A new way for analyzing toxic chemicals in complex samples was developed by a group of researchers at the University of Leicester. The research is significant as it studies how mammals smell and taste and may lead to the reduction of the need for animals in biomedical and chemical testing laboratories.
The research involves a mixture of human skin cells and environmental-sensitive fluorescent dyes, combined in a fluorescent assay. The mixture generates unique fluorescence spectra that are used to determine certain physiological conditions. The optical signal is then processed further using multivariate data analysis to provide qualitative information as well as fast diagnostics.
The electronic noses and tongues systems are what inspired the research. The operating scheme of these systems mimic how mammals smell and taste. The systems specifically combine semi-specific sensors and chemometric techniques as the means for biochemical process monitoring.
Similar theories and concept has been used by the Leicester Biotechnology Group at the University of Leicester only that instead of using electronic sensors, they used dyes array applied to human cells.
So what does this research answer? The mixture can now tell us about the toxicity of the chemical compound based on complex fluorescent spectra.
A digitized fluorescence image, being a very high-dimensional vector, can take more 250,000 numbers each. However, the number of tested chemicals is much less.
Thus, the challenge of reducing the dimensions is the first task to be done, to which the researchers strongly agree. Of course, the dimensionality reduction must be done in such a way that all vital information is unaffected.
“Firstly, we represented each signal by its projections on other signals. Secondly, we applied the classical and very popular model reduction method, Principal Component Analysis, and found five main components of the signals,” explained Alexander Gorban, one of the researchers of this study. Gorban is also a Professor of Applied Mathematics in the above mentioned University.
Gorban and his team then used dozens of various linear and nonlinear data analysis methods for signals that take up to five dimensions. The classifiers are then validated on the data that was previously unseen.
“Our approach can be considered as ‘explicit deep learning’, an explicit version of widely popularized (implicit) deep learning algorithms,” adds Gorban.
The results are promising as the sensitivity as well as specificity both measured well above 90%.
Professor Sergey Piletsky from the Department of Chemistry of the same university agrees that this finding is significant not only in the area of toxicology but also in the development of sensor arrays for chemical analysis. Piletsky comments that the sensor arrays for rapid quantification of a wide range of analytes have always been the obstacle for most analytical scientists.
Piletsky adds that the discovery may also reduce the need for animals to be used in chemical test and health diagnostics laboratories.
More information can be found at: University of Leicester.