Big Data Mining For Blood-based Tuberculosis Novel Diagnostic Biomarkers And Clinical Validation

Despite decades of developments in immunization and drug therapy,tuberculosis(TB)caused by Mycobacterium tuberculosis(MTB)remains a global public health threat with estimated one third of the world's population latently infected.There are around 9 million newly diagnosed cases of tuberculosis and 1.5 million deaths each year.China is a country with high burden of tuberculosis,and tuberculosis is the key target of prevention and control in China.Early detection and treatment is vital to prevent and control tuberculosis.The diagnosis of tuberculosis mainly depends on the traditional MTB acid-fast staining method and MTB culture.However,the sensitivity of the former is poor,and the latter is time-consuming,which could not meet the needs of rapid diagnosis of tuberculosis today.Immunodiagnosis,molecular diagnosis and imaging diagnosis are convenient and fast,but the sensitivity of tuberculosis diagnosis needs to be improved.Despite huge progress in treatment and diagnosis,timely precise diagnosis of tuberculosis remains difficult.In order to achieve the goal of ending tuberculosis,we need to have a deeper understanding of tuberculosis.The cross-integration of data science,information technology and biotechnology and the generation of big data provide us with a new way to fight tuberculosis.We can understand the pathogenesis of tuberculosis,the relationship between MTB and human beings and the emergence of drug-resistant MTB from different perspective,and explore new diagnostic and therapeutic methods for tuberculosis.The new ideal diagnostic method should be those based on non-sputum samples and high sensitive for HIV co-infected patients.Blood-based TB diagnostic development meets WHO's target product characteristics.In contrast,blood-based biomarkers have great advantages.Blood collection is relatively easy and is a useful source of biomarker measurement throughout the treatment.Blood-based inflammation and infection markers are ready to monitor and for point of care tests.There is a large amount of data as to the host response to M.tuberculosis infection and stored in publicly accessible databases such as the National Institutes of Health Gene Expression Omnibus(NIH GEO).These data have not yet been fully utilized,and the rich value behind them needs to be further excavated,which is a new way to find the markers of tuberculosis diagnosis.In this study,we selected six whole blood expression datasets from GEO database.By using the GSE19491 dataset,we calculated the differentially expressed genes in pulmonary active TB versus latent tuberculosis and health controls,gene co-expression network was constructed and genes were assigned to expression pattern to explore key genes in tuberculosis via comparative analysis.We found four genes can be novel tuberculosis diagnostic biomarkers,this was validated by both datasets test(AUC 0.86,Specificity 81%,Sensitivity 86%)and clinical specimens.These four genes are UBE2L6(Ubiquitin/ISG15-conjugating enzyme E2 L6),BATF2(Basic leucine zipper transcriptional factor ATF-like),SERPING1(Plasma protease C1 inhibitor)and VAMP5(Vesicle-associated membrane protein 5),the functions of these genes are associated with ubiquitination,immunocellular differentiation,complement activation,and vesicle trafficking respectively.The results of our study have been applied for a patent(PCT/CN2019/080563).