Back

Multicellular Spatial Programs Define the Histopathological Architecture of Meningioma

Miyagishima, D.;Afrasiyabi, A.;McGuone, D.;Erson-Omay, E.;Yalcin, K.;Takeo, Y.;Duy, P.;Gultekin, B.;Yeung, J.;Ercan-Sencicek, A.;Henegariu, O.;Youngblood, M.;Mishra-Gorur, K.;Yasuno, K.;Wang, G.;Sestan, N.;Verhaak, R.;Moliterno, J.;Barak, T.;Gunel, M.

2026-06-17 Cancer Biology
10.64898/2026.06.16.732172 bioRxiv
Show abstract

Tumors are often described by cell types or gradients, but the organizing units of tumor tissue remain unclear. Using meningiomas, which show marked morphologic diversity despite constrained and recurrent genetics, we identify reproducible multicellular spatial molecular programs (SMPs) characterized by recurring cell-type mixtures that combine in different proportions across tumors. We built a multi-omic atlas of 147 human meningiomas (1.6 million cells/spots), integrating scRNA-seq, Visium, CosMx-RNA, and CosMx-Protein. Across platforms, SMPs mapped onto canonical whorl-lobule architecture and defined a structured ecological landscape linking hypoxic, immune-evasive states to vascularized, matrix-remodeling, and mineralization-rich states. For translation, we developed MeningNet, a hybrid ConvNeXt-Vision Transformer that infers SMPs directly from hematoxylin-and-eosin sections. MeningNet generalized across an internal replication cohort and 465 external whole-slide images, recovering cross-platform inference and showing significant association with CNS-WHO grade. These findings establish meningioma architecture as a reproducible histopathological framework for inferring spatial molecular state from routine pathology.

Matching journals

The top 8 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.