A genetic algorithm developed by UK scientists could aid the use of infrared (IR) spectroscopy in cancer diagnosis. 

IR spectroscopy is a non-destructive method for the analysis of cells, tissues and fluids that has been used to detect different diseases and stages of malignancy in infected prostate, cervical and colon tissues. But methods to process the cells before analysis are not standard throughout different laboratories, which can make the results difficult to interpret by non-specialists in hospitals.  

A collaboration led by Peter Gardner at the University of Manchester has investigated the IR signatures from prostate cancer using cells from the same source in the body. By biologically and chemically transforming the cells, the researchers mimicked tumour progression finding that distinguishing between the signatures of cells with different tumour behaviour is very difficult. 

Gardner developed a genetic algorithm - a computer program that learns and optimises solutions - to process the cells' IR data before it is analysed. They used 50 independent genetic algorithm runs to optimise the number of correctly classified spectra. With this method, they could separate this family of closely related cells studied with a high degree of accuracy. 

Max Diem, an expert in spectral diagnosis from Northeastern University in Boston, US, says this paper 'is a pioneering application of this newly developed technology, and outiles the power of discriminating cancer cell lines of different levels of invasiveness by spectral methods'. 

'The aim is to identify those tumours most likely to spread,' Gardner explains, and adds that with more work, the identification of IR signatures will be an invaluable tool which could be used to aid clinical decisions regarding treatment. 'At the moment it is beyond the proof of principle stage but not yet ready for full clinical trials,' he concludes. 

Amaya Camara-Campos 

Enjoy this story? Spread the word using the 'tools' menu on the left or add a comment to the Chemistry World blog.