On the Cusp of the AI Revolution?

Tim Sandle, Head of GxP Compliance, Kedrion BioPharma, UK

Introduction

The application of artificial intelligence (AI) to support pharmaceutical processes has been made possible through advances in forms of machine intelligence and the accumulation of large-scale pharmaceutical data sets (so termed ‘Big Data’ sets). The twinning of computational power and data has, for example, helped to advance the screening and prioritizing of candidate drug pairs.¹

Advances have also been made to support pharmaceutical microbiology and, although at an early stage, the twinning of AI with advances in technology (including rapid microbiological methods) offer advantages to microbiologists working within this field. In addition, the projected technological curve suggests that further developments with AI for pharmaceutical microbiology are set to occur and that the use of AI will become more commonplace within the laboratory setting.

This article surveys the currently available technology; considers how technology may advance further; and weighs up some of the limitations and areas for concern.

What is ‘Artificial Intelligence’?

Many readers will be knowledgeable of AI and its distinct phases, but as a way of benchmarking AI can be categorized in diverse ways: the degree of intelligence; by function; and by application. These categories include:^2-4

Capability (Spectrum of Intelligence)

• Artificial Narrow Intelligence (ANI)/Weak AI: Excels at specific tasks (for example, Siri or Alexa, or a spam filter applied to email).
• Artificial General Intelligence (AGI)/Strong AI: Hypothetical AI with human-level cognitive abilities to learn and apply intelligence across various tasks.
• Artificial Superintelligence (ASI): Theoretical AI surpasses human intelligence in all aspects. 

Artificial superintelligence has yet to be developed in any meaningful way.

Function (Evolutionary Stages)

• Reactive Machines: Cannot form memories; only react to current inputs (e.g., Deep Blue chess AI).
• Limited Memory: Uses past data to improve decisions (e.g., self-driving cars).
• Theory of Mind: Can understand human emotions, beliefs, and intentions (future AI).
• Self-Aware: Possesses consciousness and self-awareness (theoretical future AI). 

The theory of mind and self-aware functions do not, as yet, exist.

Application

• Machine Learning (ML) & Deep Learning: Algorithms learning from data to find patterns and make predictions.
• Natural Language Processing (NLP): Enabling computers to understand, interpret, and generate human language (e.g., chatbots).
• Computer Vision: Allowing machines to “see” and interpret visual information (e.g., facial recognition, image analysis).
• Robotics: Building and programming robots for physical tasks, often integrating ML/CV.
• Generative AI: Creates added content like text, images, or code (e.g., ChatGPT, Midjourney).
• Expert Systems: Mimic human decision-making in specific domains using rules.

The expert systems application does not yet exist.

For most pharmaceutical microbiology applications, machine learning and deep learning are the applications of current interest. Machine learning is an AI system that acquires knowledge from data in order to generate predictions and decisions. It operates by examining data and recognizing patterns or associations within the data, through the use of algorithms. Deep Learning is a more advanced AI system that utilizes neural networks to solve intricate problems, such as image recognition and natural language processing, through artificial neural networks with multiple layers.⁵

Artificial Intelligence and Pharmaceutical Microbiology

The primary areas where AI assists pharmaceutical microbiology are:

• Microbial identification
• Integration with molecular diagnostics
• Antimicrobial susceptibility testing 
• Detection of ‘objectionable’ organisms in real-time
• Workflow efficiency
• Predictive environmental monitoring

Microbial Identification and Pattern Recognition

To aid with the counting and identification of microorganisms, AI-powered image analysis and machine learning algorithms aid with the rapid identification of microbial species and strains from microscopic images or colony morphology, outperforming manual methods in speed and accuracy.⁶

This form of technology relies on Convolutional Neural Networks (CNNs), which analyze microscopic images of colonies and cells to classify microbial species. The objective is to accelerate image reviews through automation and to reduce the reliance on manual interpretation of Gram stains and colony morphology, which is time-consuming and error-prone. An example is with AI algorithms like Deep Colony which achieved >95% accuracy in identifying bacterial strains from culture plates, significantly speeding up diagnostics.⁷ Such processes can also be integrated with identification systems. For example, there are several developments with image analysis and machine learning-assisted MALDI-TOF MS.^8,9

In addition, algorithms can aid with the sorting of data to look for patterns and trends¹⁰ (such as screening for spore formers) or comparing different zones within a cleanroom or between cleanrooms of different grades (especially where cleanrooms interface, such as airlocks). This type of application fits with the holistic contamination control strategy concept required by EU GMP Annex 1.

For researchers looking for new drug candidates, especially antimicrobials, AI can also integrate with genomic sequencing to predict antimicrobial resistance patterns.¹¹ This occurs at three levels: prediction of antimicrobial resistance using genomic data; using machine learning to gain insight into the cellular functions disrupted by antibiotics; and applying AI for antimicrobial stewardship, using data extracted from electronic health records.

Generative models (e.g., Variational Autoencoders) are particularly useful in accelerating pharmaceutical innovation by the virtual screening of millions of compounds to identify new antimicrobials and predicting drug–microbe interactions (such as predicting binding affinity and activity against microbial targets). Machine learning models analyze multi-omics data (genomics, proteomics, metabolomics) to identify novel drug targets and predict pathogen evolution. Microbiome datasets are vast, complex, and multi-omics (genomics, transcriptomics, metabolomics). Machine learning and deep learning algorithms (e.g., clustering, dimensionality reduction, and convoluted neural networks) have assisted in extracting meaningful biological patterns from such high-dimensional datasets. The advantages are with the identification of microbial community dynamics, host–microbe interactions, and functional genomics at scale.¹² The goal for many medical applications, in terms of improving treatment outcomes, is facilitating personalized medicine by analyzing a patient’s microbiome data to tailor treatments for individual patients. 

As examples, AI-driven platforms predict protein structures (e.g., DeepMind’s AlphaFold), enabling faster validation of targets for antimicrobial drugs¹³ As a real-world example, AI-assisted screening identified halicin and abaucin - novel antibiotics effective against multidrug-resistant bacteria.¹⁴ This came about by AI predicting ADMET (Absorption, Distribution, Metabolism, Excretion, Toxicity) properties and optimizes molecular structures for safety and efficacy.¹⁵

Integration with Molecular Diagnostics

In terms of supporting laboratory analyses, AI can enhance PCR and next-generation sequencing (NGS) workflows by interpreting complex genomic data rapidly. This enables precise identification of pathogens and prediction of antimicrobial resistance patterns. As an example, machine learning models combined with DNA sequencing improve pathogen identification and susceptibility testing, reducing turnaround times from days to hours.¹⁶

Rapid Antimicrobial Susceptibility Testing (AST)

AI can automate AST by analyzing growth patterns and zone diameters in culture plates. Through this application, this provides faster and more accurate antibiotic recommendations, supporting antimicrobial stewardship. Within laboratories, there are several applications of AI-driven AST systems delivering real-time susceptibility profiles and potentially improving treatment decisions.¹⁷

Real-Time Objectionable Organism Detection

As an example of artificial intelligence aiding the detection of organisms of concern, AI integrates with Raman spectroscopy and Surface Enhanced Raman Spectroscopy (SERS) for label-free, rapid microbial detection. A Raman spectrum is a unique “fingerprint” of a material, generated by shining a laser on it and analyzing the light scattered back. In terms of impact, this combination detects specific organisms within minutes, without the need for extensive sample preparation.

In different studies, AI-assisted Raman spectroscopy has achieved >95% accuracy in microbial identification in under 5 minutes.¹⁸

Workflow Efficiency and Cost Reduction

Within the laboratory setting, AI, through automation, can tackle repetitive tasks like colony counting, Gram stain interpretation, and culture plate reading. The impact of Reinforcement Learning AI models can be to reduce manual labor, minimize errors, and accelerates assay workflows. The impact is greater when AI is integrated with advanced robotics and software.¹⁹

In terms of results interpretation, AI can provide the autoverification of results from bioburden samples, with autoresulted by screening for alert and action level excursions, thereby eliminating the need for human intervention in routine tasks.²⁰

Predictive Environmental Monitoring

AI is capable of drawing rich inferences from data sets, not only for understanding what has occurred but also for making predictions as what might occur – especially with or without interventions (such as increased cleaning and disinfection). This is through AI analyses of environmental data trends to predict contamination risks, using machine learning approaches initially developed to screen for disease outbreaks (such as Random Forest and Support Vector Machines).²¹ Hence, such technologies can support proactive contamination control and deploy predictive models to help to forecast microbial contamination in pharmaceutical environments. 

An example of AI predictive modelling for contamination risks is where AI predictive models analyze historical environmental monitoring data including particle counts, microbial trends, temperature, humidity, and pressure, in order to identify subtle deviations that precede contamination events. Machine learning algorithms are able to detect the patterns that humans might miss. The advantage here is to enable early intervention before significant shifts with the contamination profile occur. AI can also move the contamination control paradigm beyond periodic reviews to continuous predictive analysis. By correlating data with contamination control strategies, AI can learn to forecast contamination risks in cleanrooms and controlled spaces.

AI can be used to trigger automated disinfection protocols and generate validated compliance reports for GMP. This type of predictive modelling ensures that interventions are timely and documented, potentially reducing regulatory risks. 

AI systems can also assist with real-time monitoring when they combine with Internet-of-Things-enabled sensors (for environmental conditions, such as a Facility Management System) and newer technologies like biofluorescent particle counters. This continuous monitoring can feed predictive algorithms to flag anomalies, allowing proactive corrective actions rather than reactive responses after contamination has already happened. 

AI can also provide predictive maintenance. Contamination risk often correlates with equipment performance. AI can predict failures in HEPA filters, HVAC systems, and sensors by analyzing degradation patterns, enabling maintenance before breakdowns compromise sterility. This reduces downtime and contamination incidents. 

Preparing for AI

While AI is potentially powerful, it will not be effective if its integration has not been carefully considered. As with any laboratory item (including rapid microbiological methods), planning is key. As a way of example, consider the incorporation of AI into microbial identification. A potential eight phase ‘road-map’ is:

Phase 1: Strategic Planning

• Define Objectives: Such as to speed up microbial identification, improve accuracy, and reduce manual workload.
• Scope: Identify target processes (e.g., colony counting, Gram stain interpretation, antimicrobial susceptibility testing).

Phase 2: Data Preparation

• Digitize Historical Data: Collect and standardize past microbial identification records, images, and genomic data.
• Ensure Data Quality: Validate datasets against ALCOA principles (Attributable, Legible, Contemporaneous, Original, Accurate).
• Label Training Data: Annotate images and genomic sequences for supervised learning models.

Phase 3: Infrastructure Setup

• Hardware: High-resolution imaging systems and Internet of Things-enabled sensors.
• Software: Secure cloud or on-premises platform for AI model training and deployment.
• Integration: Connect AI tools with LIMS (Laboratory Information Management System) for seamless data flow. When AI integrates with LIMS and Electronic Lab Notebooks, this can optimize scheduling, resource allocation, and experiment sequencing.

Phase 4: AI Model Development

• Image Analysis Models: Train convolutional neural networks (CNNs) for colony and cell morphology classification.
• Genomic Interpretation Models: Implement machine learning algorithms for rapid sequencing data analysis.
• Validation: Test models against known microbial datasets and regulatory standards.

Phase 5: Pilot Deployment

• Start Small: Apply AI to one diagnostic workflow (e.g., Gram stain interpretation).
• Monitor Performance: Compare AI predictions with manual results for accuracy and turnaround time.
• Adjust Models: Refine algorithms based on pilot feedback.

Phase 6: Full-Scale Implementation

• Expand Coverage: Integrate AI across all microbial identification and diagnostic workflows.
• Automate Reporting: Generate compliance-ready reports for GMP and regulatory audits.
• Link to Predictive Analytics: Combine diagnostic data with contamination risk modelling for proactive control.

Phase 7: Training and Change Management

• Upskill Staff: Provide training on AI tools and interpretation.
• Develop SOPs: Standardize processes for AI-assisted diagnostics.

Phase 8: Continuous Improvement

• Retrain Models: Update AI algorithms with new data regularly.
• Monitor KPIs: Track improvements in turnaround time, accuracy, and contamination prevention.
• Feedback Loop: Incorporate user feedback for ongoing optimization.

The above roadmap illustrates how AI applications can be integrated into the pharmaceutical microbiology workflow.

Challenges and Considerations

The current and future integration of AI into pharmaceutical microbiology workstreams is not without challenges. The need for careful planning has been noted above; yet there are other areas to consider, not least that the reliability of any AI is only as good as the data input. This means considerable effort needs to be put into ensuring good data quality and standardization. In addition, the datasets used must meet data integrity requirements. If these concerns are not adhered to, then poor or inconsistent data will lead to inaccurate predictions. 

All forms of AI face ethical concerns around data privacy and algorithmic bias. AI systems often process sensitive data, including genomic and patient-related information and sufficient safeguards need to be in place.²²

It is important to establish interpretable AI models to meet both regulatory and clinical standards.²³ Too often AI algorithms, especially deep learning models, are mysterious domains and a lack of interpretability will hinder both trust and regulatory acceptance, especially where decision making logic is not clearly defined for each prediction. It also needs to be noted that regulatory frameworks are still evolving for AI applications. It is probable regulators will expect model transparency and explainability;²⁴ validation protocols; change management; effective data governance; the use of proven data validation tools; and clear audit trails for all data inputs and AI outputs. There will also need to be a ‘human-in-the-loop’ to review all critical decisions.

A further challenge is cost and the infrastructure and skills gap. Many microbiology laboratories will lack the infrastructure for AI deployment as well as not necessarily having personnel trained in bioinformatics and machine learning. There may also be resistance to change among personnel and integration into existing workflows.²⁵

Summary

As this article is provided, AI continues to make strides in the laboratory space and this includes some of the specific requirements of pharmaceutical microbiology. AI accelerates microbial identification and diagnostics by combining image analysis, genomic interpretation, rapid AST, real-time detection, and workflow automation - leading to faster, more accurate, and cost-effective microbiological testing. The most powerful component is predictive AI. This fits with the concept of proactive contamination control and this could offer the advantage of reduced batch failures.

Nonetheless, developing these tools is not without its challenges. Implementation takes time and requires high-quality, standardized data and expert interpretation by a pharmaceutical microbiologist remains essential for context-specific decisions. Furthermore, while AI offers transformative potential for pharmaceutical microbiology it introduces challenges around data integrity, regulatory validation, ethics, infrastructure, and cybersecurity. It also remains uncertain what the regulatory position and additional requirements will be.

References

1. Shtar G, Azulay L. Nizri O. et al. CDCDB: a large and continuously updated drug combination database. Sci Data. 2022: 9(1): 263
2. Wang MH. Artificial intelligence and subfields. Santa Clara University, Department of Computer Engineering. 2017
3. Gordon JS. What is artificial intelligence? In The impact of artificial intelligence on human rights legislation: A plea for an AI Convention 2023 (pp. 15-28). Cham: Springer Nature Switzerland.
4. Samoili, S., Cobo, M.L., Gómez, E., De Prato, G., Martínez-Plumed, F. and Delipetrev, B., 2020. AI Watch. Defining Artificial Intelligence. Towards an operational definition and taxonomy of artificial intelligence, CRC Press, USA
5. El Samad M, Nasserddine G, Kheir A. Introduction to artificial intelligence. In Artificial intelligence and knowledge processing 2023 (pp. 1-14). CRC Press,USA
6. Wang, X., Shi, Y., Guo, S. et al. A Clinical Bacterial Dataset for Deep Learning in Microbiological Rapid On-Site Evaluation. Sci Data 11, 608 (2024). https://doi.org/10.1038/s41597-024-03370-5
7. Signoroni, A., Ferrari, A., Lombardi, S. et al. Hierarchical AI enables global interpretation of culture plates in the era of digital microbiology. Nat Commun 14, 6874 (2023). https://doi.org/10.1038/s41467-023-42563-1
8. Kotwal, S.; Rani, P.; Arif, T.; Manhas, J.; Sharma, S. Automated Bacterial Classifications Using Machine Learning Based Computational Techniques: Architectures, Challenges and Open Research Issues. Arch. Comput. Methods Eng. 2022, 29, 2469–2490
9. Wu, Y.; Gadsden, S.A. Machine learning algorithms in microbial classification: A comparative analysis. Front. Artif. Intell. 2023, 6, 1200994
10. Alnowaiser, K. MicrobeNet: An Automated Approach for Microbe Organisms Prediction Using Feature Fusion and Weighted CNN Model. Int J Comput Intell Syst 18, 55 (2025). https://doi.org/10.1007/s44196-025-00777-9
11. Anahtar, M.N.; Yang, J.H.; Kanjilal, S. Applications of Machine Learning to the Problem of Antimicrobial Resistance: An Emerging Model for Translational Research. J. Clin. Microbiol. 2021, 59
12. Zhou T, Zhao F AI-empowered human microbiome research, Gut, 2025. doi: 10.1136/gutjnl-2025-335946
13. May, M. How to fight antibiotic resistance. Nature Medicine. 2023; 29 (7): 1583–1586. doi:10.1038/d41591-023-00043-5
14. Liu G, Catacutan DB, Rathod K, et al. Deep learning-guided discovery of an antibiotic targeting Acinetobacter baumannii. Nature Chemical Biology. 2023; 19 (11): 1342–1350. doi:10.1038/s41589-023-01349-8
15. Catacutan, D.B., Tran, V., Arnold, A. et al. Discovery and artificial intelligence-guided mechanistic elucidation of a narrow-spectrum antibiotic. Nat Microbiol 2025; 10, 2808–2822 https://doi.org/10.1038/s41564-025-02142-0
16. Santos-Junior, C. D., Torres, M. D. T., Duan, Y., et al. Discovery of antimicrobial peptides in the global microbiome with machine learning. Cell 2024; 187, 3761–3778.e16. doi: 10.1016/j.cell.2024.05.013
17. Meenakshi R. The Role of AI in Microbiological Diagnostics: Innovations and Future Prospects, Int. J. of Pharm. Sci., 2025, 3 (2): 327-328. https://doi.org/10.5281/zenodo.14810967
18. Rahman MH-U, Sikder R, Tripathi M, et al. Machine Learning-Assisted Raman Spectroscopy and SERS for Bacterial Pathogen Detection: Clinical, Food Safety, and Environmental Applications. Chemosensors. 2024; 12(7):140. https://doi.org/10.3390/chemosensors12070140
19. Ford BA, McElvania E. Machine learning takes laboratory automation to the next level. Journal of clinical microbiology. 2020 Mar 25;58(4):10-128
20. Mencacci A, De Socio GV, Pirelli E, et al. Laboratory automation, informatics, and artificial intelligence: current and future perspectives in clinical microbiology. Frontiers in Cellular and Infection Microbiology. 2023 Jun 27;13:1188684
21. Li, R., Shen, M., Liu, H. et al. (2023). Do infrared thermometers hold promise for an effective early warning system for emerging respiratory infectious diseases? JMIR Form Res. 7:e42548. doi: 10.2196/42548
22. Miller S, Chiu C, Rodino Kg, et al. Should we be performing meta-genomic next-generation sequencing for infectious disease diagnosis in the clinical laboratory? J clin Microbiol. 2019;58(3). doi: 10.1128/JcM.01739-19
23. Tsitou, V-M., Rallis, D., Tsekova, M. and Yanev, N. (2024) Microbiology in the era of artificial intelligence: transforming medical and pharmaceutical microbiology, Biotechnology & Biotechnological Equipment, 38:1, 2349587 DOI: 10.1080/13102818.2024.2349587
24. Brall C, Schröder-Bäck P, Maeckelberghe E. Ethical aspects of digital health from a justice point of view. Eur J public health. 2019;29(Supplement_3):18e22–18e22. doi: 10.1093/eurpub/ckz167
25. Scott IA, Cook D, Coiera EW, et al. Machine learning in clinical practice: prospects and pitfalls. Med J Aust. 2019;211(5):203–205.e1. doi: 10.5694/mja2.50294.

Subscribe to our e-Newsletters
Stay up to date with the latest news, articles, and events. Plus, get special
offers from American Pharmaceutical Review delivered to your inbox!
Sign up now!