Welcome to the official repository for MDRIPred, a computational platform developed for predicting inhibitors against drug-tolerant Mycobacterium tuberculosis (H37Rv). The system was designed to identify compounds effective against both replicative and non-replicative phases of tuberculosis under carbon starvation conditions.

This platform supports tuberculosis drug discovery by integrating machine learning, molecular fingerprints, pharmacophore analysis, and structural feature analysis.

Web Server: http://crdd.osdd.net/oscadd/mdri/

Tuberculosis (TB), caused by Mycobacterium tuberculosis (M.tb), remains one of the deadliest infectious diseases worldwide. The emergence of multidrug-resistant (MDR) and extensively drug-resistant (XDR) strains has created an urgent need for discovering new anti-tubercular compounds.

This study focuses on identifying inhibitors effective against:

• Replicative phase M.tb
• Non-replicative (drug-tolerant) phase M.tb

The computational models were developed using high-throughput screening datasets and machine learning approaches to accelerate anti-tuberculosis drug discovery.

The datasets were obtained from PubChem BioAssay screens:

• AID-492952 (Replicative phase)
• AID-488890 (Non-replicative phase)

• Total compounds: 2135
• Inhibitors: 1355
• Non-inhibitors: 780

• Total compounds: 2135
• Inhibitors: 1206
• Non-inhibitors: 929

The compounds were processed after removing salts and ion-containing molecules.

Support Vector Machine (SVM)-based classification models were developed using multiple molecular fingerprint classes:

• PubChem fingerprints
• MACCS fingerprints
• EState fingerprints
• SubStructure fingerprints

The models predict whether a compound acts as an inhibitor or non-inhibitor against drug-tolerant M.tb.

• Best MCC: 0.45
• Best model: MACCS fingerprint model
• Accuracy: ~73%

• Best MCC: 0.28
• Best model: Hybrid fingerprint model
• Accuracy: ~64%

The study identified important physicochemical properties associated with anti-tubercular activity.

• Molecular weight
• Polar surface area (PSA)
• Hydrogen bond acceptor count
• Rotatable bond count
• logP

Preferred properties include:

• Molecular weight > 300 Da
• Hydrogen bond acceptor > 5
• Rotatable bond count > 6
• Polar surface area > 88 Ų

Preferred properties include:

• Molecular weight < 380 Da
• Rotatable bond count between 2–4

These rules may assist researchers in designing new anti-tubercular molecules.

Substructure fragment analysis identified important molecular motifs associated with inhibitory activity.

• hetero_N_nonbasic
• heterocyclic
• carboxylic_ester
• hetero_N_basic_no_H

• hetero_O
• ketone
• secondary_mixed_amine
• vinylogous_halide

The following fragments were important in both phases:

• Nitro
• Alkyne
• Enamine

SMART filtering was applied using:

• Abbott ALARM filters
• Glaxo filters
• Pfizer LINT filters

This analysis helped identify undesirable chemical fragments related to toxicity and poor drug-like properties.

Pharmacophore models were generated using first-line and second-line anti-tubercular drugs including:

• Rifampicin
• Ethambutol
• Streptomycin
• Ethionamide
• Cycloserine
• Kanamycin
• Amikacin

These pharmacophore models were used to identify structurally similar compounds within the datasets.

The study introduced a novel feature selection method called:

MCCA was used for selecting informative fingerprints and descriptors for model development.

The method showed performance comparable or superior to frequency-based feature selection methods.

The MDRIPred webserver allows users to:

• Predict inhibitors against drug-tolerant M.tb
• Analyze compounds for replicative and non-replicative activity
• Perform pharmacophore similarity analysis
• Upload molecular structures
• Draw structures using JME editor

Users can submit molecules using:

• JME molecular editor
• MOL/MOL2 file format
• File upload option

The webserver was developed using:

• Linux environment
• CGI-Perl scripts
• SVMlight package
• PaDEL descriptor software
• Pharmagist software

MDRIPred can be used for:

• Anti-tuberculosis drug discovery
• Virtual screening
• Lead optimization
• QSAR studies
• Molecular property analysis
• Pharmacophore screening
• Drug resistance research

The webserver is freely available for academic and research purposes.

Web Server:
http://crdd.osdd.net/oscadd/mdri/

Bioinformatics Centre
CSIR-Institute of Microbial Technology
Chandigarh, India

Email: raghava@iiitd.ac.in

This work is distributed under the
Creative Commons Attribution License

The authors acknowledge support from:

• Council of Scientific and Industrial Research (CSIR)
• Open Source Drug Discovery (OSDD)
• Department of Biotechnology (DBT)
• Government of India

We also acknowledge the PubChem BioAssay resource and the scientific community contributing to tuberculosis drug discovery research.