Multiclass Classification of Drug Resistance in MTB clinical isolates (version 2), abbreviated as MCDR-MTB-V2, is a webserver that uses Multi-Layer Perceptron (MLP) model to predict class of drug resistance in Mycobacterium tuberculosis (MTB) isolates from variant calling format (VCF) files of whole genome sequencing (WGS) data. It can classify the MTB isolates into four classes- 1)Extensively drug resistance (XDR), 2)Pre-extensively drug resistance (Pre-XDR), 3)Multidrug resistance (MDR), and 4)Drug susceptible (Susc).
The VCF files contain the sequence variation information such as SNPs, InDels and other type of mutations obtained from WGS data analysis. The variant calling was performed on 32 targeted regions associated with anti-tuberculosis drugs. The model achieved an overall accuracy of 91.78%. The predicted class is determined from the prediction scores and the prediction confidence is measured using reliability index (RI).
This webserver can perform multiclass classification of single MTB isolate from a VCF file. In case users have large number of MTB WGS data (either in FASTQ or VCF), users can use the standalone version of MCDR-MTB that is available here to predict the drug resistance class.

Upload input .vcf file. (Maximum file size = 18MB)

Single-sample VCF file
An example of VCF file

The VCF file input for the MCDR-MTB webserver can be generated using the scripts provided in GitHub repository.