This sorting and shifting process has allowed researchers from Linköping University to discover new groups of disease-related genes. The basis of the technology should help to advance precision (personalized) medicine leading to more reliable forms of individualized treatments for different conditions.
The biotechnology is orientated towards constructing maps of biological networks. These networks relate to how different proteins or genes interact with each other. This is a complex task and the application of artificial intelligence has helped to make the process easier.
The specific form of artificial intelligence used is a type of deep learning called “artificial neural networks“. To develop the ‘connectionist system’, the platform needs to be trained using biological data.
Over time the researches demonstrated how their “artificial neural network is capable of analyzing vast amounts of complex data.
This was demonstrated by using a data set containing 20,000 genes relating to hundreds of people. The information was presented “unsorted.” The artificial intelligence was then used to assess which gene expression patterns related to people with diseases, in contrast to those genes that related to healthy people. To do so, the system needed to find hidden datasets among protein proteins and cell types.
This exercise enabled the researchers to train the artificial intelligence to find patterns of gene expression in new datasets.
According to lead scientist Sanjiv Dwivedi: “We have for the first time used deep learning to find disease-related genes. This is a very powerful method in the analysis of huge amounts of biological information, or ‘big data’.”
The next phase of the project will involve teaming up with other research facilities in order to explore how medicines tailored to specific patients can be developed, by an assessment of a given patient’s genetic pattern.
The research has been published in the journal Nature Communications, where the research paper is titled “Deriving disease modules from the compressed transcriptional space embedded in a deep autoencoder.”