Antibiotic Resistance Prediction: Genomics, AI, and Digital Twins

Why AMR Surveillance Is Looking Backward
Can Bacterial Evolution Be Predicted?
Building a Predictive AMR System
What Would a Superbug Digital Twin Look Like?
Why Predicting Resistance Remains Difficult
From Detecting Superbugs to Anticipating Them
References and Further Reading


Advances in bacterial genomics, experimental evolution, and computational modeling are enabling researchers to predict potential antibiotic resistance pathways. Digital twins could integrate these approaches to anticipate resistance evolution before it becomes a clinical threat.

Woman, scientist and laptop screen with microscope in laboratory, microbiology and research study.Image credit: PeopleImages/Shutterstock.com

Why AMR Surveillance Is Looking Backward

Antibiotic resistance continues to threaten the effectiveness of treatments against bacterial infections. According to the World Health Organization (WHO), approximately one in six laboratory-confirmed bacterial infections worldwide covered by its 2025 surveillance report was resistant to antibiotics in 2023.1

Current antimicrobial resistance (AMR) surveillance relies heavily on laboratory testing of clinical samples to identify bacterial pathogens and assess their susceptibility to antibiotics. Systems such as the World Health Organization's Global Antimicrobial Resistance and Use Surveillance System (GLASS) help standardize this process while incorporating clinical and epidemiological information to provide a broader picture of resistance patterns and trends. GLASS also monitors antimicrobial consumption and supports the early detection and risk assessment of emerging resistance.2

Antimicrobial Resistance - Act Now: Protect Our Present, Secure Our Future

Video credit: who/youtube.com

While these systems can identify emerging threats, they largely depend on resistance that has already developed and been detected. But what if researchers could anticipate how bacteria might develop resistance before it becomes a clinical problem?

This article explores how advances in bacterial genomics, experimental evolution, and computational modeling could support resistance forecasting, and how these approaches might contribute to the development of AMR digital twins.

Can Bacterial Evolution Be Predicted?

Although genetic mutations arise without necessarily providing an adaptive benefit, the evolutionary pathways bacteria follow are shaped by biological and environmental conditions. Under antibiotic exposure, mutations that improve bacterial survival may be favored, depending on antibiotic concentration, nutrient availability, and the bacterium's existing genetic makeup. Resistance can also come at a cost: some mutations protect bacteria against antibiotics but slow their growth, a trade-off known as a fitness cost. The effect of a mutation can also depend on other mutations already present, an interaction known as epistasis.3,4,9

Researchers investigate these relationships through experimental evolution, in which bacterial populations are grown under controlled conditions to observe how resistance develops over successive generations. By comparing bacterial fitness, how successfully bacteria survive and reproduce under particular conditions, researchers can identify which adaptations provide an evolutionary advantage.4

A 2021 study demonstrated how metabolic fitness models could help predict antibiotic resistance evolution. By examining how resistance mutations affect bacterial growth, the model predicted changes in resistance levels and which resistance mechanisms would be favored under different antibiotic concentrations and nutrient conditions. These predictions were supported by experimental growth measurements and genomic data from E. coli populations, demonstrating that aspects of resistance evolution can be predicted under defined experimental conditions. However, extending these findings to a broader range of antibiotics and biological environments remains an important research challenge.3,5

These predictions remain probabilistic rather than deterministic. Researchers aim to identify evolutionary pathways that are more likely under particular conditions, not predict the exact mutations an individual bacterium will acquire. Different evolutionary pathways can also lead to similar resistance outcomes, meaning that predicting an outcome does not necessarily require predicting every intermediate evolutionary step.4

Download your PDF copy by clicking here.

Building a Predictive AMR System

Developing a system capable of forecasting antibiotic resistance evolution would involve combining different sources of biological data and computational approaches. Genomic surveillance provides large collections of bacterial genomes and information about known resistance genes and mutations. Artificial intelligence (AI) and machine learning can analyze these data to identify relationships between bacterial genetic characteristics and observed resistance, helping predict whether an existing isolate is likely to be resistant. This type of susceptibility prediction is distinct from forecasting how resistance might evolve in the future.3,6,10

However, identifying existing resistance patterns is only part of the challenge. Researchers must also account for how bacterial populations respond to antibiotic exposure and how resistance affects their ability to survive and reproduce. Evolutionary and mathematical models help explore these dynamics by simulating processes such as natural selection, fitness trade-offs, population changes, and possible resistance trajectories. Mechanistic and molecular models add another layer of understanding by examining how particular mutations affect bacterial proteins or cellular processes, potentially altering antibiotic susceptibility.3,6

But how well do these computational predictions reflect what actually happens as bacteria evolve? Experimental evolution provides a way to investigate this question. By growing bacterial populations under controlled antibiotic conditions, researchers can compare the resistance mechanisms that emerge with those predicted by computational models.5

Integrating these complementary approaches could help researchers move beyond identifying existing resistance toward forecasting which evolutionary pathways are more likely to be favored under future conditions. Still, such forecasts would need to communicate their uncertainty and be evaluated against independent observations before being considered reliable.4,6,10

A medical professional uses a transparent digital interface to analyze a virtual human body surrounded by holographic bacteria and viruses, symbolizing AI in healthcare and futuristic diagnostics.Image credit: frank60/Shutterstock.com

What Would a Superbug Digital Twin Look Like?

One possible extension of predictive AMR modeling is a digital twin, a computational representation of a real-world system that is continually updated with new data. Unlike a conventional standalone simulation, a digital twin maintains an ongoing connection with the system it represents, enabling researchers to refine their predictions as new information becomes available.7

The potential applications of this technology in antimicrobial resistance research span several biological and epidemiological scales. A digital twin might represent a bacterial population, an infection within a patient, or the treatment environment in which bacteria are exposed to antibiotics. At a broader scale, related epidemiological models could help researchers understand how resistant bacteria spread through healthcare settings.8,9

Bringing these computational approaches together, a 2026 review proposed the Computational Antimicrobial Resistance Ecosystem (C-AMRE), a conceptual framework that integrates biological data, predictive modeling, mechanistic simulations, and clinical feedback into a continuously learning system. Within this framework, a digital twin of a bacterial infection could use genomic and antibiotic exposure data to simulate how resistance might evolve under different treatment conditions, with the longer-term aim of supporting personalized treatment strategies.8

Although this approach offers a potential bridge between evolutionary forecasting and precision medicine, AMR digital twins remain an emerging research concept. Further development and validation are needed before their predictions can reliably guide clinical decisions.8

Why Predicting Resistance Remains Difficult

Despite advances in computational modeling, predicting how antibiotic resistance will evolve remains challenging. Beyond mutations and antibiotic exposure, bacteria's evolutionary pathways are shaped by their genetic background, surrounding environment, host conditions, and interactions with other microorganisms. Adding to this complexity is horizontal gene transfer, through which bacteria can acquire resistance genes from other bacteria.9

These evolutionary processes also operate across different biological and epidemiological scales. While resistance may emerge within an individual patient, its subsequent spread across populations is shaped by additional factors. This distinction helps explain why findings from controlled laboratory experiments may not translate directly to clinical or population-level settings.4,9

Capturing this biological complexity in computational models presents a further challenge, particularly because their reliability depends on the quality and diversity of the underlying data. A model trained on known bacterial strains and resistance mechanisms may perform well under familiar conditions but struggle when applied to previously unseen bacterial populations or datasets collected in different settings. Predicting resistance pathways that have not yet emerged is particularly difficult, as models may lack the data needed to recognize entirely new evolutionary possibilities.6,10

Compounding these limitations are differences in laboratory susceptibility measurements, genomic data processing, dataset composition, and validation methods, which can make it difficult to compare models or reproduce their reported performance. Consequently, high predictive accuracy in one dataset does not necessarily translate into reliable predictions elsewhere.10

Improving forecast reliability will therefore require models that account for biological variation and are tested across different bacterial strains, antibiotics, and environments. Equally important will be independent validation and transparent reporting of predictive uncertainty, helping researchers establish where these models can provide reliable forecasts and where their limitations lie.6,10

From Detecting Superbugs to Anticipating Them

Although predicting the exact course of antibiotic resistance evolution remains challenging, advances in genomic surveillance, experimental evolution, mechanistic biology, and computational modeling are opening new possibilities for forecasting likely evolutionary pathways.4,5

Such forecasts could have several applications. Researchers could use these forecasts to investigate how bacteria might develop resistance to new antibiotics, explore treatment strategies that discourage particular resistance pathways, and identify mutations that warrant closer surveillance. Computational simulations could also help evaluate infection-control strategies before they are implemented.4,8

Rather than replacing surveillance, predictive approaches would build on it. As surveillance data become more comprehensive, they could provide the foundation for models that ask not only where resistance exists today, but where it might emerge next.2

The prospect of a superbug digital twin therefore represents a research direction rather than an established clinical capability. The next challenge is to establish whether these models can reliably anticipate how resistance will evolve outside the laboratory.4,6,8,10

References and Further Reading

  1. World Health Organization. Global antibiotic resistance surveillance report 2025: WHO Global Antimicrobial Resistance and Use Surveillance System (GLASS) – Summary [Internet]. World Health Organization; 2025 [cited 2026 Nov 16]. Available from: https://iris.who.int/handle/10665/383069 DOI:10.2471/B09585, https://iris.who.int/handle/10665/383069
  2. Global Antimicrobial Resistance and Use Surveillance System (GLASS) [Internet]. [cited 2026 Nov 12]. Available from: https://www.who.int/initiatives/glass
  3. Pinheiro F. Predicting the evolution of antibiotic resistance. Curr Opin Microbiol. 2024 Dec;82:102542. DOI:10.1016/j.mib.2024.102542, https://doi.org/10.1016/j.mib.2024.102542
  4. Charlebois DA. Quantitative systems-based prediction of antimicrobial resistance evolution. Npj Syst Biol Appl. 2023 Sep 7;9(1):40. DOI:10.1038/s41540-023-00304-6, https://www.nature.com/articles/s41540-023-00304-6
  5. Pinheiro F, Warsi O, Andersson DI, Lässig M. Metabolic fitness landscapes predict the evolution of antibiotic resistance. Nat Ecol Evol. 2021 Mar 4;5(5):677–87. DOI:10.1038/s41559-021-01397-0, https://www.nature.com/articles/s41559-021-01397-0
  6. Hossain E, Yousefi N. Computational paradigms for antimicrobial resistance prediction: integrating multi-omics, structural modeling, and foundation artificial intelligence systems. Brief Bioinform. 2026 May 11;27(3):bbag219. DOI:10.1093/bib/bbag219, https://academic.oup.com/bib/article-lookup/doi/10.1093/bib/bbag219
  7. National Academies of Sciences, Engineering, and Medicine. Foundational Research Gaps and Future Directions for Digital Twins [Internet]. Washington, D.C.: National Academies Press; 2024 [cited 2026 Nov 14]. Available from: https://www.nationalacademies.org/publications/26894 DOI:10.17226/26894, https://www.nationalacademies.org/publications/26894
  8. Patil A, Jadhav M, Hussain U, Kakde R. Reimagining antimicrobial resistance: AI-driven predictive epidemiology and the C-AMRE framework for next-generation antibiotic discovery. J Antibiot (Tokyo). 2026 Nov;79(11):699–718. DOI:10.1038/s41429-026-00951-x, https://www.nature.com/articles/s41429-026-00951-x
  9. Souque C, González Ojeda I, Baym M. From Petri Dishes to Patients to Populations: Scales and Evolutionary Mechanisms Driving Antibiotic Resistance. Annu Rev Microbiol. 2024 Nov 10;78(1):361–82. DOI:10.1146/annurev-micro-041522-102707, https://www.annualreviews.org/doi/10.1146/annurev-micro-041522-102707
  10. Elseifi N, Rogers H, Ferraro M, Boucher C. Bridging generalizability and reproducibility in machine learning-based antimicrobial resistance prediction. Zhai B, editor. mSphere. 2026 Nov 6;e00647-26. DOI:10.1128/msphere.00647-26, https://journals.asm.org/doi/10.1128/msphere.00647-26

Last Updated: Nov 18, 2026

Citations

Please use one of the following formats to cite this article in your essay, paper or report:

  • APA

    Some, Caren. (2026, November 18). Antibiotic Resistance Prediction: Genomics, AI, and Digital Twins. AZoLifeSciences. Retrieved on November 18, 2026 from https://www.azolifesciences.com/article/Antibiotic-Resistance-Prediction-Genomics-AI-and-Digital-Twins.aspx.

  • MLA

    Some, Caren. "Antibiotic Resistance Prediction: Genomics, AI, and Digital Twins". AZoLifeSciences. 18 November 2026. <https://www.azolifesciences.com/article/Antibiotic-Resistance-Prediction-Genomics-AI-and-Digital-Twins.aspx>.

  • Chicago

    Some, Caren. "Antibiotic Resistance Prediction: Genomics, AI, and Digital Twins". AZoLifeSciences. https://www.azolifesciences.com/article/Antibiotic-Resistance-Prediction-Genomics-AI-and-Digital-Twins.aspx. (accessed November 18, 2026).

  • Harvard

    Some, Caren. 2026. Antibiotic Resistance Prediction: Genomics, AI, and Digital Twins. AZoLifeSciences, viewed 18 November 2026, https://www.azolifesciences.com/article/Antibiotic-Resistance-Prediction-Genomics-AI-and-Digital-Twins.aspx.

Comments

The opinions expressed here are the views of the writer and do not necessarily reflect the views and opinions of AZoLifeSciences.
Post a new comment
Post

While we only use edited and approved content for Azthena answers, it may on occasions provide incorrect responses. Please confirm any data provided with the related suppliers or authors. We do not provide medical advice, if you search for medical information you must always consult a medical professional before acting on any information provided.

Your questions, but not your email details will be shared with OpenAI and retained for 30 days in accordance with their privacy principles.

Please do not ask questions that use sensitive or confidential information.

Read the full Terms & Conditions.