Paul-Henry Cournède

ALHENA improves transcriptomic signature prediction and cell-type separability using H&E images alone, without requiring spatial transcriptomics data.

CellScribe is a generative framework that unifies cell identity and state by using a language model to generate cell type annotations and meta-program profiles, outperforming baselines and identifying new cell types through in-context learning.

The paper shows clonal memory in human HSPCs persists across divisions and is disrupted in AML, with potential for therapeutic modulation.

SHERLOC is a deep learning model for survival analysis using longitudinal ctDNA data, showing improved discrimination and calibration over existing methods in non-small-cell lung cancer.

The paper introduces STHELAR, a multi-tissue dataset integrating spatial transcriptomics and H&E images for cell type annotation in cancer research.

Novae is a graph-based foundation model for spatial transcriptomics that improves domain inference and batch correction across multiple tissues and technologies.

This paper introduces a bimodal model for bulk RNA-seq and DNA methylation that uses masked language modeling to learn representations, achieving state-of-the-art performance in cancer classification and survival prediction.

BulkRNABert uses a pre-trained transformer model on bulk RNA-seq data to improve cancer classification and survival prediction across multiple cancer types.

Sopa provides a technology-invariant pipeline for analyzing image-based spatial-omics data, offering unified visualization and analysis tools.