Deep learning predicts early cancer onset
2021
Max-Planck-Institut für Molekulare Genetik, Berlin, Germany
The EMOGI (Explainable Multi-Omics Graph Integration) is a machine learning method to identify cancer genes. Based on human patient data, the deep learning algorithm combines data for mutations, copy number changes, DNA methylation and gene expression with the protein-protein interaction for a more accurate prediction of early molecular signs of cancer than other methods known to date. Genetic alterations, as well as non-genetic causes, drive tumorigenesis and some non-mutated genes interact with known cancer genes. 165 novel cancer genes were proposed by this method. EMOGI may open new possibilities for therapeutic substance identification in personalized precision oncology. The method is not restricted to oncology, it can be used in other complex diseases with genetic and non-genetic biomarkers like metabolic disorders to identify disease patterns.
Integration of multiomics data with graph convolutional networks to identify new cancer genes and their associated molecular mechanisms
Annalisa Marsico
Added on: 04-26-2021
[1] www.nature.com/articles/s42256-021-00325-y[2] https://www.bionity.com/en/news/1170673/more-than-the-sum-of-mutations.html
