One of our most pressing research areas at SMCS-Psi focuses on addressing the growing challenge of antibiotic resistance, particularly in difficult-to-control plant and environmental pathogens such as Ralstonia solanacearum. This soil-borne bacterium is known for causing bacterial wilt in a wide range of crops, leading to significant agricultural losses worldwide. However, the broader implications of our research extend far beyond this single pathogen.
Using our in-house machine learning platform, we have successfully screened over 10,000 natural and semi-synthetic compounds to identify potential antibacterial agents with high predicted efficacy. These molecules undergo a rigorous multi-step validation process that includes:
Molecular Docking – Predicting the most favourable binding orientation of candidate compounds to essential bacterial proteins.
LigPlot+ Interaction Visualization – Generating intuitive 2D diagrams that reveal key hydrogen bonds and hydrophobic interactions between the ligand and protein active site.
Molecular Dynamics (MD) Simulations – Simulating real-time atomic-level interactions to evaluate the stability, flexibility, and binding free energy of the protein-ligand complex.
In-vitro Biological Assays – Experimentally testing the top compounds to confirm their antimicrobial activity under laboratory conditions.
Future Vision: A Universal Inhibitor Against Bacterial Growth
Looking ahead, this computational-experimental hybrid pipeline is being extended to develop a broad-spectrum antibacterial compound—one capable of halting bacterial growth across multiple species by targeting a highly conserved and essential bacterial pathway.
Our scientific rationale involves identifying evolutionarily conserved bacterial biomolecules, such as:
DNA gyrase/topoisomerases
Bacterial specific ribosomal proteins
MurA (peptidoglycan biosynthesis)
Fatty acid biosynthesis enzymes
Two-component regulatory systems
These proteins are indispensable for bacterial survival and present minimal variation across Gram-positive and Gram-negative strains. By using ML-based comparative proteomics and ligand-based activity prediction, we aim to pinpoint lead molecules that bind to these targets with:
High selectivity
Minimal off-target toxicity
Resistance-resilient binding mechanisms
Machine Learning as the Driver
Our ML framework is continuously being trained on curated datasets of:
Known antibiotic–protein interaction profiles
Resistance gene patterns (from databases like CARD and ResFinder)
Multi-target inhibitors
Physicochemical & pharmacokinetic properties
This enables prediction of both antibacterial activity and resistance risk, allowing us to prioritise compounds that are both potent and durable.
Additionally, explainable AI techniques (e.g., SHAP, integrated gradients) help us identify which molecular substructures contribute most to predicted activity—guiding rational design and chemical optimisation.
Broader Impact
This initiative doesn’t just aim to combat Ralstonia infections in plants but also contributes toward:
Developing agriculturally safe antimicrobials
Designing next-generation antibiotics for clinical pathogens
Building a generalisable ML framework for drug discovery in infectious diseases
In the long run, our goal is to create a first-in-class compound that can suppress or eliminate bacterial growth across diverse taxa by acting on a universal, irreplaceable bacterial function—a major milestone in global health and food security.
