCR mimic (TCRm) specific for an NY-ESO-1 peptide on HLA:A*02. bit.ly/4myrpa4
August 12, 2025 at 11:35 AM
Householder et al. used RFdiffusion and ProteinMPNN to de novo design an α-helical TCR mimic (TCRm) specific for an NY-ESO-1 peptide on HLA:A*02. bit.ly/4myrpa4
Researchers leveraged AI tools RFdiffusion and ProteinMPNN to develop a de novo protein/mini-TCR mimic targeting NY-ESO-1 (cancer testis antigen). Exciting times for drug discovery.
De novo design and structure of a peptide–centric TCR mimic binding module | Science www.science.org/doi/10.1126/...
De novo design and structure of a peptide–centric TCR mimic binding module | Science www.science.org/doi/10.1126/...
De novo design and structure of a peptide–centric TCR mimic binding module
T cell receptor (TCR) mimics offer a promising platform for tumor-specific targeting of peptide–major histocompatibility complex (pMHC) in cancer immunotherapy. In this study, we designed a de novo α-...
www.science.org
August 4, 2025 at 7:41 PM
Researchers leveraged AI tools RFdiffusion and ProteinMPNN to develop a de novo protein/mini-TCR mimic targeting NY-ESO-1 (cancer testis antigen). Exciting times for drug discovery.
De novo design and structure of a peptide–centric TCR mimic binding module | Science www.science.org/doi/10.1126/...
De novo design and structure of a peptide–centric TCR mimic binding module | Science www.science.org/doi/10.1126/...
new paper in nature where researchers have synthesised snake anti-venoms using roseTTA fold to design proteins, alphafold 2 to confirm, proteinMPNN to figure out the amino-acid sequence to get that protein, then built the custom dna sequence that produces those aminoacids, inserted it into bacteria>
August 1, 2025 at 11:50 PM
new paper in nature where researchers have synthesised snake anti-venoms using roseTTA fold to design proteins, alphafold 2 to confirm, proteinMPNN to figure out the amino-acid sequence to get that protein, then built the custom dna sequence that produces those aminoacids, inserted it into bacteria>
Evaluating zero-shot prediction of protein design success by AlphaFold, ESMFold, and ProteinMPNN [new]
Zero-shot protein design assessment filters, but accuracy predicting success is limited.
Zero-shot protein design assessment filters, but accuracy predicting success is limited.
August 1, 2025 at 8:58 PM
Evaluating zero-shot prediction of protein design success by AlphaFold, ESMFold, and ProteinMPNN [new]
Zero-shot protein design assessment filters, but accuracy predicting success is limited.
Zero-shot protein design assessment filters, but accuracy predicting success is limited.
Evaluating zero-shot prediction of protein design success by AlphaFold, ESMFold, and ProteinMPNN https://www.biorxiv.org/content/10.1101/2025.07.29.667290v1
August 1, 2025 at 8:47 PM
Evaluating zero-shot prediction of protein design success by AlphaFold, ESMFold, and ProteinMPNN https://www.biorxiv.org/content/10.1101/2025.07.29.667290v1
Evaluating zero-shot prediction of protein design success by AlphaFold, ESMFold, and ProteinMPNN https://www.biorxiv.org/content/10.1101/2025.07.29.667290v1
August 1, 2025 at 8:47 PM
Evaluating zero-shot prediction of protein design success by AlphaFold, ESMFold, and ProteinMPNN https://www.biorxiv.org/content/10.1101/2025.07.29.667290v1
TIM barrels are among nature’s most powerful enzyme scaffolds but making them from scratch with catalytic function has been a challenge.
Enter CANVAS: a computational pipeline combining Triad, RFdiffusion & ProteinMPNN to customize minimal TIM barrels into functional enzymes.
Enter CANVAS: a computational pipeline combining Triad, RFdiffusion & ProteinMPNN to customize minimal TIM barrels into functional enzymes.
July 29, 2025 at 6:33 PM
TIM barrels are among nature’s most powerful enzyme scaffolds but making them from scratch with catalytic function has been a challenge.
Enter CANVAS: a computational pipeline combining Triad, RFdiffusion & ProteinMPNN to customize minimal TIM barrels into functional enzymes.
Enter CANVAS: a computational pipeline combining Triad, RFdiffusion & ProteinMPNN to customize minimal TIM barrels into functional enzymes.
Fully functional AAV viral vectors with highly altered structural cores and subunit interfaces using ProteinMPNN [new]
AAV2 core redesign: PMPNN, capsid fxn, assembly/stable, transduction.
AAV2 core redesign: PMPNN, capsid fxn, assembly/stable, transduction.
July 27, 2025 at 6:02 AM
Fully functional AAV viral vectors with highly altered structural cores and subunit interfaces using ProteinMPNN [new]
AAV2 core redesign: PMPNN, capsid fxn, assembly/stable, transduction.
AAV2 core redesign: PMPNN, capsid fxn, assembly/stable, transduction.
Fully functional AAV viral vectors with highly altered structural cores and subunit interfaces using ProteinMPNN https://www.biorxiv.org/content/10.1101/2025.07.24.666527v1
July 27, 2025 at 5:46 AM
Fully functional AAV viral vectors with highly altered structural cores and subunit interfaces using ProteinMPNN https://www.biorxiv.org/content/10.1101/2025.07.24.666527v1
Fully functional AAV viral vectors with highly altered structural cores and subunit interfaces using ProteinMPNN https://www.biorxiv.org/content/10.1101/2025.07.24.666527v1
July 27, 2025 at 5:46 AM
Fully functional AAV viral vectors with highly altered structural cores and subunit interfaces using ProteinMPNN https://www.biorxiv.org/content/10.1101/2025.07.24.666527v1