ScribbleDom overview. ScribbleDom can work through two different pipelines, when the human annotator scribbles over the histology image (upper part) and when the output of a non-spatial clustering algorithm (e.g. mclust) is used as prior knowledge (lower part). ScribbleDom receives the prior knowledge about the spots, the preprocessed transcriptomics data, and the spatial information about each spot as its input. Data are preprocessed by taking highly variable genes (HVG) and then performing principal component analysis (PCA) on these HVGs. ScribbleDom identifies domains in the ST data using Inception by minimizing a loss function having two components—feature similarity loss and scribble loss. The results produced by ScribbleDom show significant improvement compared to state-of-the-art models.