Using computational methods to fight the antibiotic resistance crisis
In the relentless battle against antibiotic-resistant bacteria, scientists are fighting against increasingly dangerous opponents. Imagine a world where a simple scratch could lead to an untreatable infection, where routine surgeries become life-threatening procedures, and where once-controlled diseases become deadly again. This isn't science fiction—the World Health Organization warns we're moving toward this post-antibiotic era, with antimicrobial resistance directly causing approximately 1.27 million deaths globally each year 1 .
1.27 million deaths annually attributed to AMR
10 million estimated annual deaths by 2050
$1 trillion additional healthcare costs by 2050
47% increase in resistance rates since 2000
The traditional drug discovery pipeline has slowed to a trickle, with few new antibiotics reaching the market. But in this crisis, scientists are developing an ingenious solution: using advanced computational methods to repurpose existing FDA-approved drugs as antibacterial agents. By combining chemoinformatics, bioinformatics, and machine learning, researchers are finding hidden antibacterial potential in drugs created for completely different purposes—from anticancer therapies to antidepressants—potentially fast-tracking new weapons against superbugs 2 3 .
To understand this revolutionary approach, we need to break down the key technologies. Cheminformatics applies computational techniques to solve chemical problems, particularly in the field of drug discovery. It involves working with chemical structures, molecular descriptors, and physicochemical properties to predict how a molecule might behave biologically 4 .
Bioinformatics, on the other hand, focuses on biological data—especially genetic information. It helps researchers identify resistance genes, understand bacterial defense mechanisms, and pinpoint potential targets for new drugs 1 .
Together, these disciplines create a powerful pipeline for identifying non-antibiotic drugs that might accidentally have antibacterial properties.
Why look at existing non-antibiotic drugs? The advantages are significant:
Already established through previous clinical use
Already exist and are optimized
May be faster for new indications
Substantially lower than new drug development
The application of computational models capable of predicting new drug-target interactions is an interesting strategy to reposition already known drugs into potential antimicrobials 5 .
Perhaps the most exciting development in this field is the application of machine learning algorithms. These sophisticated computer programs can analyze massive chemical databases to identify patterns and predictions that would be impossible for humans to detect manually.
In one groundbreaking study, researchers used support vector machines and random forest methods to predict antibacterial compounds with impressive accuracy (mean accuracy >0.9 and mean AUC >0.9 for both models) . Their model screened FDA-approved drugs in the DrugBank database and identified 1,087 small-molecule drugs with potential antibacterial activity—154 of which were already known antibacterial drugs, representing 76.2% of the approved antibacterial drugs collected in the study.
Even more excitingly, they found 8 predicted antibacterial small-molecule compounds with novel structures compared with known antibacterial drugs, 5 of which are widely used in cancer treatment . This suggests entirely new structural classes might be discovered among existing drugs.
One pivotal study published in Assay and Drug Development Technologies demonstrated how computational predictions could be translated into laboratory confirmation 2 . The research team employed a sophisticated multi-step approach:
Using ROCS (Rapid Overlay of Chemical Structures) software, they screened 1,991 FDA-approved drugs against 17 known antibiotics, looking for three-dimensional shape similarity that might indicate similar biological activity.
The top 34 candidates were then tested in the lab using disk diffusion tests against two bacteria: Staphylococcus aureus (Gram-positive) and Escherichia coli (Gram-negative).
Finally, for those compounds showing antibacterial activity, researchers used molecular docking with HYBRID software to predict how these drugs might bind to bacterial protein targets.
The results were impressive—10 of the 34 tested drugs showed measurable antibacterial activity. Two of these (drotaverine and metoclopramide) had no previously reported antibacterial effects 2 .
| Drug Name | Primary Use | Activity Against S. aureus | Activity Against E. coli |
|---|---|---|---|
| Diclofenac | Anti-inflammatory | Yes | Weak |
| Drotaverine | Antispasmodic | Yes | No |
| Metoclopramide | Anti-emetic | Yes | No |
| (S)-Flurbiprofen | Anti-inflammatory | Yes | Weak |
| (S)-Ibuprofen | Anti-inflammatory | Yes | Weak |
| Indomethacin | Anti-inflammatory | Yes | Weak |
The molecular docking studies provided plausible explanations for how these drugs might work against bacteria. For instance, several anti-inflammatory drugs showed potential binding to the same protein targets as their similar antibiotics 2 .
This experiment demonstrated several crucial principles:
Can successfully identify non-obvious antibacterial activity in existing drugs
Is a valid approach for drug repurposing
Are essential for verification (chemoinformatics + bioassay)
Of the 1,991 drugs that were screened, 34 had been selected and among them 10 drugs showed antibacterial activity, whereby drotaverine and metoclopramide activities were without precedent reports 2 .
This approach represents a significant acceleration in antibacterial discovery—what might have taken years through traditional methods was accomplished in a fraction of the time and cost.
Behind these computational breakthroughs lie sophisticated tools and databases that make this research possible. Here's a look at some key components of the modern antibacterial discovery toolkit:
| Tool Name | Type | Primary Function | Example Use Case |
|---|---|---|---|
| ROCS | Cheminformatics Software | 3D Shape Similarity Screening | Finding non-antibiotic drugs similar to known antibiotics 2 |
| HYBRID | Molecular Docking Software | Predicting ligand-protein binding | Confirming how repurposed drugs might bind bacterial targets 2 |
| SVM/RF Models | Machine Learning Algorithms | Predicting antibacterial activity | Screening large compound databases for likely antibacterial compounds |
| ChEMBL | Chemical Database | Bioactivity data on small molecules | Training machine learning models to recognize antibacterial compounds |
| ResFinder | Bioinformatics Tool | Identifying antibiotic resistance genes | Understanding resistance mechanisms in target bacteria 1 |
While computational predictions are powerful, the ultimate test happens in the laboratory and clinic. Future directions in this field include:
Researchers are increasingly using multiple computational methods together—for instance, combining machine learning with molecular docking—to improve prediction accuracy 5 .
Some repurposed drugs may work best in combination with existing antibiotics. Computational models can help identify promising combinations 5 .
Cheminformatics approaches are also being applied to characterize natural antimicrobial products, which represent a rich source of potential antibacterial compounds 4 .
As computational predictions become more accurate, we'll likely see more rapid clinical testing of repurposed drugs for antibacterial applications.
| Drug Class | Example Compounds | Potential Antibacterial Targets |
|---|---|---|
| Anti-inflammatories | Diclofenac, Ibuprofen, Indomethacin | Bacterial enzyme inhibition 2 |
| Anticancer Drugs | 5-fluorouracil, Mercaptopurine | Nucleotide synthesis pathways |
| Antidepressants | Sertraline, Fluoxetine | Bacterial membrane disruption 5 |
| Antihistamines | Diphenhydramine, Chlorpheniramine | Unknown mechanisms 3 |
The innovative integration of chemoinformatics, bioinformatics, and laboratory validation represents a paradigm shift in how we approach antibiotic discovery. By looking at existing FDA-approved drugs through a computational lens, scientists are finding hidden antibacterial potential that went unnoticed for decades.
This approach doesn't replace traditional antibiotic development but rather complements it, offering a faster, more cost-effective pathway to identify promising candidates. As computational power grows and algorithms become more sophisticated, we can expect even more accurate predictions and unexpected discoveries.
These results prove the ability of computational mathematical prediction models to predict molecules with potential antimicrobial capacity and/or possible new pharmacological targets of interest in the design of new antibiotics and in the better understanding of antimicrobial resistance 3 .
In the relentless arms race against antibiotic-resistant bacteria, these computational methods provide something essential: hope. Hope that we can outsmart evolving pathogens, hope that we can extend the usefulness of existing drugs, and hope that we can avert the nightmare of a post-antibiotic era.
The future of antibiotic discovery may well be digital.
References will be added here in the final publication.