MOGONET offers a comprehensive analysis of biological procedures concerning human diseases

Sep, 2021 - By WMR

Researchers designed and tested MOGONET, an innovative multi-omics data analysis algorithm, and computational framework, to fully utilize developments in omics technology.

MOGONET, short for Multi-Omics Graph Convolutional Networks, performs better existing supervised multi-omics integrative analysis methods of various biomedical classification applications using microRNA expression data, mRNA expression data, and DNA methylation data, according to a study published in Nature Communications. MOGONET can also detect relevant omics signatures and biomarkers from various omics data types, according to the researchers.

MOGONET was evaluated on datasets from breast invasive carcinoma, gliomas, Alzheimer's disease, and kidney cancer, as well as datasets from healthy patients. MOGONET easily excelled existing supervised multi-omics integration algorithms, according to the researchers. MOGONET was also able to forecast the development of new cancer subtypes, as well as tumor grade and disease development. It can distinguish between Alzheimer's disease and normal brain activity.

Researchers use machine learning based on a neural network with MOGONET, the new AI tool, to capture complicated biological process interactions. They have improved the study of omics, and are discovering more about disease subtypes that biomarkers can help them distinguish, said Kun Huang, Regenstrief Institute Research Scientist. The main objective is to improve disease prediction and forecasting of disease outcomes. As a bioinformatician, he credits the MOGONET research group's diversity, which includes computer scientists, data scientists, and bioinformaticians, as well as their diverse perspectives, for the project's development and success.

Kun Huang and Jie Zhang intend to enhance their research further than omics to include imaging data, highlighting the number of brain images for Alzheimer's disease and cancer-related pathological images that can train MOGONET to identify cases it has never seen before. Following thorough clinical tests, both scientists believe MOGONET could help enhance patient care in a variety of ways.