Researchers from the University of Singapore have developed a non-invasive classification protocol to identify individual cancerous cells based on cell pH. The pH of cells ranged from 4.7 to 8.0, with cancerous cells generally having a more alkaline intracellular pH, and a more acidic extracellular pH. Using staining techniques that create quantifiable colours based on the pH of the cell, these colours were then ‘taught’ to the machine learning protocol to distinguish healthy and cancerous cells.
Visualisation techniques have previously focused on fluorescent probes or nanoparticles. Limitations to these methods exist however, requiring lengthy preparation protocols which limit their clinical utility, as well as causing irreversible damage and altering the cell physiology, meaning that the sample can no longer be used for diagnostic purposes afterwards. Using ethanol and Bromothymol Blue at a concentration of 0.5 mg/ml, the research team was able to produce a protocol that kept the cell alive and maintain the integrity of the cell physiology, whilst also ensuring adequate uptake of the staining chemical. Colour-staining profiles were then collected using various cell cultures, including healthy and cancerous breast tissue, pancreatic cancer tissue and human umbilical vein endothelial cells. These profiles were then given to a machine learning algorithm, which was able to detect cancerous cells with an accuracy of over 90%. Furthermore, the algorithm was able to distinguish between two breast cancer cells types in vivo, MCF-10A and MDA-MB-231, at rates of 78% and 77% respectively.
This technique has strong applicability to clinical practice, as it could potentially be used with non-invasive liquid biopsies. Furthermore, this protocol was realised with standard equipment that most, if not all hospitals have access to- a microscope and a colour camera, combined with open-access software. Chwee Teck Lim, one of the researchers explained “the ability to identify single cells has acquired a paramount importance in the field of precision and personalised medicine. This is because it is the only way to account for the inherent heterogeneity associated with any biological specimen”. Looking forward, the team hopes to apply this concept to a clinical setting to detect different stages of malignancies in real-time, in a fully automated system.