biorxiv.org· Sakai, S. A., Nomura, R., Ohira, M., Kurihara, K., Chi, S., Matsubara, Y., Morinaga, Y., Kato, T., Takeda-Miyata, N., Kojima, M., Sakashita, S., Ebi, H., Yamaguchi, R., Oki, E., Fujisawa, T., Bando, H., Suzuki, A., Suzuki, Y., Tsuchihara, K., Kageyama, S.-I., Yamashita, R. ·· 19 hours agoPickScore58
SPATNIC enables H&E-grounded cancer-cell identification across spatial and single-cell transcriptomic modalities in colorectal cancer
In brief
SPATNIC is a variational autoencoder trained on Xenium cells labeled through pathologist review of matched H&E images to distinguish individual cancer cells from normal cells. The model matches or exceeds existing tools in primary colorectal cancer and liver and lung metastases, and generalizes zero-shot to Visium, Visium HD, CosMx, Xenium, MERSCOPE, and single-cell RNA sequencing.
Why it matters
SPATNIC is trained on pathologist annotations and generalizes across spatial and single-cell platforms to separate cancer cells from normal cells.
Source: biorxiv.org · biorxiv.org