Sallese, M. R. | Search bioRxiv

bioRxiv2025

Mapping and reprogramming microenvironment-induced cell states in human disease using generative AI

Tissue microenvironments reprogram local cellular states in disease, yet current computational spatial methods remain descriptive and do not simulate tissue perturbation. We present MintFlow, a generative AI algorithm that learns how the tissue microenvironment influences cell states and predicts how tissue perturbations can reprogram them. Applied to three human diseases, MintFlow uncovered distinct pathogenic spatial reprogramming in inflammatory and tumor microenvironments. In atopic dermatitis, MintFlow identified a novel, spatially-imprinted, type 2 (IL13+ITGAE+) epidermal T resident memory cell population (type 2 TRM), and decoded signaling pathways within the perivascular lymphoid niche. In melanoma, MintFlow identified fibrotic stroma resembling keloid scar tissue. In kidney cancer, MintFlow resolved immunosuppressed CD8+ T cell states within tertiary lymphoid structures. Furthermore, MintFlow enabled in silico perturbations of disease-relevant cell states and tissue environments. Regulatory T cell modulation in atopic dermatitis was predicted to suppress the pro-inflammatory tissue environment, supporting manipulation of these cells as a therapeutic target. In kidney cancer, in silico T cell replacement recapitulated immune checkpoint blockade, while spatially targeted macrophage depletion reverted immunosuppressed T cell states. The corresponding gene programs correlated with survival in large kidney cancer patient cohorts. Together, these findings position MintFlow as a tool for unbiased disease mechanism prediction and in silico perturbation, accelerating translational hypothesis generation and guiding therapeutic strategies.

genomics↗

bioRxiv2025

A harmonized ovarian cancer scRNA-seq atlas to dissect disease heterogeneity underlying metastatization and chemoresistance

The advent of single cell technology has enabled researchers with the ability to achieve unprecedented resolution in the characterization of biological systems. The consequent increasing availability of single cell dataset brought about the possibility to obtain comprehensive reference atlases, to shed light on the molecular and cellular foundations of healthy and diseased tissues. This process has highlighted the need to integrate datasets from different sources while being able to distinguish true biological signal from technical confounders. While this issue is widespread to most biological settings, it holds especially true for cancer samples, which are characterized by a diffused inter- and intra-patient phenotypic heterogeneity. To address this issue, here we developed a novel integration method tailored to highly heterogeneous single cell transcriptomic data and applied it to one of the quintessential heterogeneous cancer type, namely high-grade serous ovarian cancer, to generate the first reference atlas for this disease. By identifying patient-specific cell populations and deriving metacells, we were able to preserve inter-patient biological variability. Using a variational autoencoder, we integrated metacell data, revealing an evolving landscape of cell states along disease progression and treatment for each of the main cell types constituting the dataset. Also, we showed the potential of this resource by identifying diffused and tissue/treatment-specific cell-to-cell interactions. Finally, the generated integration model allowed to expand the atlas with additional data, granting iterative refinement over time of this disease reference. Our strategy now provides a valuable resource for the cancer research community, facilitating the investigation of tumor heterogeneity towards the development of novel therapeutic strategies.

cancer biology↗