Graph Neural Learning for Antimicrobial Resensitization
Dismantling bacterial defense machinery across global pathogen genomes to resensitize multi-drug resistant strains to off-patent antibiotics.
The global antimicrobial resistance crisis cannot be solved by new antibiotic discovery alone. RAMP-GNN maps the complex genetic and chemical interactome across extensive global bacterial isolates, identifying non-antibiotic adjuvants—such as Gallium Nitrate, Deferasirox, and Pentamidine—that breach bacterial defense mechanisms and resensitize multi-drug resistant pathogens to existing frontline therapeutics.
A Crisis Outpacing De Novo Chemistry
Antimicrobial resistance is directly responsible for over 1.27 million deaths annually, projected to escalate to 10 million deaths per year by 2050. Bringing a new antibiotic class from chemical discovery to clinical approval takes 10 to 15 years and over \$1.5 billion—while bacteria evolve resistance within months of clinical introduction.
The Antibiotic Discovery Void
No novel chemical classes of antibiotics for Gram-negative pathogens have reached market deployment in over three decades. Big pharma has largely abandoned antibiotic pipelines due to economic market failures, leaving clinicians with dwindling options.
ICU Sepsis Mortality Windows
In septic shock, patient mortality increases by 7.6% for every hour of delayed appropriate antimicrobial therapy. When pathogens harbor multi-drug resistance, empirical frontline treatments fail, resulting in preventable fatal septicemia.
Adjuvant Resensitization
Instead of synthesizing novel antibiotics, non-antibiotic adjuvants dismantle bacterial protection mechanisms—iron scavenging, efflux pumps, and outer membrane barriers—restoring complete bactericidal activity to existing, off-patent antibiotics.
Unmonitored International Outbreaks
Pan-drug resistant hospital strains cross borders within days via medical tourism and international travel. Without automated genomic resistome prediction, local outbreaks transform into global healthcare crises before detection.
Overcoming Graph Over-Smoothing in Pathogen Biology
Traditional GNNs fail in biological networks because biologically dissimilar nodes (e.g. resistant mutations and adjuvant molecules) interact heterophilically. RAMP-GNN implements dedicated multi-relational aggregation to preserve critical sub-network topology.
Heterophilic Graph Convolutions
Decouples ego-embeddings from neighborhood aggregations, preventing graph over-smoothing and accurately capturing the sharp, non-linear interaction boundaries between bacterial defense genes and chemical entities.
Multi-Relational Knowledge Tensor
Integrates diverse edge classes simultaneously: pathogen-to-gene mutation, gene-to-resistance-mechanism, drug-to-target-affinity, and compound-to-compound synergistic inhibition.
Mechanism-Targeted Adjuvant Mining
Screens non-antibiotic pharmacopeias for specific phenotypic disruptions—such as iron metabolism disruption (Gallium Nitrate) and iron chelation (Deferasirox)—that selectively disarm Gram-negative defense machinery.
Sub-50ms Clinical Workstation Runtime
Optimized sparse tensor kernels enable whole-resistome synergy scoring in under 50 milliseconds per isolate, making the framework deployable directly on hospital clinical pathology workstations.
Empirical Results Across 104 Countries
Benchmarked on global public health surveillance data encompassing diverse geography, hospital settings, and critical resistant strains including Carbapenem-Resistant Enterobacteriaceae (CRE) and Acinetobacter baumannii.
Worldwide isolate surveillance representing diverse unique clinical biosamples across diverse clinical geographies.
Robust discriminative accuracy across complex multi-drug resistance phenotypes without overfitting.
Precision-recall performance maintaining exceptional reliability even in extreme class-imbalanced datasets.
Confirmed non-antibiotic adjuvants (Gallium Nitrate, Deferasirox, Pentamidine) reversing established resistance.
Actionable Responses to the AMR Emergency
Translating computational graph discoveries into rapid clinical and public health interventions.
Emergency Combination Rescue Therapy
When an ICU patient exhibits zero in vitro response to colistin or carbapenems, RAMP-GNN predicts immediate synergistic adjuvant co-therapies to break resistance and save the patient from septic collapse.
Rapid Hospital Genomic Resistome Profiling
Integrates directly with nanopore or next-generation sequencing pipelines in diagnostic pathology labs, delivering full phenotypic susceptibility predictions from raw bacterial DNA in under two hours.
Global Epidemiological Outbreak Tracking
Provides public health agencies with automated surveillance to detect emerging resistance clusters and geographic transmission vectors before localized outbreaks become uncontrollable global epidemics.
High-Throughput Repurposing Library Screening
Screens millions of approved non-antibiotic small molecules in silico, identifying potent adjuvants that disrupt bacterial efflux mechanisms and resensitize WHO critical-priority pathogens.
Clear Path to Clinical Deployment
A four-phase scientific execution plan moving from retrospective graph representation learning to clinical combination therapy.
Global Graph Representation Learning
Heterogeneous graph neural modeling across diverse global pathogen isolates and identification of top non-antibiotic adjuvant candidates.
In Vitro Adjuvant Synergism Assays
Checkerboard microdilution and fractional inhibitory concentration (FIC) assays validating Gallium Nitrate and Deferasirox combinations.
Automated ICU Pathogen Profiling
Integration into hospital sequencing workflows to provide real-time adjuvant guidance for multidrug-resistant sepsis patients.
Multicenter Combination Clinical Trials
Regulatory prospective clinical trials for repurposed adjuvant-antibiotic combination therapies in acute multidrug-resistant infections.
Questions & Scientific Inquiries
Why focus on non-antibiotic adjuvants rather than creating new antibiotics? +
Developing de novo antibiotics takes over a decade and costs upwards of \$1.5 billion, with bacteria rapidly developing resistance. Many non-antibiotic compounds (such as Gallium Nitrate and Deferasirox) are already clinically approved for other indications, possess known human safety profiles, and disrupt bacterial survival mechanics in ways that immediately restore the killing power of standard antibiotics.
How does Gallium Nitrate resensitize resistant bacteria? +
Gallium (Ga3+) acts as a molecular Trojan horse. Because its ionic radius matches Iron (Fe3+), bacteria mistakenly import it into their metabolic pathways. However, unlike iron, gallium cannot undergo reduction-oxidation reactions, paralyzing bacterial DNA synthesis, disrupting respiration, and collapsing membrane permeability.
Can RAMP-GNN be deployed locally in hospital laboratories? +
Yes. RAMP-GNN runs offline on standard clinical hardware or laboratory workstations without sending sensitive patient microbial data across public cloud networks, making it compliant with hospital data governance and HIPAA/GDPR standards.
Collaborate With Us on Antimicrobial Intelligence
We welcome partnerships with microbiology laboratories, infectious disease research institutions, and health agencies to validate resistance-breaking adjuvants.