Artificial intelligence to understand the activity of molecular functions

Scientists at Cima University of Navarra devise an innovative analysis method that improves understanding of gene activity and biological processes. This system demonstrates effectiveness in studying the progression of metastatic prostate cancer

Dr. Carlos Ruiz, Idoia Ochoa and Dr. Mikel Hernáez, from the Machine Learning in Biomedicine research group at Cima-CCUN

June 25, 2024

Researchers at the Cima University of Navarra have developed a new method for the analysis of gene expression data using artificial intelligence and machine learning, called NetActivity. The technique represents an innovation, as it allows the study of differences between molecular functions, in contrast to traditional studies that examine differences between individual genes.

"This neural network is composed of layers that represent a specific molecular function, which allows for a clearer interpretation of gene expression data. It also allows prioritizing the biologically most relevant genes for each function, facilitating their subsequent experimental validation," explains Dr. Carlos Ruiz Arenas, first author of the work and postdoctoral researcher at Cima University of Navarra.

This methodology defines activity values of the main molecular functions based on the expression values of individual genes using an autoencoder structure, a type of artificial neural network used to learn efficient data representations.

"This is a new advance that provides a deeper insight into understanding the activity of genes and their biological processes," details Dr. Mikel Hernáez, director of the Machine Learning in Biomedicine group at Cima, integrated into the Cancer Center Clínica Universidad de Navarra.

NetActivity has been successfully applied in a meta-analysis of the progression of prostate cancer, identifying gene sets related to cell division, a crucial element for the progression of the disease. "In the case of metastatic prostate cancer, changes were observed in gene sets associated with drug resistance, highlighting the ability of NetActivity to reveal relevant information to understand and treat the disease," points out Dr. Mikel Hernáez.

This innovative approach is now publicly available in the Bioconductor and GitHub repositories, offering new possibilities for the experimental validation of research results and data analysis in the field of molecular biology.

The research has received public and private funding from the Ministry of Science, Innovation and Universities through the Spanish Research Agency, the Government of Navarra through European Next Generation funds, the Scientific Foundation of the Spanish Association Against Cancer, and the U.S. Department of Defense medical research program, among other institutions.

Bibliographic reference:

NetActivity enhances transcriptional signals by combining gene expression into robust gene set activity scores through interpretable autoencodersNucleic Acids Research