Why are the results of cancer cell experiments always "different"? Cancer cells in a petri dish have nothing in common with cancer cells in the human body

Cancer has always been one of the arch nightmares among all human diseases. It is also known as the "king of all diseases" because of its complexity and incurable nature. In order to find a cure for cancer, scientists often cultivate various cancer cell lines in a laboratory environment and test the effects of different drugs on these cancer cell lines. It is worth noting that many researchers obtain cancer cell lines through different channels, but few people verify whether these cancer cell lines are correct. Recently, researchers from Johns Hopkins University School of Medicine in the United States published a research paper entitled Evaluating the transcriptional fidelity of cancer models in Genome Medicine. Based on the idea of exploring or improving research models in cancer laboratories, this research developed a new, self-learning computer model—CancerCellNet to measure the similarity between different cancer research models and native tumors. Results show that among several commonly used tumor models, human cancer cell lines grown in petri dishes are genetically the least similar to cancer cells in humans. Genetically engineered mice (https://www.creative-biolabs.com/drug-discovery/therapeutics/genetically-engineering-models-gems.htm) and 3D tumor models are more similar to cancer cells in humans. The construction of cancer models is an important way to study cancer biology and identify potential treatment methods. Among them, the most popular modeling methods are cancer cell lines (CCLs), genetically engineered mouse models (GEMMs), and human tumor xenograft models (PDXs). ) And 3D tumor models (tumoroids). The generalization and ability of the model come from the fidelity of the tumor type it represents. Therefore, no matter which model, in principle, it should be as close as possible to the corresponding cancer type in the patient. Unfortunately, many researchers are not clear about the similarities between different cancer models and corresponding native tumors. In this study, the team developed a computer model based on machine learning, CancerCellNet, using RNA in cells as the research object. Corresponding author Patrick Cahan said that RNA is a very good substitute for cell type and cell identity, which is the key to determining whether cells grown in the laboratory are similar to human cells. The RNA expression data is very standardized, which is not susceptible to the influence of mutations to confuse the research results. Experts said that perhaps it is not surprising that cancer cell lines are genetically inferior to other models. Due to the complex differences in the growth environment, cancer cell lines grown in the laboratory are not the same as human-derived cancer cells. In other words, once the tumor is removed from its natural environment, the cell line starts to change.rnFor example, the research team found that prostate cancer cells from PC3 began to look more like bladder cancer genetically. It is also possible that the cell line was initially labeled incorrectly, or that the cell line was actually extracted from bladder cancer. But in any case, only from a genetic point of view, this prostate cancer cell line does not represent a typical prostate cancer patient. All in all, the research developed the CancerCellNet tool to record cancer models with the greatest transcription fidelity to natural tumors, and found cancer models that did not match the annotated classification. By comparing models of different categories, among several commonly used tumor types, genetically engineered mice and 3D tumor models have higher transcription fidelity than human xenograft models (https://www.creative-biolabs.com/drug-discovery/therapeutics/xenograft-models.htm) and cancer cell lines. Therefore, researchers must choose carefully before applying cancer models.