Deep Learning

Foundational Models for Genomics: Decoding Genomic Sequences
Workshop hosted by BLISS.
2025-12-11 06:00 Technical University Berlin
Miguel Arbesú Andrés, Bernardo Almeida, Jonas Hirsch
Foundational Models for Genomics: Decoding Genomic Sequences
GeoGraph: Geometric and Graph-based Ensemble Descriptors for Intrinsically Disordered Proteins
GeoGraph is a lightweight, simulation-informed transformer model that predicts aggregate geometric properties of intrinsically disordered protein (IDP) conformational ensembles directly from their amino acid sequence.
Eoin Quinn and, Marco Carobene and, Jean Quentin and, Sebastien Boyer, Miguel Arbesú Andrés, Oliver Bent
PDF Cite Code DOI
GeoGraph: Geometric and Graph-based Ensemble Descriptors for Intrinsically Disordered Proteins
Generative Model for Small Molecules with Latent Space RL Fine-Tuning to Protein Targets
We introduce SAFER, an updated molecular encoding scheme that improves the correctness of the representations, and use it to build a latent-variable transformer model for generative tasks.
Ulrich A. Mbou Sob, Qiulin Li, Miguel Arbesú Andrés, Oliver Bent, Andries P. Smit, Arnu Pretorius
PDF Cite DOI
Generative Model for Small Molecules with Latent Space RL Fine-Tuning to Protein Targets
FrameDiPT: SE(3) Diffusion Model for Protein Structure Inpainting
We introduce FrameDiPT, a diffusion model for protein structure inpainting that captures conformational distributions.
Cheng Zhang, Adam Leach, Thomas Makkink, Miguel Arbesú Andrés, Ibtissem Kadri, Daniel Luo, Liron Mizrahi, Sabrine Krichen, Maren Lang, Andrey Tovchigrechko, Nicolas Lopez Carranza, Uğur Şahin, Karim Beguir, Michael Rooney, Yunguan Fu
PDF Cite DOI
FrameDiPT: SE(3) Diffusion Model for Protein Structure Inpainting