Fig. 5: PACpAInt identification of molecular components to decipher intratumor microheterogeneity. | Nature Communications

Fig. 5: PACpAInt identification of molecular components to decipher intratumor microheterogeneity.

From: Pacpaint: a histology-based deep learning model uncovers the extensive intratumor molecular heterogeneity of pancreatic adenocarcinoma

Fig. 5

a Workflow for molecular components identification at tile-level. PACpAInt-Neo is first applied to identify neoplastic regions, followed by PACpAInt-Cell type to identify tumor cells and stroma, then PACpAInt-Comp to predict the molecular subtype of tumor cells and stroma, b Representative tiles identified as tumor classical or basal-like or stroma active or inactive by PACpAInt-Comp in the TCGA validation cohort (112 μm square), c PACpAInt-Cell type tumor and stroma score in tiles identified as classical, basal-like, stroma active or inactive by PACpAInt-Comp (analysis on 100 K tiles for each category (i.e., classic, basal, etc.). Center corresponds to the median, lower, and upper hingers to the first and third quartiles, whiskers to the hist/lowest value no further than 1.5 × IQR (inter-quartile range), d Example at the tile level of areas identified as classical or basal-like by PACpAInt-Comp and stained by immunohistochemistry with classical (GATA6/Claudin18) or basal-like (KRT17) markers (scale bar = 200 μm), e Correlation between slide-wise median stromal and epithelial scores. Source data are provided as a Source Data file.

Back to article page