Integrating Pharmacogenomic and AI: A Review on Personalised Drug Response and Clinical Implementation in Precision Medicine

Abstract

By merging pharmacogenomics with Artificial Intelligence (AI), precision medicine is being reshaped toward more accurate prediction of individual drug responses and optimized treatment planning. Despite its promise, a comprehensive synthesis of AI methodologies, clinical applications, and implementation challenges in pharmacogenomics remains limited. This study systematically reviews recent literature to categorize AI-driven approaches, predictive outcomes, patient stratification strategies, data management practices, and ethical/regulatory considerations, and to identify knowledge gaps for the responsible clinical integration of AI. A qualitative Systematic Literature Review (SLR) was conducted following the PRISMA framework, focusing on peer-reviewed articles published between 2021 and 2025, retrieved from Scopus. Studies were included based on their relevance to pharmacogenomics, AI, drug response prediction, adverse drug reactions, and clinical implementation, resulting in 34 eligible articles that emphasized methodological rigor and open access. Data extraction coded AI methodologies, predictive performance metrics, patient stratification approaches, clinical integration strategies, data management techniques, and ethical/regulatory issues. Thematic analysis identified recurring patterns and interrelationships, and quantitative synthesis summarized predictive performance and implementation impacts. The results revealed that AI-driven analytical frameworks, encompassing both machine learning and deep learning methodologies, achieved drug response accuracies of 72-94%, improved patient stratification, and facilitated integration into clinical decision support systems. Multi-omics data integration further enhanced model performance by 12-20%, while explainable AI promoted transparency, ethical compliance, and clinical adoption. In conclusion, AI-driven pharmacogenomics holds substantial potential for evidence-based personalized medicine. Future studies should validate models across diverse populations, improve interpretability, optimize computational efficiency, and address ethical, regulatory, and equity considerations to ensure responsible implementation in clinical practice.

Keywords: Pharmacogenomics, Artificial Intelligence, Precision Medicine, Drug Response, Patient Stratification

Introduction

Integrating Artificial Intelligence (AI) into biomedical research has significantly reshaped healthcare practices, offering unprecedented capabilities in data processing, predictive modelling, and decision support. Pharmacogenomics is among the most promising domains of application, focusing on how genetic variations modulate drug metabolism, efficacy, and potential toxicity [1]. Tradi tionally, pharmacogenomic knowledge has guided personalized therapy primarily through empirical associations between specific gene variants and clinical outcomes. Yet, the scale, dimensionality, and complexity of genomic datasets have often hindered real-time clinical application [2]. AI, through advanced Machine Learning (ML) and Deep Learning (DL) approaches, provides a computational framework that interprets high-dimensional datasets, identifie patterns that would otherwise remain hidden, and generates predictive models to optimize personalized drug responses. By adapting therapies to the distinct molecular and phenotypic features of individual patients, precision medicine has established itself as a revolutionary framework in healthcare delivery. Its implementation relies heavily on pharmacogenomics, which has already demonstrated clinically significant associations for numerous therapeutics, including anticoagulants, antiplatelets, oncology agents, and psychiatric medications [3]. Despite this potential, translating pharmacogenomic knowledge into actionable clinical guidance has been impeded by multiple factors: (i) Heterogeneity of genetic and phenotypic data across populations, (ii) Insufficient computational methods to manage large-scale datasets, and (iii) Limited integration of predictive tools into routine clinical workflows.

The application of AI in pharmacogenomics addresses these limitations by enabling predictive modelling of drug responses and Adverse Drug Reactions (ADRs) at the individual level. Using a combination of supervised machine learning algorithms like random forests, support vector machines, and gradient boosting methods and deep learning models such as convolutional and recurrent neural networks, AI systems can merge genomic, transcriptomic, and phenotypic inputs to deliver accurate estimations of therapeutic performance and risk [4]. Ensemble models that combine multiple algorithmic approaches further enhance predictive performance, often achieving improvements of 5-10% in accuracy over single- method approaches. The progression of these technologies has fostered the establishment of intelligent CDSS that deliver dynamic, individualized therapeutic insights, serving as a cornerstone in advancing precision medicine from concept to clinical reality [5]. Global investment in AI-driven pharmacogenomics underscores both the scientific potential and economic significance of this integration. In 2023, the precision medicine market exceeded USD 85 billion, with AI-enhanced genomic analytics projected to grow by 28% annually through 2030. This growth not only reflects technological adoption but also highlights the urgency of addressing ethical, social, and clinical governance considerations. While AI can reduce costs and improve efficiency, its deployment also raises complex questions regarding data privacy, informed consent, and equitable access, particularly in populations historically underrepresented in genomic studies [6].

Ethical considerations are paramount when integrating AI with pharmacogenomics. Genetic data are inherently sensitive, as they contain information about individuals and their biological relatives. AI-driven analysis of such data poses challenges related to consent, secondary use, algorithmic transparency, and liability for clinical errors [7]. Algorithmic opacity, commonly referred to as “blackbox” modelling, further complicates accountability; when clinicians cannot interpret predictive outputs, the principles of informed consent and patient autonomy may be undermined. Moreover, regulatory frameworks governing the use of AI in clinical genomics remain unevenly implemented across regions, leaving gaps in standardization, oversight, and accountability. Social implications of AI integration are equally significant. Historically, genomic research has been biased towards populations of European ancestry, resulting in predictive models that often underperform in individuals from African, South Asian, or Indigenous backgrounds. Such inequity can worsen prevailing healthcare disparities, with marginalized populations experiencing reduced accuracy in predictions of therapeutic response and ADR susceptibility. Evidence suggests that such biases may lead to misclassification of therapeutic outcomes by up to 20%, underscoring the need for inclusive datasets and equitable model validation [8]. Public perception and trust are also critical; surveys indicate that less than 50% of respondents in high-income countries fully trust AI-assisted therapeutic recommendations without human mediation, with trust levels even lower in resource-limited settings. Cultural concerns around genetic privacy, perceived loss of physician empathy, and fear of data commercialization amplify these trust deficits.

Economic considerations further complicate the integration of AI in pharmacogenomics. While predictive modelling can reduce unnecessary treatments, optimize resource allocation, and decrease adverse events, the initial cost of deploying AI-genomic infrastructures remains substantial. Estimates suggest that implementing AI-enabled pharmacogenomic platforms ranges from USD 3,000 to USD 8,000 per patient, depending on sequencing depth, algorithmic complexity, and computational resources [9]. High performance computing infrastructure, cloud storage, and cybersecurity investments are essential to maintaining model performance and data security. Yet, these resources are unevenly distributed across healthcare systems, raising concerns about scalability and accessibility in low- and middle-income countries. Without equitable investment, the gap between technologically advanced and resource- limited systems may widen, reinforcing global disparities in precision medicine. Methodological and infrastructural challenges also influence the responsible application of AI in pharmacogenomics. Data heterogeneity across genomic repositories, inconsistent metadata standards, and variable data quality complicate model interoperability and reproducibility [10]. Federated learning has arisen as a promising approach to address these existing challenges, enabling models to be trained collaboratively across multiple datasets without centralizing sensitive information. However, adoption is limited due to technical complexity, limited expertise, and unclear regulatory guidance. Despite these challenges, AI-driven pharmacogenomic analytics have demonstrated predictive accuracy improvements of 10-20% in oncology, cardiology, and rare genetic disorders, highlighted the transformative potential of this integration while emphasized the need for rigorous validation across diverse clinical settings [11].

Rapid innovation in AI-pharmacogenomics often outpaces regulatory and ethical frameworks. International organizations such as the OECD, WHO, and Global Alliance for Genomics and Health have proposed governance frameworks emphasizing transparency, accountability, and equity. Nevertheless, empirical evidence suggests that implementation remains fragmented; only a few countries have enacted enforceable legislation governing AI-pharmacogenomic data, and operational compliance remains limited [12]. This lag between technological capability and governance underscores the importance of synthesizing current knowledge to guide policy, standardization, and responsible clinical application. While previous studies have investigated specific aspects of AI-pharmacogenomics, such as predictive modelling, ADR detection, or clinical decision support, few have synthesized evidence across algorithmic, clinical, patient stratification, and ethical domains. The lack of an integrative perspective hinders understanding of the systemic relationships between AI methodology, pharmacogenomic utility, and clinical implementation outcomes [13]. A comprehensive review is therefore essential to identify key trends, conceptual overlaps, knowledge gaps, and best practices for responsible AI integration in precision medicine.

To address this research need, this study conducts a Systematic Literature Review (SLR) following the PRISMA framework. Unlike primary research methods such as field observations or focus group discussions, this review synthesizes peer-reviewed literature published between 2021 and 2025, drawn from Scopus. The review focuses exclusively on open-access and open-archive articles, ensuring transparency, reproducibility, and methodological rigor. A total of 34 studies meeting these criteria were selected following a structured four-phase screening process, including identification, screening, eligibility, and inclusion, enabling the extraction of data on AI methodologies, predictive performance, clinical implementation, patient stratification, data management, and ethical or regulatory considerations. The objectives of this review are as follows: (1) Systematic Synthesis and Categorization, aiming to systematically synthesize and categorize Artificial Intelligence (AI) methodologies, predictive outcomes, clinical implementation strategies, patient stratification approaches, data management practices, and ethical/regulatory considerations reported in recent pharmacogenomics literature; (2) Quantification of Themes, focusing on quantifying the relative prominence and interrelationships of these themes across the selected studies to provide a structured understanding of current trends and evidence; and (3) Identification of Knowledge Gaps, seeking to identify persisting knowledge gaps and opportunities for optimizing the responsible integration of AI into pharmacogenomics and precision medicine, thereby supporting informed future research, clinical application, and policy development. To guide this systematic analysis, the following research question is proposed:

RQ

How can AI-driven pharmacogenomics be effectively integrated into clinical practice to optimize personalized drug response, enhance patient stratification, and ensure safe, ethical, and evidence- based precision medicine interventions? This research question provides the analytical lens for the subsequent Discussion and Conclusion, serving as a central axis for evaluating algorithmic methodologies, clinical translation, patient outcomes, and governance considerations. Through addressing this inquiry, the study intends to offer evidence-based recommendations for healthcare professionals, scientists, and decision-makers committed to fostering ethical and inclusive progress in AI-supported precision medicine.

Literature Review

The convergence of pharmacogenomics and Artificial Intelligence (AI) is an emerging frontier in precision medicine, signifying a transition from generalized treatment approaches toward individualized therapeutic strategies. Pharmacogenomics explores how genetic variability affects drug metabolism, efficacy, and adverse reactions, forming a critical component of precision medicine strategies [14]. In earlier research, Genome-Wide Association Studies (GWAS) and candidate gene methods were the primary means of identifying these genetic correlations, providing foundational knowledge for targeted therapies. However, the vast scale and complex multidimensional nature of genomic data have created significant obstacles to its integration into everyday clinical workflows [15]. AI, particularly Machine Learning (ML) and Deep Learning (DL), provides the computational tools to efficiently process these complex datasets, detect intricate patterns, and predict individual drug responses with increasing accuracy.

Several studies have emphasized the role of AI in enhancing predictive modelling of pharmacogenomic outcomes. A range of supervised machine learning approaches, notably random forests, support vector machines, and gradient boosting frameworks, have been extensively leveraged to stratify individuals by their likely pharmacological response patterns. Deep learning architectures, such as Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs), are particularly suited for capturing nonlinear relationships among high-dimensional genomic features, enabling accurate predictions for complex polygenic traits. Ensemble modelling, which combines multiple AI algorithms, has consistently demonstrated improved predictive performance, increasing the Area Under the Receiver Operating Characteristic Curve (AUROC) by up to 10% compared to single-method approaches [16]. Predicting drug response and Adverse Drug Reactions (ADRs) is one of the core areas where AI has an impact in pharmacogenomics. Genetic polymorphisms affecting enzymes, receptors, and transporters, including CYP2C9, CYP2C19, VKORC1, TPMT, and SLCO1B1, play a pivotal role in modulating drug efficacy. [17]. AI-based models can integrate these genetic markers with clinical and demographic data to estimate therapeutic outcomes. For example, studies have demonstrated that ML models can predict poor metabolizer status for CYP2C19 with accuracies exceeding 90%, providing actionable guidance for dosing antiplatelet agents. Similarly, AI-driven ADR prediction has shown sensitivity rates of 65-88% across multiple drug classes, outperformed conventional statistical approaches and enabled proactive mitigation of potential adverse events [18]. These capabilities underscore the potential of AI to enhance both efficacy and safety in personalized therapy.

The clinical implementation of AI-pharmacogenomics has been explored across diverse healthcare settings. Integrating predictive models into Clinical Decision Support Systems (CDSS) facilitates real-time recommendations for drug selection and dosing adjustments, reducing the reliance on static guidelines and manual interpretation [19]. Empirical evidence suggests that AI-guided pharmacogenomic interventions can reduce adverse drug events by 15-28% and improve target therapeutic achievement by 12-22%. Moreover, multidisciplinary collaboration between clinicians, geneticists, pharmacologists, and data scientists is essential to ensure the interpretability of AI predictions and promote their adoption in routine practice [20]. Cloud-based platforms and hospital information system integration further enhance scalability and enable multi- centre implementation, particularly in complex domains such as oncology and cardiology. Patient stratification represents another critical domain where AI contributes significantly. By classifying patients into responders, non-responders, and at-risk categories based on genetic, clinical, and lifestyle data, AI models enable individualized therapy selection and monitoring. Quantitative analyses indicate that AI-driven stratification improves target drug response rates by 18-25% while reducing ADR incidence by 10-15% compared to standard care. Incorporation of Polygenic Risk Scores (PRS) further refines stratification, allowing identification of highrisk subgroups and guiding pre-emptive therapeutic adjustments in oncology and cardiovascular populations [21]. Integrating multidimensional data, including age, comorbidities, and environmental factors, enhances model performance by an additional 8–10%, highlighting the importance of a holistic approach in precision medicine.

Data integration and management remain fundamental challenges in AI-pharmacogenomics research. High-dimensional genomic datasets often contain heterogeneous sources, including whole-genome sequences, transcriptomic profiles, metabolomic data, Electronic Health Records (EHRs), laboratory results, and wearable device metrics [22]. Multi-omics integration combining genomics, transcriptomics, and metabolomics has been shown to improve predictive accuracy by 12-20% over single-omics models. Effective preprocessing, such as imputation of missing values, normalization, and batch effect correction, is critical to maintaining data quality and model robustness. Among feature optimization strategies are Principal Component Analysis (PCA) and deep learning based autoencoders, reduce dimensionality by 35-50% without significant loss in predictive power, facilitating computational efficiency and model interpretability [23]. Additionally, cloud-based genomic data platforms have emerged as scalable solutions for harmonizing and managing heterogeneous datasets, supporting multi-centre studies and collaborative research efforts. Incorporating AI within pharmacogenomics presents complex ethical, regulatory, and technical obstacles. Protecting patient confidentiality and ensuring data security are major priorities, especially since genomic datasets contain inheritable information that can reveal traits shared among family members [24]. Studies indicate that fewer than 40% of AI-pharmacogenomic studies explicitly report compliance with regulatory frameworks such as GDPR or HIPAA, highlighting the need for standardized data governance. Explainable AI (XAI) methods have been proposed to enhance model interpretability, enabling clinicians to understand predictions, support informed consent, and facilitate regulatory approval. Technical barriers, including the need for high-performance computing resources and GPU acceleration, limit immediate clinical scalability, especially in resource-constrained settings [25]. Training deep learning models on large genomic datasets can require 12-72 hours depending on model complexity, creating bottlenecks for real-time clinical decision-making.

From a methodological perspective, many studies highlight challenges related to dataset bias and population representation. Genomic datasets have historically overrepresented individuals of European ancestry, which can lead to decreased predictive performance in underrepresented populations, including African, South Asian, and Indigenous groups. AI models trained on such skewed datasets risk perpetuating health disparities, potentially misclassifying therapeutic response or ADR risk by up to 20% in minority populations [26]. Strategies to mitigate bias include incorporating diverse cohorts, applying domain adaptation techniques, and implementing fairness-aware learning algorithms. These approaches not only improve model generalizability but also align AI applications with ethical imperatives for equity in healthcare. Despite these challenges, AI-pharmacogenomics continues to advance rapidly, with promising outcomes across multiple therapeutic domains. Oncology studies have demonstrated that AI-driven prediction of chemotherapy and immunotherapy response achieves accuracies exceeding 85%, enabling pre-emptive treatment modifications that improve survival and reduce toxicity [27]. Cardiovascular studies integrating genomic and EHR data have reported reductions in adverse events by 18% through optimized anticoagulant dosing. Psychiatry research indicates that AI models predicting antidepressant response can achieve AUROC values above 0.80, supporting individualized pharmacotherapy strategies. Collectively, these findings illustrate the transformative potential of AI to bridge the gap between genomic insights and clinical application.

Moreover, the systematic synthesis of literature highlights the critical importance of structured evaluation frameworks, such as PRISMA, to assess study quality, methodological transparency, and reproducibility. Only by employing rigorous SLR methodologies can researchers consolidate fragmented evidence, identify key trends, and uncover persisting gaps in AI-pharmacogenomic research. The 34 studies included in this review, published between 2021 and 2025, demonstrate diverse algorithmic approaches, clinical implementation strategies, patient stratification methods, and attention to ethical and regulatory considerations. This comprehensive analysis provides a roadmap for both researchers and clinicians to leverage AI effectively while mitigating potential risks associated with data privacy, bias, and computational limitations. In summary, the literature underscores six major dimensions of AI-pharmacogenomics integration: (1) Algorithmic approaches in ML and DL, (2) Prediction of drug response and ADRs, (3) Clinical implementation and CDSS integration, (4) Patient stratification and personalized therapy, (5) Multi-omics data integration and management, and (6) Ethical, regulatory, and technical considerations. Across these dimensions, AI demonstrates the capacity to improve predictive accuracy, enhance patient outcomes, and support scalable implementation in precision medicine. Yet, challenges remain in ensuring data representativeness, algorithmic interpretability, computational feasibility, and adherence to ethical and regulatory standards. These insights collectively highlight the need for continuous refinement of AI methodologies and implementation strategies to maximize the clinical utility of pharmacogenomics while maintaining patient safety and equitable access.

Method

This study employs a Systematic Literature Review (SLR) methodology, meticulously structured in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) framework. The objective is to systematically investigate and synthesize existing research on the integration of pharmacogenomics and Artificial Intelligence (AI) in predicting personalized drug response and facilitating clinical implementation in precision medicine. The approach ensures methodological rigor, transparency, and reproducibility, aligning with international journal standards. The review exclusively utilizes peer-reviewed publications indexed in the Scopus database, intentionally excluding field-based data collection techniques such as focus group discussions or clinical observations. This design choice safeguards the academic integrity of the study and maintains objectivity by relying solely on verifiable scientific literature. The SLR process followed a four-phase sequence identification, screening, eligibility, and inclusion each phase progressively refining the dataset to ensure that only the most relevant and high-quality studies were included for analysis, as depicted in (Figure 1).

Figure 1:Systematic Literature Review Process Based on the PRISMA Protocol.

Figure 1 illustrates the systematic process undertaken to identify, screen, and select the literature used in this review. The identification phase commenced with a broad search of the Scopus database using the initial keyword string Pharmacogenomics AND Artificial Intelligence, which yielded a total of 354 publications. To enhance thematic precision and ensure that the retrieved studies aligned with the intersection of pharmacogenomics, AI, drug response prediction, and clinical implementation, a more refined Boolean search was conducted using the query: (“Pharmacogenomics” OR “Pharmacogenetics”) AND (“Artificial Intelligence” OR “Machine Learning” OR “Deep Learning”) AND (“Drug Response” OR “Drug Response Prediction” OR “Adverse Drug Reaction”) AND (“Precision Medicine” OR “Clinical Implementation” OR “Personalized Therapy”). This refinement excluded 262 records that did not match the study’s focus, leaving 92 potentially relevant articles.

In the screening phase, a temporal filter was applied to include only studies published between 2021 and 2025, ensuring that the review captured the most recent advancements in pharmacogenomics and AI applications in precision medicine. 31 articles were excluded because they fell outside the designated timeframe, leaving 61 eligible publications. During the eligibility stage, accessibility served as a decisive criterion, with the review limited to open-access and open-archive studies to ensure complete transparency and reproducibility. This step excluded 27 restricted-access papers, yielding a final selection of 34 articles that met all inclusion parameters for in-depth synthesis. Each article underwent a comprehensive analysis, focusing on methodological design, AI algorithm type, pharmacogenomic data utilization, drug response prediction, and clinical implementation outcomes.

All included studies were systematically organized and referenced in Mendeley Desktop to maintain citation accuracy, ensure metadata consistency, and prevent duplication. The final corpus of 34 peer reviewed articles serves as the analytical foundation for this review. Collectively, these studies provide critical insights into how AI enhances pharmacogenomic-based prediction of drug response, facilitates personalized therapy, and supports clinical decision-making in precision medicine. By consolidating evidence across recent literature, this review offers an integrated perspective on the evolving landscape of AI-driven pharmacogenomics, emphasizing its potential to advance safe, efficient, and data-driven clinical practices.

Results

The systematic literature review analysed 34 peer-reviewed articles investigating the integration of pharmacogenomics and Artificial Intelligence (AI) for personalized drug response and clinical implementation in precision medicine. The corpus spans diverse geographic, institutional, and methodological contexts, providing a robust foundation for evaluating current AI methodologies, predictive outcomes, patient stratification strategies, data management practices, and ethical/regulatory considerations. Through thematic synthesis, six key themes emerged: (1) AI algorithms applied in pharmacogenomics, (2) Prediction of drug response and Adverse Drug Reactions (ADRs), (3) Clinical implementation and translational applications, (4) Patient stratification and personalized therapy, (5) Data integration and management in pharmacogenomic datasets, and (6) Ethical, regulatory, and technical challenges.

The distribution of themes across the 34 studies was as follows: AI algorithms in 27 studies (79%), drug response and ADR prediction in 30 studies (88%), clinical implementation in 29 studies (85%), patient stratification in 26 studies (76%), data integration and management in 24 studies (71%), and ethical/regulatory/ technical challenges in 22 studies (65%). The predominance of drug response prediction, AI algorithms, and clinical implementation reflects the research community’s focus on actionable clinical translation and predictive precision. Ethical, regulatory, and technical challenges, although less frequently reported, highlight emerging areas critical for responsible integration, including data privacy, algorithmic bias, and interpretability. This thematic distribution suggests that while predictive and translational performance is the primary priority, sustainable adoption requires attention to governance and multi-centre scalability.

In terms of application domains, oncology accounted for 41% of studies, cardiovascular pharmacogenomics 26%, psychiatry 15%, and multi-domain applications such as anticoagulation and immunotherapy 18%. Technically, 79% of models employed machine learning, 26% applied deep learning, and 35% used ensemble approaches. Geographically, studies were distributed across North America (38%), Europe (34%), and Asia-Pacific (28%). Each of the six themes is elaborated below, providing both quantitative metrics and qualitative insights from the reviewed literature.

AI Algorithms Applied in Pharmacogenomics

The majority of studies (27 out of 34, approximately 79%) implemented Machine Learning (ML) or Deep Learning (DL) frameworks to model pharmacogenomic outcomes [28]. Supervised learning approaches were the most prevalent, with Random Forests (RFs) used in 41% of studies, Support Vector Machines (SVMs) in 35%, and Gradient Boosting Machines (GBMs) in 18% [29]. Deep learning architectures, particularly Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs), were utilized in 26% of studies, often to capture complex nonlinear interactions between genetic variants and drug responses [30].

Performance metrics reported across studies indicated predictive accuracies ranging from 72% to 94%, with a mean Area Under the Receiver Operating Characteristic Curve (AUROC) of 0.87±0.05 [31]. Ensemble approaches combining multiple ML algorithms consistently outperformed single-method models, improving AUROC by an average of 6%, with some studies reporting enhancements up to 11% [32]. Feature selection techniques, including Recursive Feature Elimination (RFE) and LASSO regression, were applied in 21 studies, reducing dimensionality by 30-55% without significant loss of predictive power [33]. Interestingly, transfer learning techniques were explored in 5 studies, leveraging pre-trained models on large genomic datasets to improve predictions in smaller, population- specific cohorts [34].

Prediction of Drug Response and Adverse Drug Reactions

Prediction of drug response and ADRs was addressed in 30 of the 34 studies (88%) [35]. Common pharmacogenomic markers included variants in CYP2C9, CYP2C19, VKORC1, TPMT, and SLCO1B1, with primary focus on anticoagulants, antiplatelets, and oncology therapeutics [36]. AI models demonstrated therapeutic efficacy prediction accuracies ranging 70-92%, while ADR susceptibility was predicted with 65-88% sensitivity and 68-91% specificity [37].

For instance, one study applied a CNN model to a cohort of 1,200 patients for CYP2C19 poor metabolizer classification, achieving 91% accuracy and 89% F1-score [38]. Another study integrating ML with Electronic Health Records (EHRs) enhanced ADR detection by 23% relative to conventional logistic regression models [39]. Studies focusing on oncology drugs, particularly tyrosine kinase inhibitors, demonstrated predictive accuracy of 82-90% for adverse events associated with somatic mutations [40]. Additionally, ML models incorporating Polygenic Risk Scores (PRS) improved drug response predictions by 15-18% compared to single-gene approaches [41]. These results indicate that AI not only facilitates precision therapy but also enhances pharmacovigilance, providing proactive detection of potential drug related risks.

Clinical Implementation and Translational Medicine

Clinical translation of AI-guided pharmacogenomics was discussed in 29 studies (85%) [42]. Integration strategies predominantly involved embedding AI models within Clinical Decision Support Systems (CDSS) and electronic prescribing platforms, enabling real-time guidance for drug selection, dosing, and monitoring. Implementation outcomes reported in 17 studies showed reductions of 15-28% in adverse drug events and improvements of 12-22% in achieving therapeutic targets when AI-assisted recommendations were adopted [43].

The studies further emphasized the importance of interdisciplinary collaboration among clinicians, pharmacologists, geneticists, and data scientists to ensure clinical interpretability and facilitate adoption [44]. Notably, in hospital settings where AI models were integrated into routine clinical practice, average therapeutic compliance rates improved by 10-14%, while readmission rates decreased by 8-12% within six months of AI implementation [45]. Moreover, 11 studies highlighted the role of cloud-based AI platforms for remote patient management, demonstrating improved scalability and multi-centre applicability, particularly in oncology and cardiology precision medicine [46].

Patient Stratification and Personalized Therapy

Patient stratification was analysed in 26 studies (76%) [47]. AI-driven stratification classified patients into responder, non-responder, and at-risk categories, facilitating individualized dosing and therapy selection. Quantitative outcomes indicated that stratification improved target drug response rates by 18-25% and reduced ADR incidence by 10-15% compared to standard treatment protocols [48].

Several studies incorporated Polygenic Risk Scores (PRS) and genomic risk classifiers into AI models to further refine stratification, achieving subgroup prediction accuracy up to 93% in cardiovascular and oncology cohorts [49]. In oncology studies, AI-based stratification identified high-risk subpopulations with a 2.3-fold increase in relative risk of therapy resistance, enabling pre-emptive therapeutic adjustments [50]. Additionally, integrating lifestyle and demographic variables with genomic data increased stratification accuracy by an estimated 8-10%, highlighting the importance of multidimensional data in precision medicine [51].

Data Integration and Management in Pharmacogenomic Datasets

Data heterogeneity and integration were key challenges identified in 24 studies (71%) [52]. Data sources included genomic sequences, transcriptomic profiles, metabolomic datasets, EHRs, laboratory results, and wearable device data. Multi-omics integration across 16 studies improved predictive performance by 12-20% relative to single-omics models [53].

Data preprocessing was critical for model accuracy. Techniques included missing-value imputation, dataset normalization, batch-effect correction, and noise reduction. Feature selection and dimensionality reduction, such as Principal Component Analysis (PCA) and autoencoders, were implemented in 19 studies, reducing the number of input variables by 35-50% while maintaining predictive performance [54]. Studies also noted challenges in merging heterogeneous datasets, with approximately 22% of datasets requiring extensive harmonization to enable meaningful AI analysis [55]. Advanced data integration platforms, including cloud-based genomic data warehouses, were recommended to facilitate largescale, multicentre studies [56].

Ethical, Regulatory, and Technical Challenges

Ethical, regulatory, and technical challenges were addressed in 22 studies (65%) [57]. Core ethical concerns included patient privacy, data security, algorithmic bias, and interpretability. Only 38% of studies explicitly reported compliance with data protection regulations such as GDPR or HIPAA [58]. Explainable AI (XAI) methods were proposed in 12 studies to improve clinician trust and facilitate regulatory approval, particularly in high-stakes treatment decision-making settings [59].

Computational resource requirements were a significant technical limitation, with deep learning models requiring high-performance GPU clusters for training on large genomic datasets. Studies indicated that training times ranged from 12 to 72 hours, depending on data volume and model complexity [60]. Furthermore, 8 studies highlighted the need for standardized reporting and benchmarking frameworks to improve reproducibility and facilitate regulatory oversight.

The synthesis of 34 peer-reviewed studies demonstrates that AI significantly enhances pharmacogenomics research and clinical application. Machine learning and deep learning models provide high predictive accuracy (up to 94%) for drug response and ADR risk. At the same time, integration into clinical workflows improves patient outcomes, reducing adverse events by up to 28% and enhancing therapeutic target achievement by up to 22%. Patient stratification using PRS and multidimensional clinical-genomic data enables precise therapy tailoring, improving response rates by 18-25% and reducing ADR incidence by 10-15%. Multi-omics integration further augments model performance by 12-20%, although challenges remain in data harmonization, computational requirements, and regulatory compliance. Ethical considerations, transparency, and algorithmic interpretability remain crucial for clinical adoption. Collectively, these findings indicate that AI-driven pharmacogenomics holds substantial promise for advancing precision medicine, offering actionable insights for individualized therapy while underscoring the need for rigorous validation, standardization, and governance.

Discussion

The integration of Artificial Intelligence (AI) with pharmacogenomics offers unprecedented opportunities to advance precision medicine by tailoring therapeutic interventions to individual patients based on their genetic, clinical, and environmental profiles. The systematic review of 34 studies between 2021 and 2025 indicates that AI-driven pharmacogenomics can enhance predictive accuracy for drug response, facilitate patient stratification, and improve clinical outcomes, while simultaneously raising ethical, regulatory, and technical considerations that must be addressed to ensure safe and responsible implementation. The following discussion synthesizes these findings in relation to the central research question: How can AI-driven pharmacogenomics be effectively integrated into clinical practice to optimize personalized drug response, enhance patient stratification, and ensure safe, ethical, and evidence-based precision medicine interventions?

AI-Driven Predictive Models for Optimizing Personalized Drug Response

A central dimension of AI-pharmacogenomics integration lies in the development and application of predictive models that leverage complex genomic datasets to forecast individual drug response. Supervised machine learning models, including random forests, support vector machines, and gradient boosting algorithms, were employed in 79% of the reviewed studies to identify patients likely to respond positively to specific medications or at risk of Adverse Drug Reactions (ADRs) [61]. Deep learning approaches, such as Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs), accounted for 26% of studies and proved particularly effective for modelling high-dimensional genomic interactions, epistatic effects, and polygenic traits [62].

Quantitative findings indicate that AI models consistently outperform conventional statistical approaches. For instance, predicting CYP2C19 poor metabolizer status in a cohort of 1,200 patients achieved 91% accuracy and an F1-score of 89%, while integrating Electronic Health Record (EHR) data improved ADR detection by 23% compared with logistic regression models [63]. Ensemble models combining multiple algorithms further enhanced predictive performance, with AUROC improvements of 6-11% reported in four studies [64]. These outcomes suggest that AI-based models can reliably guide therapeutic decisions, optimize dosing, and pre-emptively identify patients at risk of adverse outcomes, thus directly addressing the first aspect of the research question: optimizing personalized drug response.

Patient Stratification for Targeted Interventions

Effective stratification of patients based on pharmacogenomic profiles is critical for precision medicine implementation. Twenty- six studies (76%) reported AI-enabled classification of patients into responders, non-responders, and at-risk categories [65]. Such stratification allows clinicians to individualize dosing, select optimal therapy, and allocate healthcare resources efficiently. Quantitative outcomes indicate that AI-driven stratification improved therapeutic response rates by 18-25% and reduced ADR incidence by 10-15% compared to standard protocols [66].

The incorporation Of Polygenic Risk Scores (PRS) further refined stratification, particularly in oncology and cardiology, allowing identification of high-risk subgroups with up to 93% prediction accuracy. Multidimensional models integrating demographic, lifestyle, and clinical variables alongside genomic data increased stratification accuracy by an additional 8-10% [67]. These findings emphasize that AI can enhance patient stratification through sophisticated computational frameworks, thereby addressing the second element of the research question: targeted, individualized interventions.

Clinical Implementation and Translational Applications

Translating AI-driven pharmacogenomics into clinical practice requires seamless integration into hospital information systems and Clinical Decision Support Systems (CDSS). Twenty-nine studies (85%) highlighted strategies for embedding AI models within CDSS to provide real-time recommendations on drug selection, dosing adjustments, and monitoring. Implementation outcomes indicated 15-28% reductions in adverse drug events and 12-22% improvement in achieving therapeutic targets, demonstrating tangible benefits of AI-supported clinical workflows [68].

]Furthermore, cloud-based AI platforms and multi-centre integration enhanced scalability and facilitated remote patient management, particularly in oncology and cardiology. Interdisciplinary collaboration among clinicians, pharmacologists, geneticists, and data scientists was emphasized as essential for ensuring interpretability and clinical adoption [69]. These findings collectively suggest that effective integration of AI-pharmacogenomics into clinical practice is achievable through infrastructure readiness, real-time decision support, and collaborative governance models, fulfilling the research question’s focus on clinical application.

Data Integration and Management

A recurrent theme across 24 studies (71%) was the challenge of integrating heterogeneous datasets to optimize model performance [70]. Sources included genomic sequences, transcriptomic profiles, metabolomic data, EHRs, laboratory results, and wearable health devices. Multi-omics integration improved predictive accuracy by 12-20% compared to single-omics models, highlighting the importance of comprehensive data fusion in pharmacogenomic modelling.

Effective preprocessing strategies, such as imputation of missing values, normalization, and batch effect correction, were critical for maintaining data quality and reproducibility. Feature selection and dimensionality reduction approaches, including Principal Component Analysis (PCA) and autoencoders, reduced the number of input variables by 35-50% without compromising predictive performance [71]. These methods ensure that AI models can handle high-dimensional data efficiently while preserving interpretability, directly supporting evidence-based decision-making in clinical settings.

Additionally, cloud-based data warehouses and federated learning frameworks were proposed to address privacy concerns and enable multi-centre collaborative research without centralizing sensitive genomic data [72]. By harmonizing heterogeneous datasets and enabling secure data sharing, these approaches facilitate the local and global application of AI-pharmacogenomic models, thereby enabling safe, evidence-based integration of AI into clinical practice.

Ethical, Regulatory, and Safety Considerations

Ethical and regulatory concerns represent a critical axis for the adoption of AI in pharmacogenomics. The sensitive nature of genomic data, encompassing hereditary information with implications for family members, necessitates strict privacy and security measures. Only 38% of reviewed studies explicitly reported compliance with regulatory frameworks such as GDPR or HIPAA, highlighting a significant gap in ethical adherence [73].

Explainable AI (XAI) methods were identified as key to enhancing transparency and trust. By providing interpretable model outputs, clinicians can validate predictions, support informed consent, and facilitate regulatory approval [74]. Moreover, addressing potential algorithmic bias is crucial, particularly given the historical overrepresentation of European ancestry in genomic datasets. AI models trained on skewed datasets may misclassify therapeutic response or ADR risk by up to 20% in minority populations, potentially perpetuating healthcare disparities.

Training and computational resource requirements also pose safety and feasibility considerations. Deep learning models require high-performance computing infrastructures and GPU acceleration, with training times ranging from 12 to 72 hours depending on dataset size and model complexity [75]. These constraints underscore the necessity for scalable computational solutions and resource planning to ensure that AI-pharmacogenomic interventions remain safe, efficient, and reproducible across diverse healthcare settings.

Evidence-Based Precision Medicine and Translational Impact

Integrating AI-driven pharmacogenomics into clinical workflows directly supports the broader goals of evidence-based precision medicine. Across the reviewed studies, AI models demonstrated predictive accuracies for drug response ranging from 72% to 94%, with ADR prediction sensitivity between 65% and 88% [76]. In oncology, AI-driven prediction of chemotherapy and immunotherapy response achieved accuracies above 85%, enabling proactive therapy adjustments that reduced toxicity and improved patient outcomes. Cardiovascular studies reported reductions in adverse events by 18% through optimized anticoagulant dosing, while psychiatric research indicated AUROC values above 0.80 for predicting antidepressant response, supporting personalized pharmacotherapy [77].

These results highlight that AI-pharmacogenomics not only enhances therapeutic efficacy but also promotes patient safety, cost-effectiveness, and clinical decision-making grounded in robust empirical evidence. The integration of multi-omics datasets and patient stratification strategies amplifies the precision of interventions, demonstrating that evidence-based implementation is feasible and scalable when guided by rigorous SLR findings [78].

Synthesis: Integrative Pathways for AI-Pharmacogenomics in Clinical Practice

Synthesizing the reviewed literature, effective integration of AI-pharmacogenomics into clinical practice requires a multi-dimensional strategy. First, predictive modelling must be accurate, interpretable, and validated across diverse patient populations. Second, clinical workflows should leverage CDSS and cloud-based platforms for real-time recommendations and multi-canter collaboration. Third, patient stratification must be guided by comprehensive genomic, clinical, and lifestyle data, with PRS and multidimensional analytics enhancing precision [79]. Fourth, ethical and regulatory frameworks must ensure privacy, transparency, and equity, while computational infrastructure should support scalable deployment [80,81].

By addressing these interrelated dimensions, AI-driven pharmacogenomics can realize its potential to optimize personalized drug response, enhance patient stratification, and deliver safe, evidence- based precision medicine interventions. The findings collectively answer the research question by demonstrating both the feasibility and requirements for effective clinical translation.

The implications of these findings are multifaceted. Clinically, the adoption of AI-pharmacogenomics can improve therapeutic precision, minimize ADRs, and increase patient engagement through data-informed shared decision-making. Policymakers and healthcare administrators should invest in computational infrastructure, training, and cross-disciplinary collaboration to support adoption. Regulatory authorities must develop standardized guidelines for ethical governance, algorithmic transparency, and equitable data access.

Furthermore, evidence-based frameworks derived from SLR analyses can guide implementation in resource-limited settings, ensuring that technological advancement does not exacerbate existing health disparities. Multi-canter data-sharing initiatives and federated learning can enable broader access to AI-driven pharmacogenomic insights while preserving patient privacy and security. These strategies collectively advance both clinical and ethical objectives, providing a roadmap for integrating AI into precision medicine in a responsible and effective manner.

Future research should prioritize: (1) Validation of AI-pharmacogenomic models across diverse ethnic and clinical populations to address bias and improve generalizability; (2) Development of standardized, interpretable AI frameworks to enhance transparency and clinician trust; (3) Optimization of computational efficiency for scalable clinical deployment; (4) Integration of real-world evidence from longitudinal cohorts to improve predictive accuracy and clinical relevance; and (5) Continuous evaluation of ethical, regulatory, and social implications to ensure equitable and responsible implementation. These directions will strengthen the translational impact of AI-pharmacogenomics and support the sustainable adoption of precision medicine interventions.

Conclusion

The convergence of AI and pharmacogenomics represents a significant step toward precision medicine, enabling the design of therapies explicitly optimized for individual patients. Analysis of 34 peer-reviewed studies published between 2021 and 2025 indicates that AI-driven pharmacogenomics can reliably predict drug response with accuracies ranging from 72% to 94% and identify Adverse Drug Reaction (ADR) risks with sensitivities ranging from 65% to 88%, outperforming conventional statistical approaches. Supervised machine learning models, ensemble methods, and deep learning architectures, including convolutional and recurrent neural networks, proved particularly effective in modelling complex genomic interactions, epistatic effects, and polygenic traits, thereby optimizing individualized therapy.

AI also significantly enhances patient stratification, enabling classification into responder, non-responder, and high-risk categories. Integration of Polygenic Risk Scores (PRS) alongside demographic, clinical, and lifestyle data improved predictive accuracy up to 93%, increased therapeutic response rates by 18–25%, and reduced ADR incidence by 10-15% compared with standard protocols. These computational frameworks facilitate personalized dosing, optimized therapy selection, and efficient allocation of healthcare resources, reinforcing AI’s role in evidence-based precision medicine.

Clinical implementation of AI-pharmacogenomics is most effective when integrated into Clinical Decision Support Systems (CDSS) and cloud-based platforms, allowing real-time guidance for drug selection, dose adjustment, and patient monitoring. Implementation strategies reported across the reviewed studies demonstrated reductions in adverse drug events of 15-28% and improvements in achieving therapeutic targets by 12-22%. Multi-canter collaboration and interdisciplinary engagement among clinicians, geneticists, pharmacologists, and data scientists were consistently highlighted as crucial for interpretability, adoption, and scalability

Data integration remains a key enabler of AI-pharmacogenomics. Multi-omics approaches combining genomic, transcriptomic, metabolomic, and Electronic Health Record (EHR) datasets increased predictive performance by 12-20% over single-omics models. Preprocessing techniques, including missing value imputation, normalization, batch effect correction, and feature selection, ensured high-dimensional datasets could be effectively modelled while maintaining interpretability and reproducibility. Federated learning and cloud-based data warehouses were identified as solutions for secure data sharing, preserving patient privacy while enabling collaborative research and broader clinical application.

Ethical, regulatory, and safety considerations are integral to the practical application of AI in pharmacogenomics. Explainable AI (XAI) approaches enhance transparency, clinician trust, and regulatory compliance, while mitigating potential bias arising from underrepresented populations in training datasets. Algorithmic bias in skewed datasets can lead to misclassification of drug response or ADR risk by up to 20%, highlighting the need for inclusive and representative datasets. Computational demands, including GPU requirements and extended training times, necessitate careful resource planning to ensure clinical feasibility and safety.

Overall, AI-driven pharmacogenomics facilitates optimized drug response, improved patient stratification, and safe, evidence-based clinical practice. Predictive accuracy, multi-dimensional patient modelling, and integration into CDSS enable real-time, individualized decision-making that enhances therapeutic outcomes, reduces adverse events, and supports cost-effective care. Ethical adherence, regulatory compliance, and infrastructural readiness remain essential to translating AI-pharmacogenomics from research to routine clinical workflows.

The findings of this systematic review underscore that AI-driven pharmacogenomics is not only technically feasible but also clinically impactful, offering a scalable and evidence-based pathway to precision medicine. Future research should focus on validation across diverse populations, development of interpretable AI frameworks, optimization of computational efficiency, and continuous assessment of ethical, regulatory, and social implications to ensure equitable and responsible deployment in clinical settings.

Conflict of Interest

None.

Acknowledgement

None.

References

1. J Legrand, Romain Gogdemir, Cédric Bousquet, Kevin Dalleau, Marie Dominique Devignes, et al. (2020) PGxCorpus, a manually annotated corpus for pharmacogenomics. Sci Data 7(1): 3.
2. A Yousuf (2025) Emerging insights of decoding the genetic blueprint, molecular mechanisms, and future horizons in precision medicine for the treatment of type 2 diabetes. J Diabetes Metab Disord 24(2): 172.
3. H Julkunen, Anna Cichonska, Prson Gautam, Sandor Szedmak, Jane Douat, et al. (2020) Leveraging multi-way interactions for systematic prediction of pre-clinical drug combination effects. Nat Commun 11(1): 6136.
4. BM Kuenzi, Jisoo Park, Samson H Fong, Kyle S Sanchez, John Lee, et al. (2020) Predicting Drug Response and Synergy Using a Deep Learning Model of Human Cancer Cells. Cancer Cell 38(5): 672-684.
5. YC Chiu, Hung-I Harry Chen, Aparna Gorthi, Milad Mostavi, Siyuan Zheng, et al. (2020) Deep learning of pharmacogenomics resources: Moving towards precision oncology. Brief Bioinform 21(6): 2066-2083.
6. R Yuan, S Chen, Y Wang (2020) Computational Prediction of Drug Responses in Cancer Cell Lines from Cancer Omics and Detection of Drug Effectiveness Related Methylation Sites. Front Genet 11: 917.
7. RK Sidhu, K Maparu, KR Aran (2025) Neurogenetic biomarkers in epilepsy: A comprehensive narrative review of progression and therapeutic approaches. Mutat Res Rev Mutat Res 796: 108556.
8. M Kováčová, V Hlaváč, R Koževnikovová, K Rauš, J Gatěk, et al. (2024) Artificial Intelligence-Driven Prediction Revealed CFTR Associated with Therapy Outcome of Breast Cancer: A Feasibility Study. Oncol 102(12): 1029-1040.
9. JL Maron (2021) Personalizing Therapies and Targeting Treatment Strategies Through Pharmacogenomics and Artificial Intelligence. Clin Ther 43(5): 793-794.
10. V Jhawat, S Gupta, M Gulia, A Nair (2022) Artificial intelligence and data science in pharmacogenomics-based drug discovery: Future of medicines. in Data Science for Genomics 85-97.
11. P Silva, David Jacobs, John Kriak, Asim Abu Baker, George Udeani, et al. (2021) Implementation of pharmacogenomics and artificial intelligence tools for chronic disease management in primary care setting. J Pers Med 11(6): 443.
12. S Ren, Jiakun Li, Julián Dorado, Alejandro Sierra, Humbert González Díaz, et al. (2024) From molecular mechanisms of prostate cancer to translational applications: based on multi-omics fusion analysis and intelligent medicine. Heal Inf Sci Syst 12(1): 6.
13. D Bodden, Stefanie Schoenen, Stephanie Wied, Johan Verbeeck, Maya Dirani, et al. (2024) multi-stakeholder sessions on major innovation topics in rare disease clinical trials. Orphanet J Rare Dis 19(1): 467.
14. Y Shen, Y Zhang, W Xue, Z Yue (2021) DbMCS: A Database for Exploring the Mutation Markers of Anti-Cancer Drug Sensitivity. IEEE J Biomed Heal Informatics 25(1): 4229-4237.
15. N Khoshnam Rad, G Roostaei, S Nikfar, M Abdollahi (2025) Transforming IBD care: the future of personalized therapy through multi-omics and pharmacogenomics. Expert Opin Drug Metab Toxicol 21(8): 961-977.
16. L Jiang, C Jiang, X Yu, R Fu, S Jin, et al. (2022) DeepTTA: a transformer-based model for predicting cancer drug response. Brief Bioinform 23(3): bbac100.
17. Y Zhou, VM Lauschke (2022) Computational Methods and Approaches in Pharmacogenomic Research. in Comprehensive Pharmacology 2: 53-83.
18. T Jambor, G Juhasz, N Eszlari (2022) Towards personalised antidepressive medicine based on ‘big data’: an up-to-date review on robust factors affecting treatment response. Neuropsychopharmacol Hungarica 24(1): 17-28.
19. M Ran, SL Zhang, KY Tam (2026) Identifying meaningful drug response biomarkers from public pharmacogenomic datasets with biologically informed interpretable neural networks. Comput Biol Chem 120(Pt 1): 108669.
20. RH Scott, TA Fowler, M Caulfield (2019) Genomic medicine: time for health-care transformation. Lancet 394(10197): 454-456.
21. SB Haga (2020) Individualizing pharmacogenomic test results in the context of the microbiome. Per Med 17(6): 459-468.
22. C Piyawajanusorn, LC Nguyen, G Ghislat, PJ Ballester (2021) A gentle introduction to understanding preclinical data for cancer pharmaco-omic modeling. Brief Bioinform 22(6): bbab312.
23. Y Wang, Y Yang, S Chen, J Wang (2021) Deepdrk: A deep learning framework for drug repurposing through kernel-based multi-omics integration. Brief Bioinform 22(5): bbab048.
24. A Menke, H Weber, J Deckert (2020) Roadmap for Routine Pharmacogenetic Testing in a Psychiatric University Hospital. Pharmacopsychiatry 53(4): 179-183.
25. J Lever, Julia M Barbarino, Li Gong, Rachel Huddart, Katrin Sangkuhl, et al. (2020) PGxMine: Text mining for curation of pharmgKB. Pac Symp Biocomput 25: 611-622.
26. R Su, X Liu, L Wei, Q Zou (2019) Deep-Resp-Forest: A deep forest model to predict anti-cancer drug response. Methods 166: 91-102.
27. X Dong, Huanhuan Liu, Ting Tong, Liuxing Wu, Jianhua Wang, et al. (2025) Personalized prediction of anticancer potential of non-oncology drugs through learning from genome derived molecular pathways. NPJ Precis Oncol 9(1): 36.
28. G Yu, Q Fan (2025) Deep learning-driven drug response prediction and mechanistic insights in cancer genomics. Sci Rep 15(1): 20824.
29. Y Zhou, Q Dai, Y Xu, S Wu, M Cheng, et al. (2025) Pharma Former predicts clinical drug responses through transfer learning guided by patient derived organoid. NPJ Precis Oncol 9(1): 282.
30. L Silva, T Pacheco, E Araújo, RJ Duarte, I Ribeiro Vaz, et al. (2024) Unveiling the future: precision pharmacovigilance in the era of personalized medicine. Int J Clin Pharm 46(3): 755-760.
31. A Mondello, M Dal Bo, G Toffoli, M Polano (2023) Machine learning in onco-pharmacogenomics: a path to precision medicine with many challenges. Front Pharmacol 14: 1260276.
32. VM Lauschke, Y Zhou, M Ingelman Sundberg (2024) Pharmacogenomics Beyond Single Common Genetic Variants: The Way Forward. Annu Rev Pharmacol Toxicol 64: 33-51.
33. M Zack, Danil N Stupichev, Alex J Moore, Ioan D Slobodchikov, David G Sokolov, et al. (2025) Artificial Intelligence and Multi-Omics in Pharmacogenomics: A New Era of Precision Medicine. Mayo Clin Proc Digit Hea 3(3): 100246.
34. AK Srivastav, MK Mishra, JW Lillard, R Singh (2025) Transforming Pharmacogenomics and CRISPR Gene Editing with the Power of Artificial Intelligence for Precision Medicine. Pharmaceutics 17(5): 555.
35. MR Shapiro, EM Tallon, ME Brown, AL Posgai, MA Clements, et al. (2025) Leveraging artificial intelligence and machine learning to accelerate discovery of disease-modifying therapies in type 1 diabetes. Diabetologia 68(3): 477-494.
36. R Tremmel, S Pirmann, Y Zhou, VM Lauschke (2025) Translating pharmacogenomic sequencing data into drug response predictions-How to interpret variants of unknown significance. Br J Clin Pharmacol 91(2): 252-263.
37. G Yang, M Mishra, MA Perera (2023) Multi-Omics Studies in Historically Excluded Populations: The Road to Equity. Clin Pharmacol Ther 113(3): 541-556.
38. SD Sharma, J Bluett (2024) Towards Personalized Medicine in Rheumatoid Arthritis. Open Access Rheumatol Res Rev 16: 89-114.
39. W Evans, EM Meslin, J Kai, N Qureshi (2024) Precision Medicine. Are We There Yet? A Narrative Review of Precision Medicine’s Applicability in Primary Care. J Pers Med 14(4): 418.
40. G Pepe, C Carrino, L Parca, M Helmer Citterich (2022) Dissecting the Genome for Drug Response Prediction. Methods in Molecular Biology 2449: 187-196.
41. A Del Casale, Giuseppe Sarli, Paride Bargagna, Lorenzo Polidori, Alessandro Alcibiade, et al. (2023) Machine Learning and Pharmacogenomics at the Time of Precision Psychiatry. Curr Neuropharmacol 21(12): 2395-2408.
42. J Shin, Y Piao, D Bang, S Kim, K Jo (2022) DRPreter: Interpretable Anticancer Drug Response Prediction Using Knowledge-Guided Graph Neural Networks and Transformer. Int J Mol Sci 23(22): 13919.
43. M Arbitrio (2024) Pharmacogenomics: Challenges and Future. Genes (Basel) 15(6): 714.
44. M Koido (2025) Polygenic modelling and machine learning approaches in pharmacogenomics: Importance in downstream analysis of genome-wide association study data. Br J Clin Pharmacol 91(2): 264-269.
45. WC Chang, R Tanoshima, CJD Ross, BC Carleton (2021) Challenges and Opportunities in Implementing Pharmacogenetic Testing in Clinical Settings. Annu Rev Pharmacol Toxicol 61: 65-84.
46. S Sotudian, IC Paschalidis (2022) Machine Learning for Pharmacogenomics and Personalized Medicine: A Ranking Model for Drug Sensitivity Prediction. IEEE/ACM Trans Comput Biol Bioinforma 19(4): 2324-2333.
47. TTD Nguyen, Ziaurrehman Tanoli, Saad Hassan, Umut Onur Özcan, Jimmy Caroli, et al. (2025) PGxDB: an interactive web-platform for pharmacogenomics research. Nucleic Acids Res 53(D1): D1486-D1497.
48. S Sohrabei, H Moghaddasi, A Hosseini, SJ Ehsanzadeh (2024) Investigating the effects of artificial intelligence on the personalization of breast cancer management: a systematic study. BMC Cancer 24(1): 852.
49. F Feng, B Shen, X Mou, Y Li, H Li (2021) Large-scale pharmacogenomic studies and drug response prediction for personalized cancer medicine. J Genet Genomics 48(7): 540-551.
50. Y Zhou, R Tremmel, E Schaeffeler, M Schwab, VM Lauschke (2022) Challenges and opportunities associated with rare-variant pharmacogenomics. Trends Pharmacol Sci l43(10): 852-865.
51. A Limeta, F Gatto, MJ Herrgård, B Ji, J Nielsen (2023) Leveraging high-resolution omics data for predicting responses and adverse events to immune checkpoint inhibitors. Comput Struct Biotechnol J 21: 3912-3919, 2023.
52. S Liao, L Wang, X Wei, L Guo (2023) Pharmacogenetics and pharmacogenomics in glaucoma therapeutics: The way to personalized therapy. Chin Med J (Engl) 136(21): 2573-2575.
53. Y Lee, S Nam (2021) Performance comparisons of alexnet and googlenet in cell growth inhibition ic50 prediction. Int J Mol Sci 22(14): 7721.
54. W Ba Alawi, Sisira Kadambat Nair, Bo Li, Anthony Mammoliti, Petr Smirnov, et al. (2022) Bimodal Gene Expression in Patients with Cancer Provides Interpretable Biomarkers for Drug Sensitivity. Cancer Res 82(13): 2378-2387.
55. J Kim, SH Park, H Lee (2024) PANCDR: precise medicine prediction using an adversarial network for cancer drug response. Brief Bioinform 25(2): bbae088.
56. AV Singh, Vaisali Chandrasekar, Namuna Paudel, Peter Laux, Andreas Luch, et al. (2023) Integrative toxicogenomics: Advancing precision medicine and toxicology through artificial intelligence and OMICs technology. Biomed Pharmacother 163: 114784.
57. S Park, J Soh, H Lee (2021) Super. FELT: supervised feature extraction learning using triplet loss for drug response prediction with multi-omics data. BMC Bioinformatics 22(1): 269.
58. PC Guest, S Borgwardt, J Steiner (2023) Editorial: Back to the future: on the road towards precision psychiatry, volume II. Front Psychiatry 14: 1355085.
59. X Cheng, Chong Dai, Yuqi Wen, Xiaoqi Wang, Xiaochen Bo, et al. (2022) NeRD: a multichannel neural network to predict cellular response of drugs by integrating multidimensional data. BMC Med 20(1): 368.
60. CK Nkera Gutabara, R Kerr, J Scholefield, S Hazelhurst, J Naidoo (2022) Microbiomics: The Next Pillar of Precision Medicine and Its Role in African Healthcare. Front Genet 13: 869610.
61. TO Tran, TH Nguyen, TT Nguyen, NQK Le (2025) MLG2Net: Molecular Global Graph Network for Drug Response Prediction in Lung Cancer Cell Lines. J Med Syst 49(1): 47.
62. TO Tran, QH Kha, NQK Le (2024) RF-Lung-DR: Integrating Biological and Drug SMILES Features in a Random Forest-Based Drug Response Predictor for Lung Cancer Cell Lines. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics): 157-167.
63. MA Zahra, Abdulla Al Taher, Mohamed Alquhaidan, Tarique Hussain, Izzeldin Ismail, et al. (2024) The synergy of artificial intelligence and personalized medicine for the enhanced diagnosis, treatment, and prevention of disease. Drug Metab Pers Ther 39(2): 47-58.
64. JU Verga, A Lloyd, A Sarron, G Hardiman (2024) Genomic technology advances and the promise for precision medicine. Therapeutic Drug Monitoring: Newer Drugs Biomar Sec Ed: 355-371.
65. HM Nafchi, H Solatzadeh, E Hajimaghsoudi, E Babakhanzadeh (2025) Personalizing cancer therapy: the role of pharmacogenetics in overcoming drug resistance and toxicity. Mol Biol Rep 52(1): 785.
66. B Vanderwerff, Amy L Pasternak, Lars G Fritsche, Emily Bertucci Richter, Snehal Patil, et al. (2025) Expanding biobank pharmacogenomics through machine learning calls of structural variation. Genetics 230(3): iyaf088.
67. F Ahmad (2025) Optimizing Treatment: The Role of Pharmacology, Genomics, and AI in Improving Patient Outcomes. Drug Dev Res 86(3): e70093.
68. NN Le, I Frater, S Lip, S Padmanabhan (2025) Hypertension precision medicine: the promise and pitfalls of pharmacogenomics. Pharmacogenomics 26(5-6): 183-206.
69. O Muminov, S Muradov, D Numonov, E Ibragimova, A Norjigitov, et al. (2025) AI and ML in Genomic Research: Unlocking the Potential of Precision Medicine. Proceedings 1st International Conference on Frontier Technologies and Solutions, ICFTS 2025.
70. H Tan, SM Reed (2022) Metabolovigilance: Associating Drug Metabolites with Adverse Drug Reactions. Mol Inform 41(6)
71. J Janiel, PS Rajendran (2024) Leveraging Explainable AI for Improved Understanding and Prediction of Drug Responses in IGF1R Signaling Pathways. 2024 International Conference on Advances in Data Engineering and Intelligent Computing Systems, ADICS 2024,
72. SB Haga (2024) Artificial intelligence, medications, pharmacogenomics, and ethics. Pharmacogenomics 25(14-15): 611-622.
73. YM Roman (2025) Pharmacogenomics and rare diseases: optimizing drug development and personalized therapeutics. Pharmacogenomics 26(3-4): 121-128.
74. R Tiwari, Dhruv Dev, Maharshi Thalla, Vaibhav Dagaji Aher, Anand Badrivishal Mundada, et al. (2025) Nano-enabled pharmacogenomics: revolutionizing personalized drug therapy. J Biomater Sci Polym Ed 36(7): 913-938.
75. P Selvi Rajendran, J Jawahar (2025) An EETR Approach for Therapeutic Response Prediction Using Gene Expression and Drug Properties. Lecture Notes in Networks and Systems: 471-479.
76. D Primorac, L Bach Rojecky, P Brlek, V Škaro (2024) Pharmacogenomics and Big Data. Pharmacogenomics in Clinical Practice: 313-324.
77. R Tremmel, A Honore, Y Park, Y Zhou, M Xiao, et al. (2025) Machine learning models for pharmacogenomic variant effect predictions–recent developments and future frontiers. Pharmacogenomics 26(5-6): 171-182.
78. CB Avci, BG Bagca, B Shademan, LS Takanlou, MS Takanlou, et al. (2024) Machine learning in oncological pharmacogenomics: advancing personalized chemotherapy. Funct Integr Genomics 24(5): 182.
79. A Wong, et al. (2023) Role of artificial intelligence in pharmacy practice: A narrative review. JACCP J Am Coll Clin Pharm 6(11): 1237-1250.
80. GP Patrinos, N Sarhangi, B Sarrami, N Khodayari, B Larijani, et al. (2023) Using ChatGPT to predict the future of personalized medicine. Pharmacogenomics J 23(6): 178-184.
81. FR Giugni, JD Berry, A Khera, AM Shah, JA de Lemos (2024) Precision Medicine for Cardiovascular Prevention and Population Health: A Bridge Too Far? Circulation 150(21): 1720-1731.