Nature: AI artificial intelligence algorithm discovers 165 new oncogenes

Nature: AI artificial intelligence algorithm discovers 165 new oncogenes.  In cancer, cancer cells lose control. They multiply rapidly and transfer to other tissues in the body, destroying organs, and ultimately causing people to lose their lives.

This uncontrolled growth is usually caused by the accumulation of DNA mutations in oncogenes, such as mutations in these genes that control cell development. But some cancers have very few mutated genes, which means that in these cases, there are other reasons for cancer to appear.

On April 12, 2021, the Annalisa Marsico team of the Max Planck Institute for Molecular Genetics in Germany published a titled Integration of multiomics data with graph convolutional networks to identify new cancer genes and their associated in the journal Nature Machine Intelligence Research papers on molecular mechanisms.

A new algorithm can predict which genes will cause cancer, even if their DNA sequence has not changed.

The research team combined a variety of data and used artificial intelligence (AI) to analyze it and successfully identified 165 new oncogenes. It opens up new prospects for targeted cancer therapy and the development of biomarkers in personalized medicine.

The research team developed a new algorithm called “EMOGI” based on artificial intelligence (AI) and successfully identified 165 previously unknown oncogenes. These genes do not have to be mutated to cause cancer. Some of them are caused by expression disorders. . All these newly discovered oncogenes interact closely with known well-known oncogenes. And cell experiments confirmed that they are vital to the survival of tumor cells.

The research team developed a machine learning algorithm based on Graph Convolutional Networks (GCN)-EMOGI (Explainable Multiomics Graph Integration). The algorithm integrates tens of thousands of data sets generated from patient samples. These data sets include mutated DNA sequence data, DNA methylation, individual gene activity, and protein interaction information in cellular pathways. In this data, deep learning algorithms can detect patterns and molecular principles that lead to the development of cancer.

Different from traditional cancer treatment methods (such as chemotherapy and radiotherapy), personalized treatment methods can precisely adjust the treatment methods and drugs according to the type of cancer. The purpose is to choose the best treatment for each patient, that is, the most effective treatment with the least side effects. In addition, cancer can be identified at an early stage based on the molecular characteristics of the patient. Only by knowing the causes of diseases can they be effectively eliminated or corrected. This is why we need to determine as much as possible the mechanism that induces cancer.

So far, most cancer research has focused on mutations in gene sequences. In fact, studies in recent years have shown that epigenetic or gene expression disorders may also cause cancer.

This is why the research team integrated gene mutation sequence data with DNA methylation, gene expression activity, protein interaction and other information. First of all, the research team confirmed that mutations or multiplication of genome fragments are indeed the main driving force of cancer. . Then, the research team further identified candidate genes that are not directly related to cancer driver genes.

The interaction of proteins and genes can be mapped into a mathematical network, that is, a graph. Think of it as a railway network. Each site corresponds to a protein or gene, and each interaction between them is like a train route.

With the help of artificial intelligence algorithms, the research team analyzed thousands of different interaction network diagrams of 16 different cancer types.

Through this algorithm, the research team found genes that are not mutated in cancer, but they can regulate energy supply and are therefore closely related to the development of cancer. These genes are affected by methylation and other ways, and their expression is unregulated, thus affecting the development of cancer.

These genes are potential cancer treatment targets, but because they are so hidden, they can only be discovered with the help of bioinformatics and the latest artificial intelligence algorithms.

The research team also discovered that many interesting details are hidden in the data. The pattern we see depends on the specific cancer and tissue, and the research team believes this is evidence that tumors are triggered by different molecular mechanisms in different organs.

Finally, the research team emphasized that the EMOGI algorithm can not only be used for cancer research. In theory, it can be used to integrate various biological data sets and find patterns from them, so it can be used for complex diseases where other genes play an important role, such as metabolic diseases such as diabetes.

(source:internet, reference only)

Disclaimer of medicaltrend.org