DGIdb 2.0, an amazing application of biomedical text mining

Nowadays, the volume of texts produced at any moment is increasing. This large volume of text, which is usually unstructured, can be processed and understood by the computer. Text mining is the task of extracting meaningful information from the text, which has attracted the attention of many researchers in recent years. Texts are generated due to infinite quantities in different forms such as social networks, patients ‘ cases, health insurance information, news, and so on. Because of the growing importance of these large data, valuable information can be extracted. The understanding and processing of this wide range of text for humans are almost impossible, so the computer is needed. Text mining has been mainly studied by researchers in the field of medicine and biology, especially in recent years, so they are tied to each other as the biomedical text mining field. In this post, I want to share a fantastic application of biomedical text mining which helped many medical experts, DGIdb2: mining clinically relevant drug-gene interactions.

I read this paper last week, and It was so amazing that I decided to share it with you. Considering the full range of medical resources in the field of gene-drug, study and obtain information for students and physicians in this field, is tough and time-consuming. The DGIdb Gene-Drug Interaction Database is a web resource that displays drug-gene interactions. It provides a graphical user interface to search for drug-gene information.

DGIdb, with extensive efforts, reflects the combined information of twenty-seven sources. In DGIdb 2.0, significant updates have been made to increase the content and increase its usefulness as a source for medicinal purposes. Specifically, three new sources of drug-gene interactions have been added, with two focusing specifically on interactions with clinical trials. That’s more than double the drug-gene interactions of the previous version. The total number of genes also increased by 5%. Most importantly, most of the unlimited and accessible resources available in DGIdb resources are now automatically updated weekly and provide the most information for these resources. Finally, a new perspective and a new interface have been developed to explore the possibility of interacting with drug identifiers to complement existing gene-based functions.

With this update, DGIdb represents a comprehensive and user-friendly tool for extracting a drug-converting genome for precision medical applications. Problems have been fixed to some extent in the initial version of DGIdb, but the following issues have caused this release:

1. Gene-gene search problem
2. Some medications were not a problem due to the lack of initial references
3. The problem of constantly updating used resources and lagging behind DGIdb

In this article, taking into account the gene-drug information available in medical authorities, extracting and coding the references and should keep in mind that this information should be continuously updated as the program continues to change. They have also predicted an increase in the number of resources to increase accuracy and accountability.

Well, as this article uses medical preparations to fill the database, the idea of using deep neural networks as an alternative comes to mind. In fact, if we could use all of the unstructured medical texts and learn the words ourselves, instead of using this information that is naturally restricted, we could have a much higher and much more accurate model.