In the push to develop better cancer treatments and to work towards a potential cure for this insidious disease, the study of cancer cells grown in culture dishes is crucial – but new research highlights some key genetic differences between these cells and the cancer cells that grow in the human body.
While that doesn’t mean lab research using lab-grown cells can’t be useful and informative, it’s important for scientists to know what these differences are as they look at ways to stop tumors from spreading and causing damage.
Researchers developed a machine learning model called CancerCellNet (CCN) to compare cancer cells in the body with cancer cells from four other sources: 26 mice models engineered to develop cancer; 415 mice with transplanted human cancer cells (xenografts); 131 balls of 3D tissue grown in a lab to mimic tumors (tumoroids); and 657 traditional cancer cell lines (cancer cells grown in culture dishes).
By comparing the RNA sequences of these cells – the biological instructions that determine how proteins develop – against a cancer genome database, the team was able to work out how similar they were to in vivo cancers on a genetic level.
“It may not be a surprise to scientists that cancer cell lines are genetically inferior to other models, but we were surprised that genetically engineered mice and tumoroids performed so very well by comparison,” says molecular biologist and geneticist Patrick Cahan from Johns Hopkins University.
On average, the genetically engineered mice and tumoroids had RNA sequences most closely matching actual human cancer in about 80 percent of the tumor types tested, including breast, lung, and ovarian cancers.
Cancer cell lines didn’t fare so well, with more discrepancies to the human tumors on record. In one example mentioned in the study, a cell line known as PC3 for prostate cancer actually more closely resembled bladder cancer. It seems that cell lines start to change once they’re out of their natural environment.
“RNA is a pretty good surrogate for cell type and cell identity, which are key to determining whether lab-developed cells resemble their human counterparts,” says Cahan.
“RNA expression data is very standardized and available to researchers, and less subject to technical variation that can confound a study’s results.”
The benefits of CancerCellNet are that it’s versatile and speedy: it’s certainly faster and less expensive than transplanting cancers into mice to see how they develop, which is one of the methods that scientists currently use to compare different models.
There are limitations to the study to bear in mind. As good as RNA is as a way of comparing cells, it doesn’t tell the whole story, and the researchers want to add more data to their CCN training database to make it more accurate.
In addition, it’s worth nothing that the study also looked at relatively few engineered mouse models and tumoroids, which may have skewed the results somewhat.
While this is just the start for CCN, it shows plenty of promise in being able to help researchers figure out just how realistic their cancer models are – and how reliable studies based on them are going to be when it comes to turning them into actual treatments. What’s more, it can be easily adapted for future cancer models too.
“Because CCN is open-source and easy to use, it can be readily applied to newly generated cancer models as a means to assess their fidelity,” the researchers explain in their paper.
The research has been published in Genome Medicine.
