Anomaly pathway activities and transcriptome burden reveal cellular plasticity of human cancer

Abstract

Abstract Background: It is now widely acknowledged that the adaptive evolutionary process of cancer cells is characterized by genomic and epigenetic alterations. Nevertheless, the extreme heterogeneity of cancer cells and the extreme sparsity of scRNA-seq data limited the understanding of cancer cell reconfiguration of anomaly pathway activity and adaptive evolution under selection pressure. Methods: We collected a total of 7,955 tumor samples, including 32 common human cancer types as the training set. Five sets of scRNA-seq data from different studies covering 5 different cancer types were used as the validation set. Results: In this study, we used a contrastive learning approach to analyze scRNA-seq data and developed a sceAPA model to determine the spatial mapping correlation between RNA features and genomic variation-associated anomaly pathway activities in cancer. The dynamic transcriptional profile revealed by scRNA-seq reflects how cancer cells reconfigure their transcriptome pattern and develop various APA patterns in response to selective pressure to gain an evolutionary advantage. We found that the plasticity of human cancer was negatively related to the transcriptome burden, and increasing transcriptome burden was associated with more chaotic APA patterns. Conclusions: Our study shows that cancer cells possess greater plasticity by reprogramming immune-associated transcript patterns, thereby conferring greater resistance to selection pressure. The plasticity of cancer cells is closely related to adaptive evolution, drug resistance, distant metastasis, and immune escape.