Recent scientific breakthroughs have underscored the utility of AI in unearthing new classes of antibiotics, with ongoing research endeavors focused on developing more pinpointed treatment methodologies
According to the World Health Association, Antimicrobial Resistance (AMR) accounts for 4.95 million deaths worldwide. There is an urgent need to develop new, more effective antibiotics. This challenge can be confronted using artificial intelligence techniques.
While the application of AI in drug development is nascent, in the past two years, medications conceptualized via artificial intelligence have commenced the preliminary stages of clinical trials. The burgeoning interest in its use, especially for the development of targeted antibiotics, is also due to the potential for AMR to inflict even greater damage in the future.
By 2050, the UNEP forecasts an annual death toll of 10 million. Furthermore, there are predicted to be severe economic repercussions. Without control, antimicrobial resistance could slash the global GDP by 3400 billion dollars a year and plunge 24 million people into extreme poverty over the next decade, as asserted by the United Nations Environment Programme. Therefore, exploring new methods and solutions, including the application of artificial intelligence, becomes imperative to avert this scenario.
TAKEAWAYS
The Use of AI Against Antimicrobial Resistance
The integration of artificial intelligence in the development of new antibiotics is of immense interest as it facilitates a rapid response to the urgent need for drugs capable of combating antimicrobial resistance. Ordinarily, the creation of a bespoke drug can span up to fifteen years, while a microbe might develop resistance within just two years.
Nevertheless, the pharmaceutical industry’s investment has dwindled in recent years, as antibiotics are perceived as less profitable compared to other medical devices. To provide context: the global market for cancer drugs is projected to double, increasing from 140 billion in 2022 to 303 billion in 2031. The market value for anti-inflammatory treatments is expected to grow from 94.6 billion dollars to 149.5 billion dollars by 2031. The global antibiotics market, valued at 43 billion dollars in 2022, is anticipated to experience significant growth, but this pales in comparison [source: Precedence Research – Antibiotics Market: Global Industry Analysis, Size, Share, Growth, Trends, Regional Outlook, and Forecast 2023-2032].
Moreover, the expense of developing new compounds is a crucial consideration. Producing thousands of novel natural substances and testing their efficacy with conventional methods is virtually impractical. Hence, the role of AI becomes critical. A team of researchers led by Jon Stokes, a professor in the Department of Biochemistry and Biomedical Sciences at McMaster University, Canada, and James Collins, a professor of biological engineering at MIT in Boston, has recently pioneered a machine learning algorithm capable of simultaneously analyzing thousands of compounds to determine their efficacy against specific bacteria.
The researchers accomplished their objective: they discovered a new compound capable of eradicating Acinetobacter baumannii, a bacterium frequently encountered in hospitals, known to cause pneumonia, meningitis, and other grave infections. This bacterium has the ability to endure in hospital environments for prolonged periods, acquiring antibiotic resistance genes from its surroundings. To secure this significant breakthrough, the scientists utilized a machine learning algorithm, analyzing a repository of nearly seven thousand potential pharmaceutical compounds. The analysis was completed in a mere two hours. The algorithm was calibrated to determine whether a chemical compound could inhibit the bacterial growth. Notably, the developed antibiotic proved effective specifically against this bacterium and not others, an advantageous attribute as a more targeted antibiotic can substantially reduce the risk of rapid microbial resistance development against the medication.
For these reasons, Stokes, Collins, and their collaborating team believe they can replicate this modelling approach to identify new antibiotics for various types of drug-resistant infections.
Opportunities to Seize
AI proves instrumental in decreasing the failure rate of drugs in clinical trials. Current estimates suggest that 90% of clinical drug development fails [source: Duxin Sun et al., Science Direct]. Beyond this, artificial intelligence can aid in the early detection of bacterial infections, crucial for diagnosing healthcare-associated infections, as well as enhancing disease diagnosis and optimizing antibiotic prescriptions.
One of the AI applications pertains to interpreting antimicrobial susceptibility tests, a vital parameter for examining bacterial responses to antimicrobial agents and assessing AMR.
Nevertheless, the most poignant interest in AI usage is linked to drug development opportunities, particularly in antibiotics. The latest discovery by a team of scientists in the United States has garnered significant attention.
These scientists are credited with developing the first antibiotic identified through AI’s capabilities. They merged in silico predictions with empirical research to discover a new broad-spectrum antibiotic, allicin, an effective inhibitor of Escherichia coli growth. Approximately 2300 compounds were analyzed to predict germ growth inhibition; the resultant data trained a deep neural network to forecast the chemical’s antimicrobial properties. Allicin, structurally distinct from conventional antibiotics, exhibits bactericidal effectiveness against various pathogens.
Blending biology with artificial intelligence represents another successfully explored approach. An example comes from a team at the Max Planck Institute for Terrestrial Microbiology, Germany. They successfully integrated the two fields to develop a more efficient method for discovering and creating new antimicrobial peptides effective against a broad spectrum of bacteria. The team established a cell-free protein synthesis system for the rapid and cost-effective production of antimicrobial peptides directly from DNA templates. Specifically, the German institute’s scientists initially used generative deep learning to design thousands of these specific proteins from scratch, followed by predictive DL to narrow down to 500 candidates. This process identified several peptides demonstrating broad-spectrum activity against multi-resistant pathogenic microorganisms, showing no development of bacterial resistance.
The application of machine learning to new drug development, starting with antibiotic treatments for tuberculosis (with 230,000 annual deaths globally from multidrug-resistant tuberculosis), is the aim of the NextAID (Neuro-explicit AI for Drug Discovery) project at Saarland University, Germany. Machine learning models will be employed to expedite compound screening, conduct in silico predictions on the pharmacokinetic properties and toxicity of prospective drugs. The goal is to utilize AI to generate active ingredients for laboratory testing to evaluate their effectiveness.
Glimpses of futures
The deployment of AI in combatting antimicrobial resistance heralds significant implications for the development of suitable therapies. Presently, substantial efforts are being made to apply artificial intelligence techniques to the antibiotics of the future, paving new avenues in research. The adoption of machine learning and deep learning models is poised to deepen our understanding of antimicrobial resistance and foster the creation of appropriate solutions. It is even feasible to estimate the quantities of antimicrobials present in water resources [source: Sara Iftikhar et al., National Library of Medicine].
Moreover, the influence of generative artificial intelligence warrants consideration. As per the McKinsey Global Institute, Generative AI-based technology could yield an annual economic value of 60 to 110 billion dollars for the pharmaceutical and medical products industry. However, we are at the outset, and numerous complexities must be navigated in applying AI to antimicrobial resistance: a great deal more remains to be discovered before extensive clinical and healthcare applications can be contemplated.
It is advisable to initiate the development of new research avenues now. For instance, a team from Texas A&M University has developed a new family of polymers capable of eliminating bacteria without inducing antibiotic resistance, marking an important stride in the battle against antimicrobial resistance.
What implications could the use of artificial intelligence in studying new antibiotics and mitigating antimicrobial resistance have? What potential futures could be envisioned?
Let us consider the prospective impacts, weighing the opportunities, benefits, and risks, according to the STEPS matrix.
S – SOCIAL: AI’s application can lead to the creation of personalized medicines, making healthcare systems more targeted and enhancing treatment efficacy while reducing side effects. However, it is vital to ensure that such tailored healthcare is universally accessible, thereby preventing social inequalities and guaranteeing fair access to research findings and new medications, especially in poorer nations disproportionately affected by AMR. Though a worldwide issue, antimicrobial resistance is intimately connected with poverty, inadequate sanitation, poor hygiene, and pollution, as the UNEP underscores in the “Bracing for Superbugs” report. Alongside targeted drugs, there is also a necessity to monitor antibiotic overuse, a prevalent issue especially in developing countries [source: Antimicrobial Resistance: A Growing Serious Threat for Global Public Health].
T – TECHNOLOGY: Technologically, AI enables the analysis of vast data sets more swiftly and efficiently than traditional methods. This could hasten the identification of potential antibiotic compounds, keeping pace with evolving resistant pathogens. Such advancements are making precision medicine increasingly viable: AI’s role here is crucial. However, there are several challenges to be overcome in the future for the effective application of artificial intelligence techniques. Take, for instance, the mechanisms of antibiotics and drugs, which remain not entirely comprehended, particularly for newly emerging diseases [source: Tabish Ali et al.]. Going forward, there is a need to ensure that medical scientific research can rely extensively on AI and its further development. While AI assists medicine, medicine in turn pushes the boundaries of technological innovation.
E – ECONOMY: AI’s application in pharmaceutical research has the potential to cut costs associated with developing new antibiotics, thus improving treatment efficacy and minimizing side effects, with significant economic and health implications. In the European Union alone, it is estimated that infections from multi-resistant bacteria cause 35,000 deaths annually and that antimicrobial resistance incurs costs of around 1.5 billion euros in healthcare expenses and productivity losses [source: European Medicines Agency].
P – POLITICAL: Alongside the opportunities, the ethical, social, and legal risks associated with AI’s use in antibiotic discovery must also be addressed. The European Parliament, in this context, published a study in 2022 titled “Artificial intelligence in healthcare. Applications, risks, and ethical and societal impacts”, which delineated both the benefits and critical concerns. These include issues regarding data security, algorithm transparency, decision-making responsibility, and the equitable distribution of advantages.
S – SUSTAINABILITY: AI-driven research could facilitate the development of more targeted antibiotics, reducing the need for high dosages or prolonged treatments. This approach would help to minimize the environmental impacts linked to the uncontrolled use of antibiotics. Nonetheless, it is crucial to acknowledge that the battle against antimicrobial resistance cannot be tackled in isolation from the triple planetary crisis, comprising climate change, biodiversity loss, pollution, and the mounting production of waste. All three factors share a common root, linked to unsustainable consumption and production patterns.
