DiffDock, a new diffusion-based ligand docking program, made s big splash earlier this year. But it’s apparently not all it’s claimed to be:
Computational Care
www.science.org
December 6, 2024 at 5:01 PM
DiffDock, a new diffusion-based ligand docking program, made s big splash earlier this year. But it’s apparently not all it’s claimed to be:
Looks like yet another case of overhyped results due to poor #bioMLeval evaluation of deep learning models -this time deep docking methods - specifically DiffDock. Look forward to the DiffDock authors response. But dont see any major flaws in this critique. Conclusion is REALLY worth reading!
Been a while since I read a paper like this:
• "What [DiffDock] appears to be doing cannot be considered" docking
• "Results are ... contaminated with near neighbors to test cases"
• "Results for DiffDock were artifactual"
• "Results for other methods were incorrectly done"
arxiv.org/abs/2412.02889
• "What [DiffDock] appears to be doing cannot be considered" docking
• "Results are ... contaminated with near neighbors to test cases"
• "Results for DiffDock were artifactual"
• "Results for other methods were incorrectly done"
arxiv.org/abs/2412.02889
Deep-Learning Based Docking Methods: Fair Comparisons to Conventional Docking Workflows
The diffusion learning method, DiffDock, for docking small-molecule ligands into protein binding sites was recently introduced. Results included comparisons to more conventional docking approaches, wi...
arxiv.org
December 5, 2024 at 11:30 PM
Looks like yet another case of overhyped results due to poor #bioMLeval evaluation of deep learning models -this time deep docking methods - specifically DiffDock. Look forward to the DiffDock authors response. But dont see any major flaws in this critique. Conclusion is REALLY worth reading!
Been a while since I read a paper like this:
• "What [DiffDock] appears to be doing cannot be considered" docking
• "Results are ... contaminated with near neighbors to test cases"
• "Results for DiffDock were artifactual"
• "Results for other methods were incorrectly done"
arxiv.org/abs/2412.02889
• "What [DiffDock] appears to be doing cannot be considered" docking
• "Results are ... contaminated with near neighbors to test cases"
• "Results for DiffDock were artifactual"
• "Results for other methods were incorrectly done"
arxiv.org/abs/2412.02889
Deep-Learning Based Docking Methods: Fair Comparisons to Conventional Docking Workflows
The diffusion learning method, DiffDock, for docking small-molecule ligands into protein binding sites was recently introduced. Results included comparisons to more conventional docking approaches, wi...
arxiv.org
December 5, 2024 at 3:36 PM
Been a while since I read a paper like this:
• "What [DiffDock] appears to be doing cannot be considered" docking
• "Results are ... contaminated with near neighbors to test cases"
• "Results for DiffDock were artifactual"
• "Results for other methods were incorrectly done"
arxiv.org/abs/2412.02889
• "What [DiffDock] appears to be doing cannot be considered" docking
• "Results are ... contaminated with near neighbors to test cases"
• "Results for DiffDock were artifactual"
• "Results for other methods were incorrectly done"
arxiv.org/abs/2412.02889
Response by the DiffDock authors on the recent preprint by Jain and colleagues.
You may have seen a recent pre-print [1] from Jain et al. with strongly worded claims against the experimental results in our DiffDock paper [2]. We initially declined to respond as we saw that this preprint contained falsehoods, misleading comparisons, seemingly deliberate omissions, ...1/n
December 8, 2024 at 9:58 PM
Response by the DiffDock authors on the recent preprint by Jain and colleagues.
A ***MUST READ*** for anyone working in computer-aided drug design
"Publication of studies such as the DiffDock report are not cost-free to the CADD field."
PLEASE DO YOUR HOMEWORK
"Publication of studies such as the DiffDock report are not cost-free to the CADD field."
PLEASE DO YOUR HOMEWORK
I'm thrilled to announce a new preprint describing collaborative work with Ajay Jain and Ann Cleves Jain, "Deep-Learning Based Docking Methods: Fair Comparisons to Conventional Docking Workflows".
arxiv.org/abs/2412.02889
arxiv.org/abs/2412.02889
Deep-Learning Based Docking Methods: Fair Comparisons to Conventional Docking Workflows
The diffusion learning method, DiffDock, for docking small-molecule ligands into protein binding sites was recently introduced. Results included comparisons to more conventional docking approaches, wi...
arxiv.org
December 6, 2024 at 5:42 PM
A ***MUST READ*** for anyone working in computer-aided drug design
"Publication of studies such as the DiffDock report are not cost-free to the CADD field."
PLEASE DO YOUR HOMEWORK
"Publication of studies such as the DiffDock report are not cost-free to the CADD field."
PLEASE DO YOUR HOMEWORK
For those following the @prof-ajay-jain.bsky.social /Cleaves/ @wpwalters.bsky.social preprint re: DiffDock @gcorso.bsky.social
I have read some of the back-&-forth between the author groups
As a practitioner in the field for > 20 yrs (academic side), here is my take: 🧵
I have read some of the back-&-forth between the author groups
As a practitioner in the field for > 20 yrs (academic side), here is my take: 🧵
@dereklowe.bsky.social honing in on the same bottom line message from @wpwalters.bsky.social @prof-ajay-jain.bsky.social
it's so true and hits so hard:
it's so true and hits so hard:
December 9, 2024 at 3:46 AM
For those following the @prof-ajay-jain.bsky.social /Cleaves/ @wpwalters.bsky.social preprint re: DiffDock @gcorso.bsky.social
I have read some of the back-&-forth between the author groups
As a practitioner in the field for > 20 yrs (academic side), here is my take: 🧵
I have read some of the back-&-forth between the author groups
As a practitioner in the field for > 20 yrs (academic side), here is my take: 🧵
While dockers are keep docking (now with diffusion and AI😀), Pat Walters and Ajay Jain show sobering assessment of ‘perceived’ accuracy of such methods! Diffdock shots fired 🔥 The conclusion is totally worth reading in full! #chemsky #compchemsky
arxiv.org/abs/2412.02889
arxiv.org/abs/2412.02889
Deep-Learning Based Docking Methods: Fair Comparisons to Conventional Docking Workflows
The diffusion learning method, DiffDock, for docking small-molecule ligands into protein binding sites was recently introduced. Results included comparisons to more conventional docking approaches, wi...
arxiv.org
December 6, 2024 at 5:33 AM
While dockers are keep docking (now with diffusion and AI😀), Pat Walters and Ajay Jain show sobering assessment of ‘perceived’ accuracy of such methods! Diffdock shots fired 🔥 The conclusion is totally worth reading in full! #chemsky #compchemsky
arxiv.org/abs/2412.02889
arxiv.org/abs/2412.02889
You may have seen a recent pre-print [1] from Jain et al. with strongly worded claims against the experimental results in our DiffDock paper [2]. We initially declined to respond as we saw that this preprint contained falsehoods, misleading comparisons, seemingly deliberate omissions, ...1/n
December 8, 2024 at 9:37 PM
You may have seen a recent pre-print [1] from Jain et al. with strongly worded claims against the experimental results in our DiffDock paper [2]. We initially declined to respond as we saw that this preprint contained falsehoods, misleading comparisons, seemingly deliberate omissions, ...1/n
I'm thrilled to announce a new preprint describing collaborative work with Ajay Jain and Ann Cleves Jain, "Deep-Learning Based Docking Methods: Fair Comparisons to Conventional Docking Workflows".
arxiv.org/abs/2412.02889
arxiv.org/abs/2412.02889
Deep-Learning Based Docking Methods: Fair Comparisons to Conventional Docking Workflows
The diffusion learning method, DiffDock, for docking small-molecule ligands into protein binding sites was recently introduced. Results included comparisons to more conventional docking approaches, wi...
arxiv.org
December 5, 2024 at 4:21 PM
I'm thrilled to announce a new preprint describing collaborative work with Ajay Jain and Ann Cleves Jain, "Deep-Learning Based Docking Methods: Fair Comparisons to Conventional Docking Workflows".
arxiv.org/abs/2412.02889
arxiv.org/abs/2412.02889
New preprint from my colleagues at Schrödinger on blending ML methods with physics-based refinement and scoring to improve small molecule docking performance - and generalization - on several tasks for early stages of drug discovery. Link👇
www.biorxiv.org/content/10.1...
#chemsky
www.biorxiv.org/content/10.1...
#chemsky
June 6, 2025 at 4:45 PM
New preprint from my colleagues at Schrödinger on blending ML methods with physics-based refinement and scoring to improve small molecule docking performance - and generalization - on several tasks for early stages of drug discovery. Link👇
www.biorxiv.org/content/10.1...
#chemsky
www.biorxiv.org/content/10.1...
#chemsky
I don't know. All the DiffDock results we report are from the original DiffDock paper.
December 6, 2024 at 11:52 PM
I don't know. All the DiffDock results we report are from the original DiffDock paper.
@prof-ajay-jain.bsky.social's response is here. www.linkedin.com/pulse/more-r...
More response to the response (DiffDock/DiffDock-L)
To amplify on what Pat Walters said (quotes from Gabriel Corso's post in italics): It only took a few days (in August 2023) for Ann Cleves, Pat, and me to figure out that the original DiffDock work ha...
www.linkedin.com
December 9, 2024 at 1:17 AM
@prof-ajay-jain.bsky.social's response is here. www.linkedin.com/pulse/more-r...
🧬 Testing beta-lactamase resistance with AlphaFold + DiffDock + GROMACS! Watch clavulanic acid bind TEM-5, CTX-M-15, KPC-2, and get rejected by TEM-30. Simulation confirms biology! 🔬💊 #gromacs #diffdock #chimerax #moleculardynamic #alphafold
Assessing TEM, CTX-M, and KPC2 With Molecular Docking and Molecular Dynamic Simulation | Everyday Is A School Day
🧬 Testing beta-lactamase resistance with AlphaFold + DiffDock + GROMACS! Watch clavulanic acid bind TEM-5, CTX-M-15, KPC-2, and get rejected by TEM-30. Simulation confirms biology! 🔬💊
www.kenkoonwong.com
February 22, 2026 at 3:49 AM
🧬 Testing beta-lactamase resistance with AlphaFold + DiffDock + GROMACS! Watch clavulanic acid bind TEM-5, CTX-M-15, KPC-2, and get rejected by TEM-30. Simulation confirms biology! 🔬💊 #gromacs #diffdock #chimerax #moleculardynamic #alphafold
You may have seen a recent pre-print [1] from Jain et al. with strongly worded claims against the experimental results in our DiffDock paper [2].
Here is our response, as also outlined in 's quoted Bluesky post.
Here is our response, as also outlined in 's quoted Bluesky post.
December 8, 2024 at 9:56 PM
You may have seen a recent pre-print [1] from Jain et al. with strongly worded claims against the experimental results in our DiffDock paper [2].
Here is our response, as also outlined in 's quoted Bluesky post.
Here is our response, as also outlined in 's quoted Bluesky post.
⚛️ cuEquivariance ⚛️ is something I've been waiting for forever. Thank you Mario Geiger and team!
Curious to see where the "equivariance is just too slow" debate goes 👀
github.com/nvidia/cuequ...
Curious to see where the "equivariance is just too slow" debate goes 👀
github.com/nvidia/cuequ...
GitHub - NVIDIA/cuEquivariance: cuEquivariance is a math library that is a collective of low-level primitives and tensor ops to accelerate widely-used models, like DiffDock, MACE, Allegro and NEQUIP, ...
cuEquivariance is a math library that is a collective of low-level primitives and tensor ops to accelerate widely-used models, like DiffDock, MACE, Allegro and NEQUIP, based on equivariant neural n...
github.com
November 21, 2024 at 9:47 PM
⚛️ cuEquivariance ⚛️ is something I've been waiting for forever. Thank you Mario Geiger and team!
Curious to see where the "equivariance is just too slow" debate goes 👀
github.com/nvidia/cuequ...
Curious to see where the "equivariance is just too slow" debate goes 👀
github.com/nvidia/cuequ...
Next @ MLSB: DiffDock-Pocket: Diffusion for Pocket-Level Docking with Sidechain Flexibility
Michael Plainer • Marcella Toth • Simon Dobers • @hannes-stark.bsky.social • @gcorso.bsky.social • Céline Marquet • Regina Barzilay
www.mlsb.io/papers_2023/...
Michael Plainer • Marcella Toth • Simon Dobers • @hannes-stark.bsky.social • @gcorso.bsky.social • Céline Marquet • Regina Barzilay
www.mlsb.io/papers_2023/...
December 15, 2023 at 8:22 PM
Next @ MLSB: DiffDock-Pocket: Diffusion for Pocket-Level Docking with Sidechain Flexibility
Michael Plainer • Marcella Toth • Simon Dobers • @hannes-stark.bsky.social • @gcorso.bsky.social • Céline Marquet • Regina Barzilay
www.mlsb.io/papers_2023/...
Michael Plainer • Marcella Toth • Simon Dobers • @hannes-stark.bsky.social • @gcorso.bsky.social • Céline Marquet • Regina Barzilay
www.mlsb.io/papers_2023/...
New Diffusion model based docking tool #MolSnapper #cheminformatics #RDKit
Recently I think relationship between ML and CADD become more closer compared to several years ago. DiffDock is one of the very nice work for interation of cheminformatics and SBDD. It uses diffusion model for…
Recently I think relationship between ML and CADD become more closer compared to several years ago. DiffDock is one of the very nice work for interation of cheminformatics and SBDD. It uses diffusion model for…
New Diffusion model based docking tool #MolSnapper #cheminformatics #RDKit
Recently I think relationship between ML and CADD become more closer compared to several years ago. DiffDock is one of the very nice work for interation of cheminformatics and SBDD. It uses diffusion model for protein-ligand pose prediction. But DiffDock can't run constrain docking which define specific location of protein pocket. Also pure docking tool AutoDock-vina also can't run constrained docking.
iwatobipen.wordpress.com
April 26, 2025 at 6:51 AM
New Diffusion model based docking tool #MolSnapper #cheminformatics #RDKit
Recently I think relationship between ML and CADD become more closer compared to several years ago. DiffDock is one of the very nice work for interation of cheminformatics and SBDD. It uses diffusion model for…
Recently I think relationship between ML and CADD become more closer compared to several years ago. DiffDock is one of the very nice work for interation of cheminformatics and SBDD. It uses diffusion model for…
I'm not happy with the way this all went down. I have tremendous respect for Gabriele and the rest of the DiffDock authors. Their work has broken new ground and helped advance machine learning in drug discovery. If I had to do it over again, I'd do things differently.
December 9, 2024 at 1:17 AM
I'm not happy with the way this all went down. I have tremendous respect for Gabriele and the rest of the DiffDock authors. Their work has broken new ground and helped advance machine learning in drug discovery. If I had to do it over again, I'd do things differently.
Upon rereading, the document is a bit callous for my taste. Temporal splits are standard in protein structure modeling (even if they shouldn't be), & I strongly suspect that the misuse of conventional docking methods when benchmarking ML docking software might not be unique to diffdock
December 5, 2024 at 4:28 PM
Upon rereading, the document is a bit callous for my taste. Temporal splits are standard in protein structure modeling (even if they shouldn't be), & I strongly suspect that the misuse of conventional docking methods when benchmarking ML docking software might not be unique to diffdock
See strong rebuttal from DiffDock authors bsky.app/profile/gcor...
Looks like the critique has several misrepresentations.
Looks like the critique has several misrepresentations.
Hi @anshulkundaje.bsky.social unfortunately, this paper from Jain et al. contains falsehoods, misleading comparisons, seemingly deliberate omissions, and is written in a tone not intended as a serious research paper. Please see our detailed response: www.linkedin.com/pulse/respon...
Response to Jain et al.
You may have seen a recent pre-print [1] from Jain et al. with strongly worded claims against the experimental results in our DiffDock paper [2].
www.linkedin.com
December 9, 2024 at 12:13 AM
See strong rebuttal from DiffDock authors bsky.app/profile/gcor...
Looks like the critique has several misrepresentations.
Looks like the critique has several misrepresentations.
In case you missed it, our Nov webinar with Gabriele Corso, Hannes Staerk, & Bowen Jing from the MIT Computer Science and Artificial Intelligence Laboratory (CSAIL)
introducing DiffDock, a state-of-the-art method for molecular docking, is now available on SBGridTV.
www.youtube.com/watch?v=_S-W...
introducing DiffDock, a state-of-the-art method for molecular docking, is now available on SBGridTV.
www.youtube.com/watch?v=_S-W...
DiffDock
Topic: DiffDock: a state-of-the-art method for molecular docking.Presenter: Gabriele Corso, Hannes Staerk, and Bowen Jing, Ph.D. students, MIT Computer Scien...
www.youtube.com
November 28, 2023 at 4:00 PM
In case you missed it, our Nov webinar with Gabriele Corso, Hannes Staerk, & Bowen Jing from the MIT Computer Science and Artificial Intelligence Laboratory (CSAIL)
introducing DiffDock, a state-of-the-art method for molecular docking, is now available on SBGridTV.
www.youtube.com/watch?v=_S-W...
introducing DiffDock, a state-of-the-art method for molecular docking, is now available on SBGridTV.
www.youtube.com/watch?v=_S-W...
Enter a "blinded challenge ring for a fair fight"
I hope that DiffDock & all the new methods coming out now, claiming amazingness (& let's be fair DiffDock is just one of many) do this at the earliest opportunity
CADD is a complex business, where MANY steps in the process can affect the results
I hope that DiffDock & all the new methods coming out now, claiming amazingness (& let's be fair DiffDock is just one of many) do this at the earliest opportunity
CADD is a complex business, where MANY steps in the process can affect the results
December 9, 2024 at 3:46 AM
Enter a "blinded challenge ring for a fair fight"
I hope that DiffDock & all the new methods coming out now, claiming amazingness (& let's be fair DiffDock is just one of many) do this at the earliest opportunity
CADD is a complex business, where MANY steps in the process can affect the results
I hope that DiffDock & all the new methods coming out now, claiming amazingness (& let's be fair DiffDock is just one of many) do this at the earliest opportunity
CADD is a complex business, where MANY steps in the process can affect the results
Many of you have seen Gabriele Corso's response to our recent preprint. If you haven't, it's here. www.linkedin.com/pulse/respon...
Response to Jain et al.
You may have seen a recent pre-print [1] from Jain et al. with strongly worded claims against the experimental results in our DiffDock paper [2].
www.linkedin.com
December 9, 2024 at 1:17 AM
Many of you have seen Gabriele Corso's response to our recent preprint. If you haven't, it's here. www.linkedin.com/pulse/respon...
DiffDock-Glide: a hybrid physics-based and data-driven approach to molecular docking [new]
Combines a diffusion model with physics-based refinement for improved molecular docking, especially on unseen protein targets.
Combines a diffusion model with physics-based refinement for improved molecular docking, especially on unseen protein targets.
June 5, 2025 at 4:32 AM
DiffDock-Glide: a hybrid physics-based and data-driven approach to molecular docking [new]
Combines a diffusion model with physics-based refinement for improved molecular docking, especially on unseen protein targets.
Combines a diffusion model with physics-based refinement for improved molecular docking, especially on unseen protein targets.
Boltz-1の開発グループはDiffdockと同じグループなんだけど、なんでかDiffdockと性能比較していないね あとよく言われるAutodockとも。
するまでもなく性能良くなっているんだろうけれど、論文の査読者的には「なぜ比較していない？」って言うのかな
するまでもなく性能良くなっているんだろうけれど、論文の査読者的には「なぜ比較していない？」って言うのかな
August 19, 2025 at 6:28 AM
Boltz-1の開発グループはDiffdockと同じグループなんだけど、なんでかDiffdockと性能比較していないね あとよく言われるAutodockとも。
するまでもなく性能良くなっているんだろうけれど、論文の査読者的には「なぜ比較していない？」って言うのかな
するまでもなく性能良くなっているんだろうけれど、論文の査読者的には「なぜ比較していない？」って言うのかな