Clemens Isert
@clemensisert
Followers
958
Following
604
Media
16
Statuses
70
New review: Many geometric #deeplearning methods use the 3D structure of macromolecules to predict properties that are relevant for #drugdesign. Preprint: Work with @atzkenneth & Gisbert Schneider. #compchem #machinelearning #drugdiscovery (1/4).
3
38
147
Really cool to see Neo-1 released. Latent diffusion to combine structure prediction and molecular generation across modalities. #ai #drugdiscovery #genai.
Announcing Neo-1: the world’s most advanced atomistic foundation model, unifying structure prediction and all-atom de novo generation for the first time - to decode and design the structure of life 🧵(1/10)
3
3
25
With the scarcity of high-quality, leakage-free benchmarking datasets in the chem-bio space, it's really exciting to see this released today!.
💥 Introducing PINDER & PLINDER.With existing evals saturating without clear advances in real life downstream tasks, current progress in AI x Bio is primarily rate-limited by better datasets & evals. In two back-to-back preprints, we address this via fantastic academic-industry
0
3
21
RT @vant_ai: Building on success in ongoing projects, we are thrilled to announce the increased scope and range of our partnership with $BP….
0
5
0
Fantastic work led by @atzkenneth which I really enjoyed being a part of! Especially seeing low micromolar activities confirmed experimentally for molecules generated by deep learning models was a highlight! #compchem #pharma.
Happy to share our new preprint. We introduce the concept of interactome-based deep learning for ligand- and structure-based de novo molecular design. 🧬👨🏻💻. @clemensisert @ETH_en @ETH @ETH_DCHAB #MachineLearning #compchem #GenAI #chemtwitter.
0
0
10
These bond-critical points describe non-covalent interactions. We feed them into a 3D-aware graph neural network and use this representation to predict binding affinities.
1
0
4
We assess the hypothesis that we can predict protein-ligand binding affinity with geometric deep learning by using a representation based on the electron density. This fundamental physical property features characteristic locations in space, so called “bond-critical points”.
1
0
2
Excited to share the next manuscript from my PhD - “Exploring protein-ligand binding affinity prediction with electron density-based geometric deep learning”. With @atzkenneth, @rinikerlab & Gisbert Schneider @ETH @ETH_en. #compchem #machinelearning #quantummechanics
2
10
68
RT @atzkenneth: Happy to share our latest preprint. We show how graph transformer neural networks can be applied to in silico reaction scre….
0
3
0
Great to see our review “Structure-based drug design with geometric deep learning” published in @COSB_CRSB. Paper: (w/ @atzkenneth & Gisbert Schneider) #compchem #deeplearning.
New review: Many geometric #deeplearning methods use the 3D structure of macromolecules to predict properties that are relevant for #drugdesign. Preprint: Work with @atzkenneth & Gisbert Schneider. #compchem #machinelearning #drugdiscovery (1/4).
0
2
18
Happy to share the work from my internship at Novartis is now published, investigating different ML methods for lipophilicity prediction: Work with @jckromann, Nik Stiefl, Gisbert Schneider, and Richard Lewis.
2
13
105
RT @josejimlun: We now have open internship positions for mid-2023. If you're interested, please apply using the link below. Shoot me a DM….
0
11
0
Here's a short blog post on QMugs & DelFTa (work w/ @atzkenneth & @josejimlun), particularly from a usability perspective. Thanks to @TOSSOUPrudencio and @hsu_jonny from @valence_ai for the invitation! . #compchem #deeplearning
1
6
13
A graphical overview classifies key developments in the field by the approached task and the chosen macromolecular representation. (4/4)
0
0
5
…and of commonly used macromolecular structure representations. (3/4)
1
0
5
We’ve compiled a brief summary of the symmetry groups & operations relevant in the drug design space… (2/4)
1
0
5
RT @atzkenneth: Glad to share our most recent preprint. We introduce a geometric #deeplearning platform utilizing automated high-throughput….
0
5
0