Leon Hetzel (@leon_het) header
Leon Hetzel (@leon_het) profile picture

👨‍💻PhD student at TUM and Helmholtz Munich, Generative Modelling, Graphs, Applications in single-cell and drug discovery

Munich, Bavaria
Joined May 2019
Photo by Leon Hetzel (@leon_het)Photo by Leon Hetzel (@leon_het)Photo by Leon Hetzel (@leon_het)Photo by Leon Hetzel (@leon_het)Photo by Leon Hetzel (@leon_het)Photo by Leon Hetzel (@leon_het)Photo by Leon Hetzel (@leon_het)
Leon Hetzel retweeted
Finding good resources on efficient AI is harder than it should be. We're fixing that! 🚀 Check out our new github.com/PrunaAI/awesome-a… repo —a curated hub of the best tools, papers, and techniques to make AI faster, smaller, and cheaper.
3
18
656
Congrats to my amazing PhD students: We have 9 papers accepted at #ICLR2025. Reliability, AI4Science, graphs, LLMs, and more (go.tum.de/936150). And if you follow the recent discussions about AI efficiency, you might like our blog and webinars (pruna.ai).
4
14
173
21,174
Unified Guidance for Geometry-Conditioned Molecular Generation • UniGuide introduces a unified framework for geometry-conditioned molecular generation using diffusion models. It provides a flexible method for controlled molecular design, eliminating the need for additional training or specialized networks. • Key innovation: UniGuide employs “condition maps” to translate geometric constraints into actionable guidance, enabling applications in structure-based, fragment-based, and ligand-based drug design with a single model. • Unlike traditional conditional models, UniGuide excels in handling diverse geometric constraints like surfaces, densities, and volumes. This versatility allows it to generalize across a range of tasks without the limitations of task-specific models. • Performance benchmarks show UniGuide’s superior or comparable results to specialized approaches in drug discovery tasks, demonstrating its adaptability and efficiency. • Applications include ligand-based drug design (LBDD) for generating novel molecules that match predefined shapes, structure-based drug design (SBDD) for ligand-pocket interactions, and fragment-based drug design (FBDD) for linking and growing molecular fragments. • UniGuide’s method decouples training from conditioning, making it robust in data-scarce scenarios. It demonstrates state-of-the-art performance across ligand similarity, binding affinity, and drug-likeness metrics. @guennemann @fabian_theis @j_m_sommer @leon_het @sirine_ayadi1 📜Paper: openreview.net/pdf/704d9702f… #MolecularDesign #DrugDiscovery #GeometryConditioning #AIinBiology #GenerativeModels
5
25
3,517
11 exciting news: Our group has 10 papers at #NeurIPS2024 (incl. 1 oral + 2 spotlights) 📃🎓. And as of October 1st, I am on entrepreneurial leave 🚀. Re papers: Congrats to all co-authors. Amazing work! go.tum.de/689644 Re startup: We are hiring! pruna.ai
5
9
131
20,738
Presenting at #ICML2024: Fragment-Biases for Molecular GNNs 🧪 Tue, 23.07: Oral session 1F @ 11:00 Poster #105 @ 11:30 🔑 Fragment-Biased GNNs outperform others, match Transformers with better generalization & linear cost! 🤝 With N. Kemper, @leon_het , @j_m_sommer , @guennemann
1
13
1
18
2,802
Mixed Models with Multiple Instance Learning (MixMIL) received an Oral & Outstanding Student Paper award at @aistats_conf last week! 🏆 MixMIL enables accurate & interpretable patient label prediction from single-cell data by adding attention to GLMMs.#singlecell #MachineLearning
2
13
1
47
12,432
Get to know #mcml junior member @leon_het 💡 He developed an algorithm called chemCPA designed to make drug discovery more effective. Read more (in German): mcml.ai/news/2024-03-26-mcml… This article was written by @leonie__fischer from @DJSde!
1
10
717
Thrilled to present our paper "(Provable) Adversarial Robustness for Group Equivariant Tasks" at #NeurIPS2023! Joint work with great collaborators @YanScholten @guennemann! Paper: arxiv.org/abs/2312.02708 Poster #800 Thursday 11 am - 13 pm #NeurIPS
1
9
16
1,687
Is binarization of scATAC-seq data really necessary? The conclusion from our analysis is that a quantitative treatment is in fact beneficial. Now out in Nature Methods! @gagneurlab @fabian_theis nature.com/articles/s41592-0… Many additions since the preprint 👇(1/n)
4
97
5
317
51,115
Do you care about uncertainty quantification? Do you like guarantees? Check out our #ICML2023 paper where we bring conformal prediction to Graph Neural Networks. tldr: Instead of a single label return a *set* that is guaranteed to contain the true label. Set size = uncertainty.
Prediction sets for node classification with a guarantee of covering the true label? Also, without any assumption on the data distribution and the model? Check our last paper at #ICML2023 #ICML23 on "Conformal Prediction Sets for Graph Neural Networks" openreview.net/forum?id=zGf8…
16
1
124
20,206
If you are interested in ML potentials and/or uncertainty estimation come and visit our presentation of "Uncertainty Estimation for Molecules: Desiderata and Methods" this week at #ICML! Tue, 11am, #415 Joint work with @n_gao96 @Bertrand_Charp @amine_ketata @guennemann 1/2
1
11
2
32
3,508
1/9 CPAis finally published. It can predict single-cell responses to combinatorial perturbation (drugs, CRISPR). This is a joint collab between @meta and @fabian_theis @HelmholtzMunich. Read the thread to understand the LEGO analogy! embopress.org/doi/full/10.15…
4
64
3
188
91,694
Happy to announce our new paper on Deterministic Uncertainty Methods @TMLunLimited #ICLR2023 ! We dissect how the design of training schemes, architecture, and prior can significantly impact feature collapse and uncertainty performance! w/ C. Zhang and @guennemann
1
12
35
3,811
📺New video: Curvature for Graph Learning📺 Presenting, among other things, current work with (i) @JoshSouthern13, @jeremy_wayland, @mmbronstein and (ii) @CorinnaCoupette and Sebastian Dalleiger. Very excited about this research direction! youtube.com/watch?v=-V0JpdAF…
1
12
3
63
8,021
A new method from @johwirth @celia4science @Meier_Lab enables cost effective spatial transcriptomics on multiple tissues in parallel. @PioneerCampus
1
2
5
7,743
There are some collaborations that are just fun and this was one of them! Big thank you to @ab_meier and Alessandra Moretti and everyone else involved!🫀
Epicardioid single-cell genomics uncovers principles of human epicardium biology in heart development and disease go.nature.com/3zr2PkH
1
2
14
1,262
If you are at #neurips2022 and are interested in the robustness of Graph Neural Networks then stop by our poster (#906) this afternoon (Poster Session 6). Joint work with @LoadingByte, @guennemann and @abojchevski
17
56
Interested in sparsity, generalization, and neural ODEs? I’ll present “Sparsity in continuous-depth neural networks” at #NeurIPS, at 11:30 am in Hall J #121. 📜: arxiv.org/abs/2210.14672 🧑‍💻: github.com/theislab/PathReg Thanks @AileeHana, @k__niki @fabian_theis, Mikhail, and Ignacio! – at New Orleans, LA
(1) Continuous-time Neural Nets are mainly used for predictive tasks such as classification or regression. To learn the underlying dynamics that govern the evolution of the observed trajectory, some inductive biases (or preferences) over the space of all functions are necessary.
5
17
We start our Adventskalender with Laura Martens @lauradmartens! Together with @soroorh1 & @ArrudaAnaLuiza, she started the amazing STEM-Tisch initiative this year & we are all looking forward to more events in 2023!😍 A special thanks to @EleZeggini & @dana_peer for your support!
1
5
37