Fig. 2

Identification of malignant cells in GGN by InferCNV and characterization of epithelia from different stage by WGCNA. A Heatmap showing large-scale CNVs for individual cells (rows) for all the EPCAM + cells as input, immune and stromal cells were as references (top), and large-scale CNVs were observed in malignant cells (bottom), with the color shows the log 2CNV ratio. Red represents gene amplifications and blue represents deletions. B Validation of inferCNV results. DEG analysis on the malignant and normal cells we inferred in the EPCAM + cells from AIS and IAC samples, and same cells randomly spited into to subsets for three times, with genes of interested was framed with black box (Left). Significantly enriched Hallmark pathways in inferred normal and malignant cells as determined by GSVA score (Right) C Scale-free topology model fitting, with the horizontal axis representing different soft thresholds and the vertical axis representing the value of the model (Left top); Average degree of connectivity, with the horizontal axis representing different soft thresholds and the vertical axis representing the average degree of connectivity (Left bottom). Clustering of different gene modules (Right middle). Abbreviation: AIS: Adenocarcinoma in situ; CNV: Copy Number Variation; DEG: Differential expressed gene; GSVA: Gene set variation analysis; IAC: Invasive adenocarcinoma; WGCNA: Weighted correlation network analysis