Drug Repurposing: Deep Learning Approach Through Data Integration Pubmed Case Study Metformin

Background: Biomedical text mining has become very crucial because of increasing number of related documents. Along-side with the development in machine learning, obtaining good knowledge from literature has become popular in academia, and deep learning has enhanced the progress in efficacy of text mining models in medical sciences. However, deep learn-ing models need a lot of training data, and application of deep learning to biomedical text mining fails frequently because of not having training data in medical sciences. Current discover-ies on training contextualized language representation models on text corpora clarifies on the likelihood of benefitting from a lot of unannotated medical text corpora. Development in welfare and diet have created noticeable enhancement of life expectancy around the world. Our knowledge about scientific ground these morbidities has been quickly improved, however, successful new medication has not been created or discovered, yet. From that, alternative drug development approaches like the reposition of already known drugs for treatment of other diseases, are now being explored. This would save a lot of time and money because the pharmacokinetics, pharmacodynamics and safety profiles of these drugs have been already character-ized, In fact drug reprofiling effectively would get around pre-clinical studies, required for newly invented medication. Met-formin is one of medication on which there has been ongoing research on its new application Previous clinical evidence show that metformin has been normally prescribed for diabetes treat-ment. This study aims at investigation into the possible effects of this medication on various diseases, published in PubMed. In other words, we studied the reported results, available in medi-cal literatures for potential of metformin to prevent or treat dif-ferent kinds of disorders.Materials and Methods: We look for relevant publication in Pub Med through using metformin as key word. This search cover studies have been done between 1994 a d 2019, which was written in English language. In this direction, we applied NER BioBERT method. Bidirectional Encoder Representations is transformation of Text Mining and specific area of language representation model, useful in wide range of texts. According to this structure, BioBERT actively relocate the information from a lot of biomedical texts to biomedical text mining models by a few modifications in task specific structure. While BERT illustrate excellent function with former models, BioBERT noticeably overtake them on entity recognition and clustering with regard to metformin impact on individuals’ health. Differ-ent disease and the trend of metformin impact is in publication during several years, based on drug effect on various disease Results: There has been some reported investigation into the association between metformin and the results of remedy in various diseases. In addition to these preclinical report, reliable biological pathways have been known which explain the mo-lecular mechanism of metformin and addressed in our study. However, the important answer to this question that the level of metformin efficacy against non-diabetic disorders. Up until there is not clear answer for that in clinical trial, the role of metformin on treatment or prevention of disease remains hy-pothetical.Conclusion: Up until there is not clear answer for that in clini-cal trial, the role of metformin on treatment or prevention of disease remains hypothetical that in this paper has been investigated all of their dimensions through clustering and NER in PubMed.

Keywords:

Drug Repurposing , Deep Learning , Metformin , Pub-med , Biobert