Technical University of Munich

A VAE trained on scRNA-seq reference data and applied frozen at inference to impute unmeasured genes and denoise spatial transcriptomics profiles.

Latent flow-matching foundation model that predicts pan-cancer spatially-resolved single-cell gene expression directly from routine H&E histology slides.

A LoRA adapter on ProstT5 that predicts per-residue probability distributions over Foldseek 3Di tokens, capturing sequence-encoded conformational flexibility from MD trajectories.

Predicts single-cell-resolution scRNA-seq coverage and scATAC-seq insertion profiles directly from DNA sequence by adapting the Borzoi predictor with a cell-specific decoder.

Decoder-only transformer that reframes ancestral recombination graph inference as next-token prediction, estimating coalescence times from genetic variation at scale.

Grounded multimodal large language model for endoscopic surgery, supporting visual dialogue, region-based question answering, and bounding-box grounding across surgical scene understanding tasks.

Transformer foundation model pretrained on 110M single-cell and spatially resolved transcriptomics profiles, enabling spatial context prediction for dissociated cells.

A bilingual protein language model that translates bidirectionally between amino acid sequences and the 3Di structural alphabet, enabling inverse folding and structure-aware embeddings.

Masked DNA language model trained on 800+ species spanning 500M years of evolution, using explicit species conditioning to capture conserved regulatory elements.

Optimized protein language model that surpasses state-of-the-art performance with fewer than 10% of the parameters of comparable models.

A suite of six protein language models — including ProtBERT and ProtT5 — trained on up to 393 billion amino acids using large-scale HPC infrastructure.