Credit: Pixabay/CC0 public domain
Transcriptomics, proteomics and metabolomics are all possible thanks to the rapid advancements in biomedical technology. How can we best take advantage of the interplays and complementar information found in omics data?
Researchers from Regenstrief Institute, Indiana, Purdue, and Tulane Universities developed and tested MOGONET to fully harness the advances in omics technology and gain a better understanding of biological processes that underlie human diseases. The integration of data from different omics gives a better understanding of the biological processes that underlie human diseases. MOGONET is open-source, which means it's free to all researchers.
The scientists published a study in Nature Communications demonstrating that MOGONET (short for Multi-Omics graph cOnvolutional Networks) outperforms other supervised multi-omics integrative analyses approaches to different biomedical classification apps using mRNA expression, DNA methylation, and microRNA expression.
They also discovered that MOGONET can identify important biomarkers and omics signatures from different types of omics data.
"With MOGONET [our new AI] tool, machine learning based upon a neural network is used to capture complex biological process relations. The study was led by Kun Huang, a Regenstrief Institute Research Scientist. "We have improved our understanding of omics and are also learning more about subtypes of disease that biomarkers can help us distinguish," he said. The ultimate goal of the study is to improve disease prognosis, and predict disease outcomes. He is a bioinformatician and credits the success of MOGONET's research group, which was made up of data scientists, bioinformaticians and computer scientists. He is the director of data sciences, informatics and information for the Indiana University Precision Health Initiative.
MOGONET was tested on data related to Alzheimer's, gliomas and kidney cancer, as well as healthy patient data. MOGONET outperformed all existing supervised multi-omics methods of integration, according to the researchers.
According to Jie Zhang (a bioinformatician and co-author of the study), MOGONET can learn and integrate intuitive recognition. MOGONET could also predict new types of cancer, the progression of disease and grade. It can distinguish between normal brain activity and Alzheimer's disease.
Drs. Drs. Both scientists agree that MOGONET can be used to improve patient care in many areas after undergoing rigorous clinical trials.
Drs. Drs. Zhang and Huang are the authors of MOGONET, which integrates multi-omics data using graph Convolutional Networks allowing patient classification, and biomarker identification. Dr. Wang was a member of Dr. Huang's lab. Dr. Ding is a former professor at Indiana University and an expert in machine learning.
Further research on Improved statistical methods to high-throughput data analysis of omics data
Tongxin Wang and colleagues, MOGONET integrates multiomics data using graph-convolutional networks to allow patient classification and biomarker recognition, Nature Communications (2021). Journal information: Nature Communications Tongxin Wang et al, MOGONET integrates multi-omics data using graph convolutional networks allowing patient classification and biomarker identification,(2021). DOI: 10.1038/s41467-021-23774-w
