As the medical industry continues its quest for accurate and reliable diagnostic tools, more is becoming understood about the science of the sniff. The olfactory system of mammals is a powerful bio-sensor, its complex network of receptors capable of receiving and distinguishing between a near-unlimited number of odours and other chemical compounds.
But how can this knowledge be used to design diagnostic tools? GlobalData’s medical editor Chris Lo investigates.
There is perhaps no better demonstrator of the power of mammalian olfaction than man’s best friend. Dogs have around double the number of olfactory biosensors when compared to humans, providing a much greater specificity of detection.
Recently, cancer screening firm BioScentDX presented a study that found that four trained beagles were able to correctly identify lung cancer samples with 96.7% accuracy, offering the prospect of a non-invasive test and more frequent screening, which is usually limited by the radiation exposure involved.
However, rather than using dogs, a University of Bristol spinout company called Rosa Biotech is taking a different approach. The company was incorporated in January 2019 to develop and commercialise products based on the academic work of Professor Dek Woolfson at the Bristol BioDesign Institute. Woolfson and his team built a series of barrel-shaped proteins that mimic the mammalian olfactory system, but with a simpler structure that makes them easier to produce and handle.”
Arrays of ‘barrels’ are loaded with dye and produce colored patterns when exposed to analysis samples. The patterns are then analyzed using machine learning.
Rosa Biotech CEO Dr Andy Boyce told GlobalData:
The more we come to understand the biology of diseases, the more we realize how little we know about how much variability there is with diseases, especially in chronic diseases, from individual to individual.
There’s a need to do an awful lot of training on our side. If and when we get to the level of releasing a diagnostic test it will have to be backed up by some very large cohorts and sample data, so that we can be totally confident that the results we give are robust and accurate. That will be underpinned by machine learning, but also all of the stats and databasing and other activities that sit behind that.
We’ve got some early proof-of-concept data, but we need to get something that is convincing to clinicians and to potential diagnostic technology partners. It depends exactly what our business model is, but we may choose to partner with one of the big diagnostic technology developers like Abbott or Roche rather than taking it all the way through the clinic ourselves.”