The molecular pathology of cancer is still limited to the use of a small number of biomarkers rather than the examination of all genes.
Part of this limitation is due to the challenges of computerised data analysis. To overcome this problem, researchers at the Royan Research Institute, Sharif University of Technology, and Colorado State University introduced deep neural networks (DNNs) that can infer the various properties of biological samples at the same time.
It encrypts all the transcriptome part of the genome on a small vector lattice (a computer-generated input image), then retrieves the mRNA and miRNA profiles based on tissue and the type of disease.
This method works much more efficiently than the original gene data to differentiate samples based on tissue and disease.
Researchers have used this method to examine 10,750 samples from 34 different groups (one healthy group and 33 groups with different types of cancer) from 27 tissues.
The results of the study, published in the International Journal of Scientific Reports, showed that the method developed in this study was significantly better than previous methods and conventional methods of using source tissue data, health status or disease, and it can predict the type of cancer in each sample.
For tissues with more than one type of cancer, 99.4% of all types of cancer were correctly diagnosed.
All in all, the results showed that using artificial intelligence can be very useful in molecular pathology of cancer and oncological research.
This project was implemented by Behrouz Azar Khalili, Dr Ali Sharifi Zarchi, Ali Saberi and their colleagues at Royan Research Institute, Sharif University of Technology, and Colorado State University.