Fig. 3: DeepTalk-CCC detects spatially CCCs for mouse visual cortex region from MERFISH data.

a Probabilistic mapping of snRNA-seq data on MERFISH data. Assigns a likelihood score to each cell subset within the three main categories, indicating their potential cell type. b Deterministic mapping of cell types. Displays segmented cells on the MERFISH slide, color-coded based on the most probable snRNA-seq profile mapped using DeepTalk-Integration. c ROC Curve Evaluation of DeepTalk-CCC (DT-CCC) using the MERFISH Dataset. This curve illustrates the performance of DT-CCC in terms of true positive rate vs. false positive rate. d DES Rankings of CCC Tools Evaluation. A comparative analysis of various cell–cell communication tools, ranked according to their performance in the evaluation process. The boxplots display data distribution where the box spans from the first to the third quartile, marking the median with a distinct line. The whiskers reach out to the maximum range within 1.5 times the interquartile range, and individual outliers are denoted by separate dots; n = 5 repeated independent tests. e Comparison of communication scores between adjacent and distant cell types. Highlights spatial proximity-based differences in communication scores between Layer 4 (L4) cells and other cell type, including L2/3 intratelencephalic neurons (IT), L5 IT, L5 pyramidal tract neurons (PT), and L6 PT. The boxplots display data distribution where the box spans from the first to the third quartile, marking the median with a distinct line. The whiskers reach out to the maximum range within 1.5 times the interquartile range, and individual outliers are denoted by separate dots; N: number of predicted L-R(ligand–receptor) pairs. f Visualization of CCC via low-dimensional embedding. Each point in the graph represents a CCC event, with the cell serving as the receptor cell. g Comparison of the CCC score between spatially adjacent and distant cells at single-cell level. This plot compares the communication scores between spatially adjacent cells and distant cells at the single-cell level, revealing differences in communication strength based on proximity. The boxplots display data distribution where the box spans from the first to the third quartile, marking the median with a distinct line. The whiskers reach out to the maximum range within 1.5 times the interquartile range, and individual outliers are denoted by separate dots; n = 169 L4 cells. h Local microenvironment viewed via low-dimensional embedding. Each point in this diagram represents a collection of CCC events where the cell of interest serves as the receiver. i Predicted CCCs mediated by Apoe - Grm5 pairs. It includes the scores of CCCs mediated by Apoe - Grm5 pairs between glutamatergic cells (left) and the predicted CCCs from L2/3 IT cells to L4 cells (right). Source data are provided as a Source Data file.