The new class of protein AI design tools are amazing, and could revolutionize many areas of science, including therapeutics, diagnostics, and biosensors. Surprisingly, one important area that I haven't seen discussed too much is how these tools could impact patents. I am not a lawyer, so obviously this post is just my basic understanding, and I'd be happy to hear corrections. If there is a more expert critique I did not find it.

Patents are wordy and convoluted by design. For proteins, because a string of amino acids defines them, there are some common elements: they often include the sequence(s) being patented, and a threshold for how similar another sequence can be before infringing. That means there is a target to hit, and AI is really good at hitting targets.

There are two major categories of protein patents: biologics (usually meaning antibodies) and enzymes.

Antibodies

According to the European Patent Office, there are two main ways to patent an antibody:

• "functional" claims, usually meaning the antibody's associated antigen or epitope;
• "structural" claims, usually meaning a sequence and sequence identity threshold, along with the epitope or some other support.

Over the past few years, the "functional" claim has been going away. In the US it was killed off by the 2023 Amgen vs Sanofi ruling, which essentially said you can't patent the concept of an antibody against PCSK9. That means antibodies are now almost exclusively patented based on their structure (more specifically, a sequence plus some supporting functional information like epitope affinity.)

For antibody sequences, it used to be common for claims to cover any sequence 80%+ identical in the heavy or light chains. These days it seems like you have to be more specific, with claims only covering 100% identity to all 6 CDRs.

To take some real examples:

• Zanidatamab, a HER2 bispecific approved in 2024, claims sequences with 100% sequence identity to its CDRs;
• Epcoritamab, a CD3/CD20 bispecific approved in 2024, also has claims sequences with 100% sequence identity to its CDRs;
• Trastuzumab, the famous HER2 antibody approved in 1998 (filed in 2013), claims sequences with 85%+ sequence identity to the heavy and light chains, and does not mention CDRs at all.

The EPO says: "the slightest modification of the CDRs can affect the recognition of the target." There is a nice breakdown of the differences between the USPTO vs EPO approach to antibody patents here.

Enzymes

For enzymes, the patent landscape is more complicated, or at least more varied. Unlike antibodies, where the patents are pretty uniformly focused on the sequence that binds an epitope, enzymes can perform any number of functions. Enzyme types include enzyme replacement therapies, industrial enzymes like detergents, and molecular biology tools like CRISPR-Cas9. It is still typical for these patents to include a sequence and supporting information.

Some examples:

• this detergent patent, granted in 2018, claims sequences with 60%+ sequence identity to the reference;
• this proteinase patent, granted in 2022, claims sequences with 90%+ sequence identity to the reference;
• this novel Taq polymerase patent, granted in 2025, claims sequences with 95%+ sequence identity to the reference.

Cas9

The Cas9 patents are unusually diverse: there are hundreds of them and they mostly cover the many applications of the invention rather than the sequences. Since the 2013 ruling against Myriad Genetics, sequences from naturally occurring enzymes like Cas9 cannot be patented. Engineered sequences can be patented with other supporting functional information. You cannot take one of the thousands of unique Cas9 sequences in GenBank and use that to circumvent the CRISPR-Cas9 patents.

There are hundreds of Cas9 patents covering everything anyone could think of

AI

Given that the amino acid sequence is so important in protein patents, I am surprised that it is not bigger news that AI has effectively broken the direct connection between sequence and function.

For patents where protein sequence identity is protected, it is now relatively straightforward to generate new sequences that fold to the same structure but have 50% or lower sequence identity.

For antibody patents where the CDR sequence is protected, I believe it is also relatively straightforward to introduce a mutation that does not disrupt binding. To be honest, I am not even sure AI is required here, since a mutation scan could perform the same function. Perhaps for this reason, a recent paper called for "comprehensive CDR scanning" to protect a panel of CDR sequences instead of just one.

ProteinMPNN, published in 2022 by Baker lab, is the most prominent tool for producing a new sequence that folds to a known structure. ProteinMPNN is widely used as a step in many protein design workflows. For example, methods like RFdiffusion generate backbone coordinates only, and ProteinMPNN turns that into an amino acid sequence.

In a follow-up ProteinMPNN paper, the authors demonstrated that they could make a myoglobin and TEV protease with comparable or better function and greater stability than the natural versions, with sequence identities as low as 40%. This is below the sequence identity threshold in any patent I have seen.

ProteinMPNN can be used to produce a new sequence for a protein while maintaining its function

Sequence vs Structure

If this ability for AI to circumvent sequence-based patents is an issue, maybe the obvious change here would be to base patent protection on structure. This is a bit more complex than sequence identity, but one way to do this would be with TM-align or a similar tool. TM-align has >3k citations so it is arguably the standard in the field. A TM-score of above 0.8 indicates "the same topology"—in other words a very close structure. I think this would work well for many proteins, though it might need to be constrained to subdomains (akin to CDRs) in some cases.

Interestingly, the only literature I found on patenting 3D structure is from 20 years ago. Maybe this has been debated already and rejected for some reason. I suspect it was just easier to use sequence though.

OpenCRISPR-1

OpenCRISPR-1 was published in 2024 by the protein AI company Profluent. This is a de novo Cas9 enzyme that is substantially different in sequence to any known Cas9 (according to the abstract, "400 mutations away in sequence [from SpCas9]"—specifically 403/1380, or 71% identity).

Cas9 is a bilobed enzyme, with a REC lobe (nucleotide recognition) and a NUC lobe (DNA cleavage and PAM recognition.) Broadly speaking, the REC lobe is the first half of the enzyme (amino acids 50–700), and the NUC lobe is the second (1–50 and 700–1350.) These two lobes are connected by a "bridge helix".

Cartoon representation of Cas9 from addgene.

The OpenCRISPR-1 enzyme is not as novel as it might seem. In fact, I found it is actually 98% identical to a sequence constructed from three Cas9's spliced together from Streptococcus cristatus, Streptococcus pyogenes and Streptococcus sanguinis (24 amino acids are unique to OpenCRISPR-1).

This raises an interesting question, which is whether you could create a "novel" Cas9 by simply stitching together the REC lobe from one species' Cas9 and the NUC lobe from another. I believe this enzyme would work, and this sequence would meet any sequence identity threshold requirements.

The Profluent paper says the OpenCRISPR-1 enzyme was released for "research and commercial applications", but there is a big caveat here. Since CRISPR-Cas9 patents post-date the Myriad decision, almost all are functional / method of use, and naturally the most protected part is the use of Cas9 in "commercial applications" like therapeutics and diagnostics.

It is commendable that Profluent tried to broaden the availability of Cas9, so I appreciate the work behind this, but as I understand it, OpenCRISPR-1 is not really more available for commercial use than any Cas9.

There is actually another "royalty-free" Cas, a "Class 2 Type V" Cas nuclease called MAD7, released by Inscripta for commercial use in 2023. I do not know how this enzyme intersects with the many Cas9 patents.

Conclusion

One upshot of all this AI work is that me-too and biosimilar antibodies will be easier to make. That saves some time and money, but does not necessarily save on the major clinical trial costs, although the probability of success could go up a lot if the antibody is functionally identical.

While many enzyme patents will be affected, patents like CRISPR-Cas9 that rely on functional or method of use claims do not seem to be impacted as much. I don't know how many enzyme patents rely on sequence identity claims vs other claims these days. It would be interesting to (get an AI to) do a proper survey.

For internal research use, it's unclear to me that using AI to reproduce a patented protein does a whole lot, since at least in drug development, the research exemption seems to allow for the use of patented material quite broadly.

I have written about protein binder design a few times now (the Adaptyv competition; a follow up). Corin Wagen recently wrote a great piece about protein–ligand binding. This purpose of this post is to review how well protein binder design is working today, and point out some interesting differences in model performance that I do not understand.

Protein design

There are two major types of protein design:

1. Design a sequence to perform some task: e.g., produce a sequence that improves upon some property of the protein
2. Design a structure to perform some task: e.g., produce a protein structure that binds another protein

There is spillover between these two classes but I think it's useful to split this way.

Sequence models

Sequence models include open-source models like the original ESM2, ProSST, SaProt, and semi-open or fully proprietary models from EvolutionaryScale (ESM3), OpenProtein (PoET-2), and Cradle Bio. The ProteinGym benchmark puts ProSST, PoET-2 and SaProt up near the top.

Many of the recent sequence-based models now also include structure information, represented as a parallel sequence, with one "structure token" per amino acid. This addition seems to improve performance quite a lot, allows sequence models to make use of the PDB, and — analogously to Vision Transformers — blurs the line between sequence and structure models.

SaProt uses a FoldSeek-derived alphabet to encode structural information

The most basic use-case for sequence models is probably improving the stability of a protein. You can take a protein sequence, make whatever edits your model deems high likelihood, and this should produce a sequence that retains the same fold, but is more "canonical", and so may have improved stability too.

An elaboration of this experiment is to find some data, e.g., thermostability for a few thousand proteins, and fine-tune the original language model to be able to predict that property. SaProtHub makes this essentially push-button.

A further elaboration is doing active learning, where you propose edits using your model, generate empirical data for these edits (e.g., binding affinity), and go back and forth, hopefully improving performance each iteration. For example, EVOLVEpro, Nabla Bio's JAM (which also uses structure), and Prescient's Lab-in-the-loop. These systems can be complex, but can also be as simple as running regressions on the output of the sequence models.

EvolvePro's learning loop

Sequence-based models are a natural fit to these kinds of problems, since you can easily edit the sequence but maintain the same fold and function. Profluent and other companies make use of this ability by producing patent-unencumbered sequences like OpenCRISPR.

This is especially enabling for the biosimilars industry. Many biologics patents protect the sequence by setting amino acid identity thresholds. For example, in the Herceptin/trastuzumab patent they protect any sequence >=85% identical to the heavy (SEQ ID NO: I) or light chain (SEQ ID NO: II).

Excerpt from the main trastuzumab patent

Patent attorneys will layer as many other protections on top of this as they can think of, but the sequence of the antibody is the primary IP. (Tangentially, it is insane how patents always give examples of numbers greater than X. Hopefully, the AIs that will soon be writing patents won't do this.)

For binder design, sequence models appear to have limits. Naively, since you do not know the positions of the atoms, then unless you are apeing known interaction motifs, you would assume binder design should be difficult?

Diego del Alamo points out apparent limits in the performance of sequence models for antibody design

Structural models

Structural models include the original RFdiffusion and the recently released antibody variant RFantibody from the Baker lab, RSO from the ColabDesign team, BindCraft, EvoBind2, foldingdiff from Microsoft, and models from startups like Generate Biomedicines (Chroma), Chai Discovery, and Diffuse Bio. (Some of these tools are available on my biomodals repo).

Structural models are trained on both sequence data (e.g., UniRef) and structure data (PDB), but they deal in atom co-ordinates instead of amino acid strings. That difference means diffusion-style models dominate here over the discrete-token–focused transformers.

There are two major classes of structural models:

• Diffusion models like RFdiffusion and RFantibody
• AlphaFold2-based models like BindCraft, RSO, and EvoBind2

The success rates of RFdiffusion and RFantibody are not great. For some targets they achieve a >1% success rate (if we define success as finding a <1µM binder), but in other cases they nominate thousands of designs and find no strong binder.

An example from the RFantibody paper showing a low success rate

BindCraft and RSO are two similar methods that produce minibinders (small-ish non-antibody–based proteins) and rely on inverting AlphaFold2 to turn structure into sequence. EvoBind2 produces cyclic or linear peptides, and also relies heavily on an AlphaFold confidence metric (pLDDT) as part of its loss.

BindCraft (top) and EvoBind2 (bottom) have similar loss functions that rely on AF2's pLDDT and intermolecular contacts

Even though these AF2-based models work very well, one non-obvious catch is that you cannot take a binding pose and get AlphaFold2 to evaluate it. These models can generate binders, but not discriminate binders from non-binders. In the EvoBind2 paper, they found that "No in silico metric separates true from false binders", which means the problem is a bit more complex than just "ask AF2 if it looks good".

According to the AF2Rank paper, the AF2 model has a good model of the physics of protein folding, but may not find the global minimum. The MSAs' job is to help focus that search. This was surprising to me! The protein folding/binding problem is more of a search problem than I realized, which means more compute should straightforwardly improve performance by simply doing more searching. This is also evidenced by the AlphaFold 3 paper, where re-folding antibodies 1000 times led to improved prediction quality.

Excerpt from the AF2Rank paper (top), and a tweet from Sergey Ovchinnikov (bottom) explaining the primacy of sequence data in structure prediction

RFdiffusion/RFantibody vs BindCraft/EvoBind2

The main comparison I wanted to make in this post is between RFdiffusion/RFantibody vs BindCraft and EvoBind2.

These are all recently released, state-of-the-art models from top labs. However, the difference in claimed performance is pretty striking.

While the RFdiffusion and RFantibody papers caution that you may need to test hundreds or even thousands of proteins to find one good binder, the BindCraft and EvoBind2 papers appear to show very high success rates, perhaps even as high as 50%. (EvoBind2 only shows results for one ribonuclease target but BindCraft includes multiple).

Words of caution from the RFantibody github repo (top) and BindCraft's impressive results for 10 targets (bottom)

There is no true benchmark to reference here, but I think under reasonable assumptions, BindCraft (and arguably EvoBind2) achieve a >10X greater success rate than RFdiffusion or RFantibody. The Baker lab is the leading and best resourced lab in this domain, so what accounts for this large difference in performance? I can think of a few possibilities:

• RoseTTAFold2 was not the best filter for RFantibody to use, and switching to AlphaFold3 would improve performance. This is plausible, but it is hard to believe that is a 10X improvement.
• Antibodies are just harder than minibinders or cyclic peptides. Hypervariable regions are known to be difficult to fold, since they do not have the advantage of evolutionary conservation. However, RFdiffusion also produces minibinders, so this is not a satisfactory explanation.
• BindCraft and EvoBind2 are testing on easier targets. There is likely some truth to this. Most (but not all) examples in the BindCraft paper are for proteins with known binders; EvoBind2 is only tested against a target with a known peptide binder. However, most of RFantibody's targets also have known antibodies in PDB.
• Diffusion currently just does not work as well as AlphaFold-based methods. AlphaFold2 (and its descendants, AF3, Boltz, Chai-1, etc.) have learned enough physics to recognize binding, and by leaning on this ability heavily, and filtering carefully, you get much better performance.

What comes next?

RFdiffusion and RFantibody are arguably the first examples of successful de novo binder design and antibody design, and for that reason are important papers. BindCraft and EvoBind2 have proven they can produce one-shot nanomolar binders under certain circumstances, which is technically extremely impressive.

However, if we could get another 10X improvement in performance, then I think these tools are being used in every biotech and research lab. Some ideas for future directions:

• More compute: One of the interesting things about BindCraft and EvoBind2 is how long they take to produce anything. In BindCraft's case, it generates a lot of candidates, but has a long list of criteria that must be met. One BindCraft run will screen hundreds or thousands of candidates and can easily cost $10+. Similary, EvoBind2 can run for 5+ hours before producing anything, again easily costing $10+. This approach of throwing compute at the problem may be generally applicable, and may be analogous to the recently successful LLM reasoning approaches.
• Combine diffusion and AlphaFold-based methods: I have no specific idea here, but since they are quite different approaches, maybe integrating some ideas from RFdiffusion into EvoBind2 or BindCraft could help.
• Combine sequence models and structure models: There is already a lot of work happening here, both from the sequence side and structure side. In the simplest case, the output of a sequence model like ESM2 could be an independent contributor to the loss of a structure model. At the very least, this could help filter out structures that do not fold.
• Neural Network Potentials: Neural Network Potentials are an exciting new tool for molecular dynamics (see Duignan, 2024 or Barnett, 2024). Achira just got funded to work on this, and has several of the pioneers of the field on board. Semi-open source models like orb-v2 from Orbital Materials are being actively developed too. The amount of compute required is prohibitive right now, but even a short trajectory could plausibly help with rank ordering binders, and would be independent of the AF2 metrics.

Tweet from Tim Duignan at Orbital Materials

Adaptyv is a newish startup that sells high-throughput protein assays. The major innovations are (a) they tell you the price (a big innovation for biotech services!) (b) you only have to upload protein sequences, and you get results in a couple of weeks.

A typical Adaptyv workflow might look like the following:

• Design N protein binders for a target of interest (Adaptyv has 50-100 pre-specified targets)
• Submit your binder sequences to Adaptyv
• Adaptyv synthesizes DNA, then protein, using your sequences
• In ~3 weeks you get affinity measurements for each design at a cost of $149 per data-point

This is an exciting development since it decouples "design" and "evaluation" in a way that enables computation-only startups to get one more step towards a drug (or sensor, or tool). There are plenty of steps after this one, but it's still great progress!

The Adaptyv binder design competition

A couple of months ago, Adaptyv launched a binder design competition, where the goal was to design an EGFR binder. There was quite a lot of excitement about the competition on Twitter, and about 100 people ended up entering. At around the same time, Leash Bio launched a small molecule competition on Kaggle, so there was something in the air.

PAE and iPAE

For this competition, Adaptyv ranked designs based on the "PAE interaction" (iPAE) of the binder with EGFR.

PAE (Predicted Aligned Error) "indicates the expected positional error at residue x if the predicted and actual structures are aligned on residue y". iPAE is the average PAE for residues in the binder vs target. In other words, how accurate do we expect the relative positioning of binder and target to be? This is a metric that David Baker's lab seems to use quite a bit, at least for thresholding binders worth screening. It is straightforward to calculate using the PAE outputs from AlphaFold.

Unusually, compared to, say, a Kaggle competition, in this competition there are no held-out data that your model is evaluated on. Instead, if you can calculate iPAE, you know your expected position on the leaderboard before submitting.

The original paper Adaptyv reference is Improving de novo protein binder design with deep learning and the associated github repo has an implementation of iPAE that I use (and I assume the code Adaptyv use.)

Confusingly, there is also a metric called "iPAE" mentioned in the paper Systematic discovery of protein interaction interfaces using AlphaFold and experimental validation. It is different, but could actually be a more appropriate metric for binders?

At the end of last month (August 2024), there was a new Baker lab paper on Ras binders that also used iPAE, in combination with a few other metrics like pLDDT.

Experiments

A week or so after the competition ended, I found some time to try a few experiments.

Throughout these experiments, I include modal commands to run the relevant software. If you clone the biomodals repo it should just work(?)

iPAE vs Kd

The consensus seems to be that <10 represents a decent iPAE, but in order for iPAE to be useful, it should correlate with some physical measurement. As a small experiment, I took 55 PDB entries from PDBbind (out of ~100 binders that were <100 aas long, had an associated Kd, and only two chains), ran AlphaFold, calculated iPAE, and correlated this to the known Kd. I don't know that I really expected iPAE to correlate strongly with Kd, but it's pretty weak.

PDBbind Kd vs iPAE correlation

# download the PDBbind protein-protein dataset in a more convenient format and run AlphaFold on one example
wget https://gist.githubusercontent.com/hgbrian/413dbb33bd98d75cc5ee6054a9561c54/raw -O pdbbind_pp.tsv
tail -1 pdbbind_pp.tsv
wget https://www.rcsb.org/fasta/entry/6har/display -O 6har.fasta
echo ">6HAR\nYVDYKDDDDKEFEVCSEQAETGPCRACFSRWYFDVTEGKCAPFCYGGCGGNRNNFDTEEYCMAVCGSAIPRHHHHHHAAA:IVGGYTCEENSLPYQVSLNSGSHFCGGSLISEQWVVSAAHCYKTRIQVRLGEHNIKVLEGNEQFINAAKIIRHPKYNRDTLDNDIMLIKLSSPAVINARVSTISLPTAPPAAGTECLISGWGNTLSFGADYPDELKCLDAPVLTQAECKASYPGKITNSMFCVGFLEGGKDSCQRDAGGPVVCNGQLQGVVSWGHGCAWKNRPGVYTKVYNYVDWIKDTIAANS" > 6har_m.fasta
modal run modal_alphafold.py --input-fasta 6har_m.fasta --binder-len 80

Greedy search

This is about the simplest approach possible.

• Start with EGF (53 amino acids)
• Mask every amino acid, and have ESM propose the most likely amino acid
• Fold and calculate iPAE for the top 30 options
• Take the best scoring iPAE and iterate

Each round takes around 5-10 minutes and costs around $4 on an A10G on modal.

# predict one masked position in EGF using esm2
echo ">EGF\nNSDSECPLSHDGYCL<mask>DGVCMYIEALDKYACNCVVGYIGERCQYRDLKWWELR" > esm_masked.fasta
modal run modal_esm2_predict_masked.py --input-fasta esm_masked.fasta

# run AlphaFold on the EGF/EGFR complex and calculate iPAE
echo ">EGF\nNSDSECPLSHDGYCLHDGVCMYIEALDKYACNCVVGYIGERCQYRDLKWWELR:LEEKKVCQGTSNKLTQLGTFEDHFLSLQRMFNNCEVVLGNLEITYVQRNYDLSFLKTIQEVAGYVLIALNTVERIPLENLQIIRGNMYYENSYALAVLSNYDANKTGLKELPMRNLQEILHGAVRFSNNPALCNVESIQWRDIVSSDFLSNMSMDFQNHLGSCQKCDPSCPNGSCWGAGEENCQKLTKIICAQQCSGRCRGKSPSDCCHNQCAAGCTGPRESDCLVCRKFRDEATCKDTCPPLMLYNPTTYQMDVNPEGKYSFGATCVKKCPRNYVVTDHGSCVRACGADSYEMEEDGVRKCKKCEGPCRKVCNGIGIGEFKDSLSINATNIKHFKNCTSISGDLHILPVAFRGDSFTHTPPLDPQELDILKTVKEITGFLLIQAWPENRTDLHAFENLEIIRGRTKQHGQFSLAVVSLNITSLGLRSLKEISDGDVIISGNKNLCYANTINWKKLFGTSGQKTKIISNRGENSCKATGQVCHALCSPEGCWGPEPRDCVSCRNVSRGRECVDKCNLLEGEPREFVENSECIQCHPECLPQAMNITCTGRGPDNCIQCAHYIDGPHCVKTCPAGVMGENNTLVWKYADAGHVCHLCHPNCTYGCTGPGLEGCPTNGPKIPSI" > egf_01.fasta
modal run modal_alphafold.py --input-fasta egf_01.fasta --binder-len 53

One of the stipulations of the competition is that your design must be at least 10 amino acids different to any known binder, so you must run the loop above 10 or more times. Of course, there is no guarantee that there is a single amino acid change that will improve the score, so you can easily get stuck.

After 12 iteratations (at a cost of around $50 in Alphafold compute), the best score I got was 7.89, which would have been good enough to make the top 5. (I can't be sure, but I think my iPAE calculation is identical!) Still, this is really just brute-forcing EGF tweaks. I think the score was asymptoting, but there were also jumps in iPAE with certain substitutions, so who knows?

Unfortunately, though the spirit of the competition was to find novel binders, the way iPAE works means that the best scores are very likely to come from EGF-like sequences (or other sequences in AlphaFold's training set).

Adaptyv are attempting to mitigate this issue by (a) testing the top 200 and (b) taking the design process into account. It is a bit of an impossible situation, since the true wet lab evaluation happens only after the ranking step.

Bayesian optimization

Given an expensive black box like AlphaFold + iPAE, some samples, and a desire to find better samples, one appropriate method is Bayesian optimization.

Basically, this method allows you, in a principled way, to control how much "exploration" of new space is appropriate (looking for global minima) vs "exploitation" of variations on the current best solutions (optimizing local minima).

Bayesian optimization of a 1D function

The input to a Bayesian optimization is of course not amino acids, but numbers, so I thought reusing the ESM embeddings would be a decent, or at least convenient, idea here.

I tried both the Bayesian Optimization package and a numpyro Thompson sampling implementation. I saw some decent results at first (i.e., the first suggestions seemed reasonable and scored well), but I got stuck either proposing the same sequences over and over, or proposing sequences so diverged that testing them would be a waste of time. The total search space is gigantic, so testing random sequences will not help. I think probably the ESM embeddings were not helping me here, since there were a lot of near-zeros in there.

This is an interesting approach, and not too difficult to get started with, but I think it would work better with much deeper sampling of a smaller number of amino acids, or perhaps a cruder, less expensive, evaluation function.

ProteinMPNN

ProteinMPNN (now part of the LigandMPNN package), maps structure to sequence (i.e., the inverse of AlphaFold). For example, you can input an EGF PDB file, and it will return a sequence that should produce the same fold.

I found that for this task ProteinMPNN generally produced sequences with low confidence (as reported by ProteinMPNN), and as you'd expect, these resulted in low iPAEs. Some folds are difficult for ProteinMPNN, and I think EGF falls into this category. To run ProteinMPNN, I would recommend Simon Duerr's huggingface space, since it has a friendly interface and includes an AlphaFold validation step.

ProteinMPNN running on huggingface

# download a EGF/EGFR crytal structure and try to infer a new sequence that folds to chain C (EGF)
wget https://files.rcsb.org/download/1IVO.pdb
modal run modal_ligandmpnn.py --input-pdb 1IVO.pdb --extract-chains AC --params-str '--seed 1 --checkpoint_protein_mpnn "/LigandMPNN/model_params/proteinmpnn_v_48_020.pt" --chains_to_design "C" --save_stats 1 --batch_size 5 --number_of_batches 100'

RFdiffusion

RFdiffusion was the first protein diffusion method that showed really compelling results in generating de novo binders. I would recommend ColabDesign as a convenient interface to this and other protein design tools.

The input to RFdiffusion can be a protein fold to copy, or a target protein to bind to, and the output is a PDB file with the correct backbone co-ordinates, but with every amino acid labeled as Glycine. To turn this output into a sequence, this PDB file must then be fed into ProteinMPNN or similar. Finally, that ProteinMPNN output is typically folded with AlphaFold to see if the fold matches.

Although RFdiffusion massively enriches for binders over random peptides, you still have to screen many samples to find the really strong binders. So, it's probably optimistic to think that a few RFdiffusion-derived binders will show strong binding, even if you can somehow get a decent iPAE.

In my brief tests with RFdiffusion here, I could not generate anything that looked reasonable. I think in practice, the process of using RFdiffusion successfully is quite a bit more elaborate and heuristic-driven than anything I was going to attempt.

Figure 1 from De novo design of Ras isoform selective binders, showing multiple methods for running RFdiffusion

# Run RFdiffusion on the EGF/EGFR crystal structure, and diffuse a 50-mer binder against chain A (EGFR)
modal run modal_rfdiffusion.py --contigs='A:50' --pdb="1IVO"

Other things

A few other strategies I thought might be interesting:

• Search FoldSeek for folds similar to EGF. The idea here is that you might find a protein in another organism that wants to bind EGFR. I do find some interesting human-parasitic nematode proteins in here, but decided these were unlikely to be EGFR binders.
• Search NCBI for EGF-like sequences with blastp. You can find mouse, rat, chimp, etc. but nothing too interesting. The iPAEs are lower than human EGF, as expected.
• Search the patent literature for EGFR binders. I did find some antibody-based binders, but as expected for folds that AlphaFold cannot solve, the iPAE was low.
• Delete regions of the protein with low iPAE contributions to increase the average score. I really thought this would work for at least one or two amino acids, but it did not seem to. I did not do this comprehensively, but perhaps there are no truly redundant parts of this small binder?

Conclusion

All the top spots on the leaderboard went to Alex Naka, who helpfully detailed his methods in this thread. (A lot of this is similar to what I did above, including using modal!) Anthony Gitter also published an interesting thread on his attempts. I find these kinds of threads are very useful since they give a sense of the tools people are using in practice, including some I had never heard of, like pepmlm and Protrek.

Finally, I made a tree of the 200 designs that Adaptyv is screening (with iPAE <10 in green, <20 in orange, and >20 in red). All the top scoring sequences are EGF-like and cluster together. (Thanks to Andrew White for pointing me at the sequence data). We can look forward to seeing the wet lab results published in a couple of weeks.

Tree of Adaptyv binder designs