An integrated data analysis and computational approach to reveal gene signatures for paediatric Tuberculosis meningitis related to non-tuberculosis patients

The global disease burden remains greatly influenced by tuberculosis. Its most critical form, Tuberculosis meningitis (TBM) affects both adults and children who face high rates of mortality and morbidity. As many as 50% of paediatric TBM cases lead to mortality, and nearly 53.9 percent of those who survive endure neurological consequences. The primary reasons for unsatisfactory TBM results are late diagnosis and the postponement of anti-tuberculosis therapy. Thus, improved case identification and prompt administration of effective treatments are crucial for controlling tuberculosis in children. The primary aim of this research is to identify new biomarkers for the early diagnosis of Tuberculosis meningitis in children by conducting RNA sequencing on data obtained from various samples. This research emphasizes the use of bioinformatics, machine learning, and diverse statistical methods, including the evaluation of fixed threshold values or p-values, which facilitate the identification of Differentially Expressed Genes (DEGs). Consequently, three new biomarkers were identified by combining the findings from RNASeq and machine learning, utilizing algorithms like multi-layer perceptron, support vector machine (SVM), and random forest (RF) to suggest the optimal model with high accuracy and performance for the identification of the new biomarkers. It was understood that the RF model surpassed the other classifiers by achieving an accuracy score of 1.0 in identifying new biomarkers of Tuberculosis meningitis in children. Then computational analysis of these genes was done with the potential drugs used for treating Tuberculosis meningitis using AutodockTools software.