A new visualization method combined with machine learning will be able to reveal previously hidden information in micrographs of cells. This will help to reveal new information about gene expression levels.
Scientists from the James Watt School of Engineering at the University of Glasgow used image analysis and machine learning as a tool to determine gene activity in cells.
The research teams used detailed image analysis to extract more than a thousand mathematical values describing each cell being analyzed, commonly called morphometric descriptors. By combining these values, they taught the computer the relationship between morphometric values and actual levels of gene expression.
This approach is similar to the types of machine vision that are already used in devices such as mobile phones and autonomous cars. In these devices, algorithms are able to identify objects based on large sets of training data. In a new work, the technique allowed not only to distinguish between cell types, but also directly predict the activity of genes in each cell.
Professor of Biomedical Engineering, Nikolai Gadegard of the University of Glasgow, noted that “the microphotographs we have collected over the years have much more information. Thanks to modern methods of computer technology, we have now seen that even small changes in the genome directly affect the cells”.
This technique could pave the way for obtaining much more information from microscopy data than is possible now.