We conducted our analysis using the Web of Science Core Collection (WoSCC): Science Citation Index Expanded (SCIE) and Social Sciences Citation Index (SSCI) as our data source. We ensured the validity of our conclusions by using the WoSCC, a globally recognized authoritative database and the most utilized resource in bibliometric research (19). Our search strategy encompassed two thematic areas: keywords relating to AI, allergies, and immunology, along with their synonyms. The search strategy was as follows: TS = (“Allergy” OR “Immunology” OR “Clinical Immunology” OR “Immunochemistry” OR “Immunoinformatics” OR “immunolog*” OR “Allergy Specialty” OR “Allergic Reactions” OR “Hypersensitivity” OR “Immune System” OR “Allergy” OR “Food Allergy” OR “Asthma” OR “Bronchial Asthma” OR “Primary immunodeficiency” OR “Primary Immune Deficiency” OR “Autoimmune Diseases” OR “Anaphylaxis” OR “Atopic Dermatitis” OR “Eosinophilic Disorders” OR “Immunodeficiency Syndromes” OR “Allergen-specific Immunotherapy”) AND TS = (“artificial intelligence” OR “computational intelligence” OR “machine learning” OR “deep learning” OR “neural networks” OR “reinforcement learning” OR “random forest” OR “support vector machine” OR “big data” OR “image processing” OR “feature learning” OR “convolutional neural networks” OR “evolutionary algorithms” OR “natural language processing” OR “Bayesian networks” OR “Bayesian learning” OR “multiagent systems”), with the search culminating on August 15, 2024. We limited the inclusion criteria to articles and reviews published in English. Two authors reviewed the titles and abstracts of 4,327 papers from WOSCC. Among the 4,346 papers retrieved from WoSCC, we excluded 432 articles belonging to the fields of information security, architecture, materials science, landscape architecture, botany, neurology, dermatology, cardiology, and food science, and excluded 12 retracted articles, yielding 3,883 papers (see Figure 1 for details). Out of 3,883 papers, 511 were reviews and 3,372 were original articles. Furthermore, of the 3,372 original articles included, 296 referenced or employed bioinformatics methodologies. Then we exported the selected publications in plain text and table-delimited formats, capturing essential details such as publication year, title, authors, abstracts, and keywords.

We imported the data into CiteSpace V5.7.R1 (Drexel University, Chaomei Chen, USA) and Excel 2019 (Microsoft, Washington, United States) for deduplication, organization, and analysis. No duplicate articles were found. CiteSpace, a bibliometric program developed by Chaomei Chen based on JAVA, is widely used in the field of bibliometrics (18, 20). We also utilized https://bibliometric.com/ and the bibliometrix package in R software version 4.0.3 to analyze collaborative relationships between different authors, countries, and regions. In the presented visualizations, the weight of the parameters determines the size of the nodes, the more significant the weight, the larger the node. Each node represents different parameters, including countries, institutions, authors, and keywords. We assign different colors to nodes and lines based on their cluster type or timeline, with identical colors denoting the same cluster or timeline. The thickness of the connecting lines indicates the strength of the linkage (17).

As of August 15, 2024, we had included 3,883 relevant publications in the WoSCC. Figure 2 illustrates the publication trends of AI-related literature in allergy and immunology. Prior to 2001, the growth of publications in this domain was slow. Between 2002 and 2015, the number of publications began to increase, with rapid growth observed after 2015. The field has experienced vigorous development in recent years, as evidenced by the publication of 2,958 articles over the past 6 years (2019–2024), accounting for 76.18% of the total output. Citation data show a rise in citations starting in 2002, with particularly sharp increases after 2016. This positive momentum has persisted in recent years. The 3,883 relevant publications had received 79,539 citations as of August 15, 2024, with an average of 20.5 citations per article (see Figure 2 for details).

Researchers from 113 countries and regions have contributed to the study of AI in allergy and immunology, leading to numerous publications in this field. In 1981, the first article in this domain was published in Experimental Pathology by German scholar Haroske et al. (21). The article discussed the utilization of automated image processing to distinguish the subtle characteristics of various groupings of human lymphocytes in unpurified, unlabeled lymphocyte suspensions obtained from peripheral blood. Despite Germany’s early contribution, the USA has emerged as the leading country in terms of publication volume. Farmer et al. (22) authored the first USA publication in this field in Physica D: Nonlinear Phenomena, detailing a dynamic model of the immune system based on the Jerne network hypothesis. Since 2016, the USA has seen a rapid increase in related research output. Between 1986 and 2015, the USA published 179 articles in this field, but from 2016 to 2024, this number surged to 964, bringing the total to 1,143 articles, representing 29.44% of the global output. China, which published its first related research in 2001, has also seen significant growth, particularly after 2016, with 1,031 articles published by 2024, accounting for 26.55% of the global research output. The USA and China now dominate this field, publishing the largest number of articles and engaging in extensive collaborations with other countries, with strong growth momentum continuing (see Figures 3, 4). Among the top 10 countries contributing to publications, the UK has participated in 395 studies, followed by Germany in 289, Italy in 214, India in 195, South Korea in 174, Spain in 164, France in 155, and Australia in 151 studies. Moreover, Figure 4 illustrates that in addition to China and the USA, nations including the UK, Germany, France, Japan, and South Korea are also performing a large amount of collaborative research in this domain. This is due to these countries presently occupying favorable positions in vaccine development, clinical trials, and AI-enhanced immunology research.

We also analyzed the affiliations of the corresponding authors for the 3,883 articles. Among these, 961 articles (24.75% of the total) had corresponding authors from China, with 821 articles having Chinese authors as the sole corresponding authors and 140 articles co-authored with corresponding authors from other countries. The USA contributed 847 articles (21.81% of the total), with 629 of these having USA-based sole corresponding authors and 218 co-authored with corresponding authors from other countries. Together, China and the USA accounted for 46.56% of the corresponding authority in this field, indicating their significant leadership (see Table 1). Additionally, 214 articles from the UK, 159 from Germany, 153 from India, 139 from South Korea, 134 from Italy, 93 from Spain, 85 from France, and 83 from Australia represented corresponding authorship.

In terms of citation counts, the USA leads with 25,580 citations, followed by China with 9,510 citations. Other top 10 countries by citation count include the UK (n = 6,578), Germany (n = 2,999), India (n = 2,390), Denmark (n = 2,330), Australia (n = 2,279), Italy (n = 2,090), France (n = 2,064), and South Korea (n = 1,938). Denmark ranks first in average citations per paper, with an impressive 75.20 citations per article. The UK follows with 30.70 citations per article, with the USA close behind at 30.20 citations per article. China, however, lags with an average of only 9.90 citations per article, indicating that it still has some ground to cover in terms of impact (see Figure 5 for details).

A total of 4,269 institutions contributed to the 3,883 published articles. Harvard University leads the field with 294 publications, followed by the University of California system with 199 articles, and Stanford University with 148 articles. Other top institutions include Harvard Medical School (134 articles), Université Paris Cité (123 articles), University of London (118 articles), Institut National de la Santé et de la Recherche Médicale (109 articles), Chinese Academy of Sciences (102 articles), University of Ohio (95 articles), and Mayo Clinic (93 articles) (see Figure 6 for details). Of the top 10 institutions, six are based in the USA, two in France, one in China, and one in the UK, further underscoring the USA’s significant dominance in this field.

In conclusion, the USA and China dominate allergology and immunology research, with the USA leading in impact and China excelling in publication volume but lagging in influence. The USA’s strong research infrastructure, high international collaboration, and leadership in pharmaceuticals and AI-driven immunology contribute to its global authority. In contrast, China’s lower international collaboration limits its influence, despite significant government investment and vaccine advancements.

European nations-France, Germany, Spain, and the UK-stand out for their high international collaboration, which enhances research impact. Meanwhile, South Korea and India are emerging players but require greater global integration to improve their influence. Strengthening international partnerships will be key for China and other rising nations to match the USA’s research impact in the future.

A total of 21,552 authors have contributed to research on AI in allergy and immunology. We have summarized and analyzed the top 15fifteen most productive authors (see Table 2 for details). Adnan Custovic from Imperial College London leads the field with 34 related publications between 2010 and 2024. His main research interests are using ML to assess the types of asthma and eczema people will have (23, 24), figuring out how allergen microarrays relate to clinical symptoms (25), and studying how allergic sensitization happens and what factors make it more likely to happen (26–28). He has also worked on developing predictive models for asthma in children and adults (29), as well as investigating the application of AI in allergy research (30, 31). Morten Nielsen, from the Technical University of Denmark, published 21 related publications from 2005 to 2024. His research interests include using ML to improve prediction algorithms for MHC Class I/II, T cells, and B cell epitopes (32–34), enhancing understanding of cellular immunopeptides (35), and improving predictions of protein drug immunogenicity (36). Angela Simpson from the University of Manchester has closely collaborated with Adnan Custovic, publishing 21 papers between 2010 and 2024. She likes to use AI to look into different sensitization patterns related to asthma (28), study the link between different allergic phenotypes and asthma (37), find asthma subgroups (38), predict phenotypes for asthma and eczema (24), and look into the relationship between cytokine expression patterns and clinical outcomes (39, 40). Gang Luo, from the University of Washington in the USA, published 14 related articles between 2015 and 2024. He has extensively worked on using AI to predict childhood asthma (41, 42), assist asthma patients in seeking medical care (43), and support clinical decision-making for physicians (44). Tesfaye B. Mersha from the University of Cincinnati published 13 related papers between 2015 and 2024. His research focuses on using ML and deep learning for multi-omics analysis (45), analyzing sex dimorphism (46), predicting individual risk genes, and promoting the development of precision medicine (47). He has also developed and validated risk prediction models for asthma (48).

Kimberly G. Blumenthal from Massachusetts General Hospital, Bjoern Peters from the La Jolla Institute for Allergy and Immunology, and Scott T. Weiss from Harvard Medical School, along with Yaron Ilan from Hebrew University in Israel and Gajendra P.S. Raghava from the Indraprastha Institute of Information Technology in India, each published 12 articles in this field. Kimberly G (see Table 2 for details). Blumenthal began her research in 2016, focusing on using ML to explore new approaches to studying the epidemiology of drug allergies (49), improving methods in electronic health records (EHRs) (50, 51), and developing models for predicting patient allergic reactions (52). Yaron Ilan published 12 articles from 2021 to 2024, advocating for and advancing the development of second-generation AI systems to improve diagnosis and detection of rare diseases (53), optimize drug dosing and study drug resistance (54, 55), and enhance global healthcare in an intelligent, digital manner to reduce healthcare costs (56). Bjoern Peters has established a strong collaboration with Morten Nielsen. In 2007, Bjoern Peters published his first article in this field, aligning his research interests with Nielsen’s. He focuses on using ML to predict the binding of peptides to MHC Class I/II molecules and exploring non-canonical patterns (57, 58). Peters has also worked on improving predictions for MHC II antigen presentation (59), as well as T cell and B cell epitope prediction (60, 61). From 2007 to 2024, Gajendra P.S. Raghava published 11 related papers, optimizing methods for predicting linear and conformational B cell epitopes in antigens (62), and improving algorithms for predicting and identifying receptors and defensins (63, 64). Scott T. Weiss studies how ML can be used to evaluate how people will react to short-acting bronchodilators (65), figure out gene expression and genetic polymorphisms (66), look into the molecular mechanisms of asthma (67), investigate childhood asthma metabolomics (68) and develop new models for using electronic health records to test and predict asthma (69).

Among other prolific contributors, Yudong Cai from Shanghai University published 11 related papers between 2011 and 2024, with a focus on using ML to identify immune gene signatures of immune cell subtypes (70), particularly in identifying immune cell markers for COVID-19 (71, 72). Salih Güneş from Selcuk University published 10 relevant publications from 2005 to 2013. Kemal Polat from Bolu Abant İzzet Baysal University, who collaborated extensively with Salih Güneş, published 10 publications between 2004 and 2009. Their research primarily explored computer-assisted medical diagnostic systems and medical decision support systems supported by algorithms like the artificial immune recognition system (AIRS), fuzzy weighted pre-processing, feature selection, and component analysis (73, 74). Nima Aghaeepour from Stanford University School of Medicine also published 10 related papers, primarily exploring the relationships between patient emotions, lifestyle, and peripheral and systemic immune responses using ML and deep learning (75, 76). Vladimir Brusic from the University of Nottingham Ningbo in China published 10 related articles between 2002 and 2022, focusing on using ML to develop new tools for analyzing immunological data (77, 78).

Among these prolific authors, Adnan Custovic ranks first in local citations with 263, followed by Angela Simpson with 178 citations and Morten Nielsen with 161 (see Table 2 for details). These highly productive and highly cited authors have contributed to varying extents to the practical application of AI in this field.

We dispersed publications related to AI in allergy and immunology across 1,247 journals. We classified these journals according to Bradford’s Law, which divides journals into three zones, each containing approximately one-third of the total publications. This classification helps identify the most influential core journals within a specific academic field (19, 79). Our analysis revealed 42 journals in Zone 1, 233 in Zone 2, and 972 in Zone 3. We summarized the impact factor (IF), quartile, and category for the top 15 journals in Zone 1 (see Table 3 for details). Frontiers in Immunology emerged as the leading journal in this field, publishing 199 papers, which account for 5.12% of the domain’s output. This was followed by Scientific Reports with 90 publications and PLOS One with 69 publications. Other journals in the top 15 include IEEE Access (n = 49), Journal of Allergy and Clinical Immunology (n = 48), BMC Bioinformatics (n = 46), Briefings in Bioinformatics (n = 46), Frontiers in Genetics (n = 45), Bioinformatics (n = 35), International journal of Molecular Sciences (n = 35), Allergy (n = 35), Cancers (n = 29), Computers in Biology and Medicine (n = 29), Heliyon (n = 29), and Journal of Allergy and Clinical Immunology-in Practice (n = 29). Together, the top 15 journals published 807 papers, accounting for 20.78% of the total publications in the field.

An analysis of local citations within the field can provide insights into foundational research and emerging hotspots. We ranked the local citations of the 3,883 included publications and identified the top 15 most-cited articles (see Table 4 for details). The most locally cited article, authored by Vanessa Jurtz in 2017 with 44 local citations, introduced NetMHCpan-4.0, a novel ML framework trained to predict MHC peptide presentation. This method demonstrated strong predictive performance in identifying validated eluted ligand data, cancer neoantigens, and T cell epitopes (58). Joseph Finkelstein published the second most locally cited article in 2017 with 41 local citations, using ML to predict asthma exacerbations early. This highlights the significant potential of ML in developing remote monitoring systems for chronic diseases (80). The third-ranking article, with 35 local citations, was authored by Ilka Hoof in 2009 and focused on predicting peptide binding to MHC Class I molecules using NetMHCpan-2.0 (81). Martin Closter Jespersen and colleagues introduced the BepiPred-2.0 tool in 2017 (82), which garnered 32 local citations and ranked fourth. Bioinformatics widely used BepiPred-2.0 for predicting B cell epitopes in antigen sequences (82). Six of the top 15 most locally cited articles used ML to identify various asthma sensitization patterns, distinguish and predict asthma phenotypes, and develop allergen prediction models (28, 37, 80, 83–85). Four articles focused on ML algorithms to identify T and B cell epitope specificity, optimize antigen specificity recognition, and predict peptide binding to MHC Class I molecules (58, 81, 82, 86). Three articles summarized applications and developments of AL in the field of asthma (1, 87, 88). Additionally, two articles discussed dynamic models of the immune system and artificial immune systems (AIS) (22, 89).

Keyword co-occurrence analysis is instrumental in identifying focal points within a domain and exploring the internal connections among them. In the 3,883 publications included in our study authors used 8,849 distinct keywords. We selected the top 50 most frequent keywords for analysis. As shown in Figure 7 “machine learning” appeared most frequently with 871 mentions followed by “asthma” with 305 mentions “artificial intelligence” with 235 mentions and “deep learning” with 226 mentions. “AI” encompasses the broadest meaning among AI-related terms and ML is a subset of AI that involves algorithms learning from data to make predictions or decisions without explicit programming. Deep learning is a more complex subset of ML that uses neural networks with many layers (deep neural networks) to analyze various disease-related factors including genetic data and medical images. Overall the field often uses “artificial intelligence,” “machine learning,” and “deep learning” interchangeably despite their differing meanings. Other high-frequency keywords include “immune system” (n = 78) “artificial immune system” (n = 73) “prediction” (n = 73) “immune infiltration” (n = 69) “support vector machine” (n = 55) “diagnosis” (n = 53) “neural networks” (n = 42) and “data mining” (n = 40) (see Table 5 for details). The keywords “artificial immune system” and “immune system” reflect the symbiotic relationship between AI and immunology. Inspired by the biological immune system the AIS mimics its mechanisms to solve complex problems like pattern recognition anomaly detection predictive analysis and adaptive learning. In allergy and immunology AIS can simulate immune system responses under different conditions improving the accuracy of diagnosing allergic and immune diseases and facilitating personalized treatment plans based on individual immunological profiles (90). High-frequency terms like “prediction” and “diagnosis” highlight AI’s specific applications in the fields of immunology and allergy while the frequent mention of “asthma” underscores AI’s widespread application in asthma research. Indeed AI applications and advancements in asthma are the subject of nine out of the top 15 most locally cited articles (1, 28, 37, 80, 83–85, 87, 88).

Additionally, we conducted a thematic map and evolutionary analysis of keywords to better understand the structure and dynamics within the research domain and predict future research focal points. As shown in Figure 8, the thematic map divides various topics into four quadrants based on centrality and density: (1) Upper Right—Motor Themes: This quadrant contains themes with high centrality and high density, indicating well-developed, mature, and reliable topics that are critical to the field. The themes of “machine learning,” “artificial intelligence,” “deep learning,” “biomarkers,” “immunotherapy,” and “prognosis” fall into this quadrant. From 2021 to 2023, ML and DL, as subsets of AI, have frequently appeared in immunology and allergy research. These technologies have applications in disease prediction, patient stratification, drug discovery, analysis of immune responses, development of personalized immunotherapies, clinical decision support, and prognosis management. Additionally, AI has made significant contributions to addressing the challenges posed by COVID-19, particularly in vaccine development, real-time tracking of viral mutations, and predicting patient outcomes. (2) Upper Left—Niche Themes: These themes exhibit low centrality but high density, indicating that they are well-researched yet have had less impact on the broader field. The themes of “immunoinformatics,” “epitope,” “immune system,” “neural networks,” and “feature extraction” belong to this quadrant. Immunoinformatics combines immunology and bioinformatics to analyze immune system data, predict immune responses, and develop computational models for vaccine design and epitope prediction. The keywords “immune system,” “epitope,” and “immunoinformatics” are closely related, as epitopes are specific regions of antigens critical for vaccine development and immunotherapy. Neural networks show great potential for enhancing epitope prediction models and analyzing immune infiltration data. These keywords represent important interdisciplinary research areas between AI and immunology, which, despite their current limited impact on the broader field, have the potential to become core themes as AI becomes further integrated into these topics. (3) Lower Right—Basic Themes: Characterized by high centrality and low density, the themes in this quadrant are less researched but are gaining influence, with the potential to become central to future studies. Keywords in this quadrant include “random forest,” “classification,” “support vector machine,” “asthma,” “big data,” and “prediction.” ML algorithms like random forest and support vector machine (SVM) excel at classifying and predicting high-dimensional and heterogeneous data. As the volume of immunology and allergy data continues to grow, the ability to effectively analyze big data will be crucial to advancing research in this field. Future studies may focus on combining multiple classification algorithms to improve the accuracy and robustness of predictive models in immunology and allergy, particularly in diseases like asthma. Keywords such as “random forest,” “classification,” “support vector machine,” “asthma,” “big data,” and “prediction” represent emerging research themes in AI within the fields of immunology and allergy. As AI continues to evolve, these themes are likely to become central to the development of more accurate diagnostic tools, predictive models, and personalized treatment strategies. (4) Lower Left—Emerging or Declining Themes: This quadrant encompasses themes with low centrality and low density, typically representing areas that are either under-researched or declining in influence. “Artificial Immune Systems” is one such theme. AIS are computational systems inspired by the principles and processes of biological immune systems. They are advantageous in mimicking immune mechanisms such as pathogen recognition, memory, and adaptive learning. While AIS was the subject of significant research from 2007 to 2020, its prominence and importance in the field have notably declined in recent years.

To further delineate the evolving themes within the field, we analyzed keywords that appeared more than eight times and divided the research into five phases, as illustrated in Figure 9: (1) Initial Phase (1986–2016): Research focused on topics such as “asthma,” “artificial immune systems,” “machine learning,” “proteomics,” “immunohistochemistry,” and “electron microscopy.” Early studies primarily targeted specific diseases, such as asthma, while developing foundational AI technologies like AIS and ML. Techniques such as proteomics, immunohistochemistry, and electron microscopy were crucial for understanding the molecular and cellular foundations of immune responses. The use of AIS reflected a growing interest in applying bio-inspired algorithms to simulate immune function. This period laid the groundwork for integrating AI with immunology and allergy research, particularly in disease modeling and molecular analysis. (2) 2017–2020: During this period, themes such as “machine learning,” “deep learning,” “epidemiology,” “cancer,” “transcriptome,” “gene expression,” “artificial intelligence,” and “immunotherapy” gained prominence. Broader topics such as epidemiology, cancer, and gene expression analysis related to immunology and allergy witnessed a marked shift in AI towards ML and DL. As research in these fields advanced, it generated vast amounts of high-dimensional, heterogeneous data, leading researchers to utilize more complex DL models for the analysis of extensive medical datasets. During this phase, immunotherapy became increasingly important, with growing research on optimizing cancer treatments and other immune-modulating therapies using AI. (3) 2019–2020: Key research hotspots during this phase included “COVID-19,” “influenza,” “autoimmune diseases,” “bioinformatics,” “gut microbiota,” “convolutional neural networks,” and “artificial immune systems.” COVID-19 became a dominant research focus, reflecting the global health crisis of this period. Researchers rapidly employed AI to track viral transmission, predict outcomes, and accelerate vaccine development. Convolutional neural networks (CNNs) and bioinformatics tools further enhanced the ability to analyze the vast amounts of genomic and clinical data associated with COVID-19. Interest in the gut microbiome and autoimmune diseases also grew during this time. The demand for simulating immune responses brought AIS research back into the spotlight. (4) 2021–2022: Themes such as “anaphylaxis,” “cytokines,” “asthma,” “immunotherapy,” “psoriasis,” “prognosis,” “autoimmune diseases,” and “deep learning” captured the attention of researchers. The focus of research during this period shifted toward more specific immune-related diseases, such as anaphylaxis, asthma, and psoriasis. DL continued to play an important role in analyzing immune data, while cytokine analysis became crucial for understanding immune responses, especially in the aftermath of COVID-19, as global health challenges persisted. During this phase, researchers increasingly employed AI to predict patient outcomes and guide treatment strategies. (5) 2023–2024: More recent themes, such as “machine learning,” “artificial intelligence,” “deep learning,” “digital health,” “gut microbiome,” “inflammation,” and “immunoinformatics,” have garnered greater attention. AI technologies like ML and DL continue to be central to immunology and allergy research. The emergence of new focus areas, such as digital health, highlights the integration of AI with healthcare delivery systems. Immunoinformatics, which combines immunology with bioinformatics, is gaining traction, reflecting the growing need for advanced computational tools to analyze immune data. Overall, the evolution of AI-related keywords in immunology and allergy research indicates that the field is increasingly reliant on advanced computational tools to address complex biological and medical problems. Clinically, attention is shifting toward AI-driven healthcare systems, with AI poised to become a key driver of research in immune mechanisms, personalized medicine, and digital health innovations.

Currently, the application of AI in immunology and allergy is focused on three primary directions: clinical research, basic research, and the development of drugs and immunotherapies. The rapid emergence of large-scale, high-dimensional, heterogeneous, and complex data—including imaging, speech, text, and omics data—has significantly increased the complexity of research and clinical diagnostics. Advances in deep CNNs and artificial neural networks (ANNs) have revolutionized image recognition technology, driving progress in histopathological immunology research and enhancing the ability to detect disease biomarkers and assist in diagnostics (91–94). Recurrent neural networks (RNNs), long short-term memory networks (LSTMs), and natural language processing (NLP) have dramatically improved the analysis of medical speech and text sequences. When combined with electronic health records (EHRs), these technologies further enhance early disease identification and management (95–99). Random forest algorithms and support vector machines (SVMs) have strengthened the ability to predict disease risk and progression based on patient histories, environmental factors, and genetic data (100–105). Meanwhile, in patient management and prognosis, applications such as chatbots (106, 107), and wearable devices (108, 109) have greatly facilitated self-management, improving patients’ quality of life.

Overall, AI and biological neuroscience have fostered a mutually reinforcing cycle of inspiration. The biological structure and function of the human brain have inspired the development of neural network models in AI (110, 111). These models, in turn, have been used to model and understand brain function and have further contributed to the prediction of factors, outcomes, and drug development for various immune diseases (112, 113). Metacognitive computing systems that simulate human thought processes provide insights into the development of more adaptive AI systems (114, 115). AI, particularly ML and DL, shows enormous promise in medical data analysis, aiding clinical decision-making, predicting diseases, improving patient care and experience, increasing healthcare system efficiency, reducing medical costs, and advancing medical education (2, 3, 116, 117). Advances in algorithms and training techniques, such as transfer learning, backpropagation, and dropout techniques, along with the expansion of computational resources (e.g., GPUs and cloud computing) and the development of genomic sequencing technologies and EHR systems, have driven the application of AI in allergy and immunology. As AI continues to evolve, it is increasingly capable of handling large-scale medical data, while advances in allergy and immunology provide more reliable datasets for AI applications. This mutual reinforcement has made AI an increasingly critical tool in allergy and immunology.

Additionally, AI-driven drug development has transformed immunology by accelerating the discovery of novel immunotherapies, optimizing vaccine design, and improving treatment personalization (112, 118, 119). ML and DL models are increasingly used to predict neoantigen-based immunotherapies, facilitating the development of personalized cancer vaccines by identifying highly immunogenic tumor epitopes (119). AI-enhanced bioinformatics has also contributed to the design of hypoallergenic derivatives, where ML models analyze allergen structures, modify epitopes to reduce immunogenicity, and ensure therapeutic efficacy, as seen in recent computational approaches to food allergy treatments (120, 121). Moreover, AI in epitope prediction enables precise identification of T-cell and B-cell epitopes, utilizing CNNs and generative adversarial networks (GANs) to model antigen–antibody interactions and predict immune responses more accurately than traditional methods (122, 123). AI-driven molecular docking further enhances therapeutic development by rapidly screening potential drug candidates, predicting their binding affinities to immune targets, and optimizing molecular modifications to improve efficacy and reduce side effects (124). Beyond drug discovery, AI significantly impacts allergy and immunology by improving prediction and prevention, where AI models forecast allergic reactions based on genomic, environmental, and clinical data, enabling early intervention (105); optimizing hospital operations through predictive analytics for immunotherapy scheduling and automated triage systems (125); enhancing patient care with AI-powered virtual assistants, wearable biosensors, and personalized treatment planning (126); and advancing diagnostics via deep learning-assisted imaging, immunological biomarker discovery, and AI-enhanced allergen testing to increase diagnostic accuracy (127). As AI continues to evolve, integrating these innovations into clinical practice will revolutionize immunological research, making therapies more effective, accessible, and personalized.

The development and application of AI in the fields of immunology and allergy are far from straightforward and face numerous challenges that must be addressed for AI to achieve its full potential. One of the fundamental issues is the quality of the data on which AI models rely. The datasets trained on these models inherently determine their performance. Given the intricate interplay of genetic, environmental, and lifestyle factors, collecting and standardizing large volumes of high-quality data to model human immune responses is no small feat. Poor data quality can introduce bias into AI-generated outputs, while insufficient data volumes can render AI models unreliable (31). The complexity of data collection, storage, and annotation further compounds the difficulty, significantly increasing the time and resource burden on researchers. Inconsistent data sets not only hamper the training of robust AI models but also limit their generalizability across different populations and clinical settings. Moreover, much of the training data currently comes from countries such as the USA, China, Australia, Japan, India, and European nations. This geographical concentration can lead to biased AI models, which may exhibit reduced effectiveness or even provide inaccurate results when applied to populations in other regions. The global applicability of AI models remains a significant challenge, particularly when attempting to deploy these models in resource-limited settings.

Another critical issue in this field is the lack of publicly accessible, high-quality datasets. Due to concerns over patient privacy, a lack of data-sharing infrastructure, and varying policies, the availability of comprehensive datasets remains limited, which in turn restricts the advancement of AI technologies in allergy and immunology. The future success of AI in these domains depends heavily on the creation of secure, shared data environments where researchers can collaborate and build more generalizable models.

Furthermore, the rapid advancement of AI technology has led to the development of increasingly complex models, which often function as “black boxes” (128). This lack of interpretability is a significant barrier to the clinical application of AI, as clinicians may struggle to understand the logic behind AI-driven recommendations (129, 130). When the reasoning process of an AI system is opaque, it can erode trust among healthcare providers and patients alike, making it difficult to integrate AI into routine clinical workflows. This raises a critical question: How should medical education systems evolve to prepare healthcare professionals to work effectively with AI tools? Integrating AI education and training into medical curricula will be essential to bridging this gap and ensuring that AI becomes a trusted component of clinical practice.

From an ethical perspective, the integration of AI in medical research and practice brings new challenges. The use of AI to analyze large-scale data raises concerns about patient privacy and the potential exposure of sensitive health information. The possibility of training AI models on non-representative datasets, driven by profit motives, could result in biased and inequitable outcomes. These ethical concerns necessitate the development of robust regulatory frameworks that keep pace with the rapid evolution of AI technologies. Ensuring that AI applications in immunology and allergy are ethically sound, transparent, and accountable will be critical to their long-term success.

Given the advances in immunology, allergology, data science, and computational resources, coupled with the widespread success of AI in radiology, oncology, and dermatology, the future of AI in allergy and immunology holds the promise of transformative growth. These pivotal experiences provide a roadmap for AI’s application and development in the field of allergy and immunology, particularly in areas such as improving diagnostic accuracy of allergic conditions through image recognition, enhancing personalized treatment plans using predictive models, accelerating the discovery of allergy therapies through AI-driven drug development, analyzing patterns of allergic reactions, and creating risk prediction tools to foresee severe allergic responses or asthma exacerbations.

The success of AI in these domains has been largely driven by collaboration among data scientists, clinicians, and biostatisticians. Similarly, advancing AI in allergy and immunology will depend on close interdisciplinary collaboration, offering diverse perspectives that drive rapid progress. This expansion will strengthen collaborative research efforts and propel the development and refinement of diagnostic tools, treatment strategies, and personalized healthcare systems.

However, AI’s application in radiology, oncology, and dermatology faces significant challenges, such as data collection, model transparency, generalizability of conclusions, and ethical and regulatory concerns. These obstacles provide valuable lessons for AI’s future development in allergy and immunology.

Considering the current state of AI research and development in allergy and immunology, as well as successes and challenges in other medical fields, we anticipate four core areas where AI will impact allergy and immunology:

Moreover, in the face of growing AI, a number of other questions arise, such as whether a reduction in interpersonal interactions will increase the frequency of psychological problems (131). Over-reliance on AI may weaken the patient-physician relationship, leading to a decline in empathy and patient satisfaction. Personal interactions between patients and healthcare providers are critical to effective care, and AI should augment, not replace, this dynamic. At the same time, relying on AI for decision-making may weaken the critical thinking skills of students and clinicians (132, 133), and may lead to diagnostic or treatment errors if the AI system malfunctions or provides incorrect advice. This set of issues deserves to be explored in depth. Overall, addressing these risks requires adequate research, developing ethical guidelines, implementing unbiased training datasets, ensuring transparency in the AI decision-making process, and establishing robust data protection regulations. By proactively addressing these challenges, the healthcare industry can fully utilize the potential of AI while safeguarding patient rights and maintaining the integrity of clinical practice.