Determining the effectiveness of drugs on a mutating strain of tuberculosis bacteria by using tuberculosis datasets under a secure cloud-based data management

Abstract

Drug-resistant tuberculosis (TB) poses an alarmingly high mortality risk, because the bacterial genes undergo complex mutations in response to anti-tuberculosis drugs. For studying gene regions, where mutations occurred in response to a specific drug, association mining, a machine learning technique, was first applied on established datasets to group the individual gene-drug pairs and their corresponding reported mutations. Secondly, a simple, yet novel, effectiveness factor is proposed which evaluated the gene-drug pair by incorporating both frequency and distribution of the mutations in a specific bacterial gene. This study applied the proposed factor to determine the effectiveness of anti-tuberculosis drugs for a specific TB strain, H37Rv. The proposed method, however, can also be applied to other strains, subject to the availability of datasets. The datasets as well as the information generated from the proposed study can be readily stored in a secure cloud storage system, either for public or private access and retrieval.