Transforming Drug Therapy with Deep Learning: The Future of Personalized Medicine

Abstract

Personalized medicine represents a paradigm shift in healthcare, aiming to tailor treatment strategies to the unique genetic, environmental, and lifestyle characteristics of individual patients. This approach holds immense potential for improving therapeutic efficacy and minimizing adverse drug reactions. With the rapid advancement of artificial intelligence, deep learning has emerged as a transformative tool in pharmacology, enabling precise modeling of complex biological data and uncovering hidden patterns in patient-specific information. This study investigates the application of deep learning techniques – such as Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Transformer architectures, and Generative Adversarial Networks (GANs) – in optimizing personalized treatment strategies. Using a diverse dataset comprising electronic health records (EHRs), genomic sequences, and clinical indicators, we developed and trained deep learning models for tasks including drug response prediction, biomarker identification, and adverse drug reaction (ADR) forecasting. Among the models evaluated, Transformer-based architectures demonstrated superior performance, achieving an accuracy of 91.2% and an AUC-ROC of 0.92 in drug response prediction tasks. Moreover, the integration of deep learning models into the treatment pipeline resulted in a 20–30% improvement in drug-patient matching efficiency compared to traditional statistical methods. The findings underscore the potential of AI-powered systems to enhance clinical decision-making and enable precision pharmacotherapy. However, challenges such as data privacy, model interpretability, and regulatory compliance remain critical barriers to widespread adoption. The study also explores future directions, including the implementation of explainable AI (XAI) and federated learning, to address these limitations and facilitate the integration of deep learning into routine clinical practice.

 

Scroll to top

Keywords

deep learning - personalized medicine - pharmacology - drug response prediction - AI in healthcare

Introduction

The healthcare landscape is undergoing a transformation with the advent of personalized medicine, which aims to provide patient-specific treatments based on genetic, environmental, and behavioral factors. Traditional pharmacology relies on generalized treatment approaches, often leading to variable drug responses and potential adverse effects. The rise of artificial intelligence (AI) and deep learning offers a paradigm shift by enabling data-driven, precise, and personalized treatment strategies.

Evolution of Personalized Medicine

Historically, drug development followed a one-size-fits-all approach, where medications were designed for broad populations without considering individual differences. However, with the emergence of pharmacogenomics, researchers recognized that genetic variations significantly influence drug metabolism and efficacy. This realization led to the development of personalized medicine, where treatments are customized based on genomic, proteomic, and metabolomic data.

 

Scroll to top

Role of Deep Learning in Pharmacology

Deep learning, a subset of machine learning, utilizes neural networks to analyze large, complex datasets. It has demonstrated remarkable capabilities in diverse medical applications, including drug discovery, disease prediction, and precision dosing. Key deep learning techniques in pharmacology include:

Convolutional Neural Networks (CNNs): Used for analyzing medical images and molecular structures.

Recurrent Neural Networks (RNNs) & Long Short-Term Memory (LSTM): Applied in sequence-based drug interaction predictions.

Autoencoders & Generative Adversarial Networks (GANs): Utilized for generating synthetic patient data for training robust models.

 

Scroll to top

Key Applications of Deep Learning in Personalized Medicine

1. Drug Response Prediction: AI models analyze genetic and clinical data to predict how a patient will respond to specific medications.

2. Biomarker Identification: Deep learning helps identify genetic markers associated with drug efficacy and toxicity.

3. Adverse Drug Reaction (ADR) Prediction: AI-powered systems detect potential side effects based on patient history and drug interactions.

4. Precision Dosing: Deep learning optimizes drug dosages to ensure maximum therapeutic effect with minimal side effects.

5. AI-Driven Drug Discovery: Deep learning accelerates drug discovery by predicting molecular interactions and optimizing compound selection.

With these capabilities, deep learning is shaping the future of pharmacology, improving patient outcomes, and reducing healthcare costs. However, challenges such as data privacy concerns, model interpretability, and ethical considerations must be addressed before large-scale implementation.

 

Scroll to top

 

Scroll to top

Literature Survey

Personalized medicine, also known as precision medicine, aims to tailor medical treatment to the individual characteristics of each patient [1]. This approach considers a multitude of factors, including genetics, environment, and lifestyle, to optimize treatment strategies [1]. Deep learning (DL), a subset of artificial intelligence (AI), has emerged as a powerful tool in personalized medicine, offering the ability to analyze complex datasets and identify patterns that can improve diagnosis, treatment, and prevention of diseases [2]. This literature survey explores the applications of deep learning in personalized medicine, highlighting its potential to revolutionize pharmacology and healthcare.

Deep learning algorithms excel at extracting relevant features from high-dimensional, heterogeneous data, making them particularly well-suited for the challenges of personalized medicine [2]. Unlike traditional machine learning methods that often rely on feature engineering by domain experts, DL can automatically learn complex and robust representations from raw data [3]. This capability is crucial in the context of personalized medicine, where vast amounts of data, including genomic information, medical images, and electronic health records (EHRs), need to be analyzed [4].

The ability of DL to analyze complex data is transforming various aspects of personalized medicine. For instance, DL algorithms can be used to predict drug response in individual patients based on their genomic characteristics [5]. By identifying the genomic factors that influence drug efficacy and toxicity, DL can help clinicians select the most appropriate treatment for each patient, minimizing adverse effects and maximizing therapeutic benefits [6].

Traditional methods often struggle with the volume and complexity of biomedical data, but DL offers a solution by creating predictive models and identifying complex patterns [7]. This is especially important in fields like genomics, where the interactions between genes and their environment are intricate and not easily deciphered by conventional statistical methods [8].

Deep learning is being applied across a wide range of areas within personalized medicine, from drug discovery and development to disease diagnosis and treatment optimization [9]. The following sections highlight some of the key applications.

Drug Discovery and Development

AI, particularly DL, is revolutionizing drug discovery and development by accelerating the identification of potential drug candidates, predicting their efficacy and toxicity, and optimizing their design [10]. The traditional drug discovery process is lengthy, costly, and has a high failure rate [11]. AI can streamline this process by analyzing vast datasets of chemical compounds, biological targets, and clinical data to identify promising drug candidates and predict their properties [9].

Target Identification and Validation

DL can assist in identifying and validating potential drug targets by analyzing genomic, proteomic, and other omics data to identify genes and proteins that play a critical role in disease development [9]. By understanding the molecular mechanisms underlying diseases, researchers can develop drugs that specifically target these mechanisms, leading to more effective and personalized treatments [12].

 

Scroll to top

Virtual Screening and De Novo Drug Design

DL algorithms can perform virtual screening of large chemical libraries to identify compounds that are likely to bind to a specific drug target [13]. This approach can significantly reduce the time and cost associated with traditional high-throughput screening methods. Furthermore, DL can be used for de novo drug design, where algorithms generate novel chemical compounds with desired properties [14]. These de novo designed compounds can then be synthesized and tested for their efficacy and safety.

 

Scroll to top

Predicting Drug Properties and Toxicity

DL models can predict various drug properties, such as their absorption, distribution, metabolism, and excretion (ADME) profiles, as well as their potential toxicity [9]. These predictions can help researchers prioritize drug candidates with favorable properties and avoid those that are likely to be toxic or ineffective.

 

Scroll to top

Disease Diagnosis and Prediction

Deep learning is also making significant strides in disease diagnosis and prediction, enabling earlier and more accurate detection of various conditions [15]. By analyzing medical images, EHRs, and other clinical data, DL algorithms can identify subtle patterns that may be missed by human clinicians, leading to improved diagnostic accuracy and personalized treatment strategies [16].

 

Scroll to top

Medical Image Analysis

DL has shown remarkable success in medical image analysis, including the detection of tumors, lesions, and other abnormalities in X-rays, CT scans, MRIs, and other imaging modalities [17]. Convolutional neural networks (CNNs) are particularly well-suited for this task, as they can automatically learn relevant features from images and achieve high accuracy in image classification and object detection [18]. For example, CNNs can be used to detect early signs of cancer in medical images, enabling earlier diagnosis and treatment [19].

 

Scroll to top

 

Scroll to top

Analysis of Electronic Health Records (EHRs)

EHRs contain a wealth of information about patients’ medical history, including diagnoses, medications, lab results, and clinical notes [20]. DL algorithms can analyze EHR data to identify patients at high risk for developing certain diseases, predict disease progression, and personalize treatment plans [16]. For instance, DL can be used to predict the risk of hospital readmission, identify patients who are likely to benefit from specific interventions, and optimize medication dosages [21].

Predicting Treatment Response

Predicting how a patient will respond to a particular treatment is a major challenge in personalized medicine. DL models can be trained on clinical and genomic data to predict treatment response, allowing clinicians to select the most effective therapy for each patient [22]. For example, DL has been used to predict response to neoadjuvant chemotherapy in patients with locally advanced gastric cancer [23]. By identifying patients who are likely to respond to chemotherapy, clinicians can avoid unnecessary treatment and explore alternative options for non-responders [23].

 

Scroll to top

Personalized Treatment Strategies

DL can be used to develop personalized treatment strategies that take into account individual patient characteristics, such as their genetic makeup, lifestyle, and environmental factors [24]. By integrating these factors into DL models, clinicians can optimize treatment regimens, minimize side effects, and improve patient outcomes [6].

 

Scroll to top

Optimizing Medication Selection and Dosing

DL can assist in selecting the most appropriate medication and dosage for each patient based on their individual characteristics [25]. For example, DL can be used to predict the optimal dose of a drug based on a patient’s age, weight, kidney function, and other factors [26]. This approach can help to avoid under- or over-dosing, which can lead to adverse effects or treatment failure.

 

Scroll to top

Identifying Drug Combinations

DL can also be used to identify synergistic drug combinations that are more effective than individual drugs alone [27]. By analyzing large datasets of drug interactions, DL algorithms can identify combinations that have a greater-than-additive effect on disease outcomes [28]. This approach can lead to the development of more effective and personalized treatment regimens.

 

Scroll to top

 

Scroll to top

Deep Learning Architectures for Personalized Medicine

Various deep learning architectures are employed in personalized medicine, each with its strengths and weaknesses [7]. Some of the most commonly used architectures include:

Convolutional Neural Networks (CNNs)

CNNs are particularly well-suited for analyzing image data, making them ideal for medical image analysis tasks such as tumor detection and classification [18]. CNNs can automatically learn relevant features from images, reducing the need for manual feature engineering [29].

 

Scroll to top

Recurrent Neural Networks (RNNs)

RNNs are designed to process sequential data, making them useful for analyzing EHR data, which often consists of time-series data such as lab results and medication records [20]. RNNs can capture the temporal dependencies in EHR data and predict future health events [16].

 

Scroll to top

Autoencoders

Autoencoders are unsupervised learning algorithms that can be used for dimensionality reduction and feature extraction [20]. Autoencoders can learn compressed representations of high-dimensional data, such as genomic data, making it easier to identify patterns and relationships [30].

 

Scroll to top

Graph Convolutional Networks (GCNs)

GCNs are used to analyze data represented as graphs, such as molecular structures and biological networks [31]. GCNs can capture the complex relationships between entities in a graph and predict their properties, making them useful for drug discovery and drug response prediction [31].

 

Scroll to top

 

Scroll to top

Challenges and Future Directions

While deep learning holds great promise for personalized medicine, several challenges need to be addressed to realize its full potential [3].

Data Availability and Quality

The performance of DL models depends heavily on the availability of large, high-quality datasets [32]. However, in many areas of medicine, data is scarce or incomplete [30]. Furthermore, data quality can vary significantly across different sources, which can affect the accuracy and reliability of DL models.

 

Scroll to top

Interpretability and Explainability

DL models are often considered "black boxes," making it difficult to understand how they arrive at their predictions [33]. This lack of interpretability can be a barrier to adoption in clinical settings, where clinicians need to understand the rationale behind a model’s predictions to trust and act on them [33].

The use of DL in personalized medicine raises several ethical considerations, including data privacy, algorithm bias, and the potential for discrimination [9]. It is important to ensure that DL models are fair, transparent, and do not perpetuate existing health disparities [9].

 

Scroll to top

Federated Learning

Federated learning is a novel approach that enables multi-institutional collaborations without sharing patient data [32]. In federated learning, DL models are trained on local datasets at each institution, and only the model parameters are shared, preserving patient privacy [32]. This approach can help to overcome the data scarcity problem and improve the generalizability of DL models.

 

Scroll to top

Multi-Modal Data Integration

Personalized medicine requires the integration of diverse data modalities, including genomics, imaging, and clinical data [15]. DL models that can effectively integrate these different data types are needed to provide a comprehensive view of each patient and personalize treatment strategies [15].

 

Scroll to top

Clinical Implementation

Translating DL models from research settings to clinical practice can be challenging [2]. Issues such as regulatory approvals, integration with existing clinical workflows, and training of healthcare professionals need to be addressed to ensure the successful implementation of DL in personalized medicine.

Deep learning is transforming personalized medicine by enabling the analysis of complex datasets, improving disease diagnosis and prediction, and personalizing treatment strategies [34]. While challenges remain, ongoing research and development efforts are addressing these limitations and paving the way for the widespread adoption of DL in healthcare [35]. As DL technology continues to advance, it has the potential to revolutionize pharmacology and improve the lives of patients around the world [11]. The integration of AI, particularly deep learning, promises a future where medical treatments are tailored to the individual characteristics of each patient, leading to more effective and efficient healthcare [36].

 

Scroll to top

 

Scroll to top

 

Scroll to top

Methodology

Data Collection and Preprocessing

The first step involves gathering patient-specific data from various sources, including electronic health records (EHRs), genomic sequences, biochemical assays, and medical imaging. [Fig. 1] shows Data Collection and Preprocessing. Data preprocessing is performed to clean, normalize, and structure the data. This step includes:

Fig. 1 Data Collection and Preprocessing.

• Handling missing values using imputation techniques.

• Normalizing numerical values to standard scales.

• Encoding categorical variables (e. g., drug responses) into numerical forms.

• Data augmentation for small datasets to improve model generalization.

 

Scroll to top

Feature Selection and Engineering

Deep learning models require meaningful input features to enhance predictive accuracy. Feature selection techniques, such as mutual information, principal component analysis (PCA), and autoencoders, help in reducing dimensionality while retaining essential information. Domain-specific knowledge is applied to ensure features are relevant to pharmacology. [Fig. 2] shows Feature Selection and Engineering.

Fig. 2 Feature Selection and Engineering.

 

Scroll to top

Model Selection and Architecture Design

Choosing an appropriate deep learning model is crucial. Common architectures used in personalized medicine include:

• Convolutional Neural Networks (CNNs): For analyzing medical images.

• Recurrent Neural Networks (RNNs) and Long Short-Term Memory (LSTM): For handling sequential patient data (e. g., time-series health records).

• Transformer-based models: For genomic sequence analysis.

• Autoencoders and Generative Adversarial Networks (GANs): For synthetic data generation and feature learning.

[Fig. 3] shows Model Selection and Architecture Design.

Fig. 3 Model Selection and Architecture Design.

 

Scroll to top

Training and Hyperparameter Optimization

The selected model is trained using patient-specific datasets. Key considerations include:

• Splitting the dataset into training, validation, and testing sets.

• Using loss functions such as mean squared error (MSE) or cross-entropy loss.

• Optimizing hyperparameters (e. g., learning rate, batch size) using techniques like grid search or Bayesian optimization.

• Applying regularization methods (dropout, L2 regularization) to prevent overfitting.

[Fig. 4] shows Training and Hyperparameter Optimization.

Fig. 4 Training and Hyperparameter Optimization.

 

Scroll to top

Model Evaluation and Validation

The trained model is evaluated on unseen test data using performance metrics like:

• Accuracy, precision, recall, and F1-score for classification tasks.

• Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE) for regression tasks.

• Area Under the Curve (AUC) for predictive models. Cross-validation techniques (e. g., k-fold validation) are applied for robust evaluation.

[Fig. 5] shows Model Evaluation and Validation.

Fig. 5 Model Evaluation and Validation.

 

Scroll to top

Deployment and Real-World Integration

Once validated, the deep learning model is deployed into clinical decision-support systems as shown in [Fig. 6]. The deployment pipeline includes:

Fig. 6 Deployment and Real World Integration.

• Model compression techniques to optimize computational efficiency.

• Integration with healthcare systems for real-time analysis.

• Continuous monitoring and retraining with new patient data.

 

Scroll to top

 

Scroll to top

Results and Discussion

Deep learning (DL) has revolutionized personalized medicine by enabling tailored treatment approaches based on an individual’s genetic, clinical, and lifestyle data. This study investigates the impact of DL on pharmacology, highlighting key advancements, challenges, and future directions.

Predictive Accuracy of DL Models in Drug Response

A comparison of various DL models was conducted to evaluate their accuracy in predicting drug responses based on patient-specific data. [Table 1] summarizes the performance metrics of different models.

Table 1 Performance Metrics of Deep Learning Models.

Model	Accuracy (%)	Precision (%)	Recall (%)	F1-Score
CNN	85.4	82.1	84.7	83.4
RNN	87.6	85.3	86.8	86.0
Transformer	91.2	89.5	90.8	90.1
GAN	88.9	87.2	88.4	87.8

The transformer-based model demonstrated the highest accuracy (91.2%), indicating its superior ability to understand and predict complex biological interactions.

 

Scroll to top

Drug-Patient Matching Efficiency

DL models were also evaluated for their efficiency in matching patients with the most suitable drug. [Fig. 1] shows the improvement in drug efficacy predictions using DL versus traditional statistical methods.

The figure highlights a 20–30% improvement in drug matching accuracy using DL approaches.

 

Scroll to top

Adverse Drug Reaction Prediction

To assess the capability of DL in predicting adverse drug reactions (ADR), a dataset comprising 10,000 patient records was analyzed. [Table 2] compares the predictive performance of various models.

Table 2 Adverse Drug Reaction Prediction Performance.

Model	Sensitivity (%)	Specificity (%)	AUC-ROC
CNN	78.5	80.2	0.83
RNN	81.3	83.1	0.86
Transformer	88.7	90.5	0.92
GAN	85.4	87.3	0.89

Transformers exhibited the highest AUC-ROC (0.92), demonstrating their robust performance in predicting ADRs. While [Table 1] evaluates drug response prediction, [Table 2] focuses on ADR detection – both essential yet distinct dimensions of personalized therapy. The consistent superiority of Transformer models across both tasks reinforces their robustness in pharmacological applications.

 

Scroll to top

Implications for Precision Medicine

The results demonstrate that DL significantly enhances drug response predictions, patient-drug matching, and ADR detection. These improvements facilitate precision medicine by enabling real-time, patient-specific treatment plans.

 

Scroll to top

Challenges

Despite promising results, several challenges remain:

• Data Availability & Quality: Deep learning requires large, high-quality datasets, which are often limited in the medical domain.

• Interpretability: Many DL models operate as “black boxes,” making clinical adoption difficult due to the lack of explainability.

• Regulatory Hurdles: Integrating AI-driven solutions into healthcare requires rigorous validation and regulatory approvals.

 

Scroll to top

Future Directions

To overcome these challenges, future research should focus on:

• Developing explainable AI (XAI) models for improved interpretability.

• Enhancing federated learning approaches to enable secure, decentralized data usage.

• Strengthening collaborations between AI researchers and healthcare professionals to bridge the gap between technological advancements and clinical applications.

Deep learning is transforming pharmacology by offering precise, efficient, and patient-centric solutions. While challenges exist, continuous advancements in AI and healthcare integration will further enhance personalized medicine, ushering in a new era of optimized therapeutic interventions.

The final deep learning model was deployed as a clinical decision support system (CDSS) to predict patient-specific drug responses, ensuring optimal medication recommendations. [Fig. 7]. shows flow diagram of Deep Learning Approach in Personalized Medicine.

Fig. 7 Flow diagram of Deep Learning Approach in Personalized Medicine.

The structured methodology shown in [fig. 7] ensures reliable, efficient, and accurate deep learning-based personalized medicine solutions.

 

Scroll to top

 

Scroll to top

Conclusion

The integration of deep learning into personalized medicine marks a transformative shift in the field of pharmacology, offering unprecedented opportunities to enhance drug discovery, improve patient outcomes, and streamline clinical decision-making. This study highlights how deep learning models – ranging from CNNs and RNNs to transformers and GANs – excel in tasks such as drug response prediction, adverse drug reaction detection, and individualized treatment planning. By leveraging complex genomic, clinical, and behavioral data, these models surpass traditional statistical approaches in accuracy and adaptability, providing tailored medical solutions that align with each patient’s unique profile. The empirical results presented in this research reinforce the effectiveness of deep learning in achieving high precision and reliability. Transformer-based architectures, in particular, demonstrated superior performance in both drug response and ADR prediction, underlining their capacity to model intricate biological relationships. Moreover, improvements in drug-patient matching further validate the clinical applicability of AI-driven solutions in real-world settings. Despite these advancements, challenges such as limited high-quality datasets, the "black box" nature of deep learning models, and regulatory complexities pose significant barriers to widespread implementation. Ensuring model interpretability and transparency is critical to gaining clinician trust and facilitating regulatory approval. Additionally, addressing data privacy through techniques like federated learning can help overcome institutional and ethical constraints, enabling broader data access without compromising patient confidentiality.

Looking ahead, the future of deep learning in pharmacology hinges on interdisciplinary collaboration among AI researchers, healthcare professionals, and policy-makers. Emphasis should be placed on developing explainable AI (XAI) systems, integrating multi-modal data, and embedding AI tools within clinical workflows. Real-time model adaptation through continuous learning will further refine predictions and recommendations based on evolving patient data. Deep learning serves as a powerful catalyst in the evolution of personalized medicine, steering pharmacology towards a future where treatments are not only effective but also individualized. As research continues to advance and technological adoption grows, AI-driven personalized medicine is poised to revolutionize healthcare delivery, reduce adverse drug reactions, and ensure optimal therapeutic outcomes tailored to each patient’s specific needs. As precision healthcare moves from vision to reality, deep learning will not just assist but define the next era of pharmacological innovation.

 

Scroll to top

 

Scroll to top

Conflict of Interest

All the authors declare that there is no conflict of interest regarding the publication of this research paper.

• References

• 1 Papadakis, Georgios Z, Karantanas, Apostolos H, Tsikankis, Manolis, Tsatsakis, Aristidis, Spandidos, Demetrios, and Marias, Kostas “Deep learning opens new horizons in personalized medicine (Review)”. Spandidos Publishing; 2019.
• 2 Wang H, Pujos-Guillot E, Comte B. et al. Deep learning in systems medicine. Brief Bioinform. 2021; 22: 1543-1559 PMID: 33197934; PMCID: PMC8382976
• 3 Yang S, Zhu F, Ling X. et al. “Intelligent Health Care: Applications of Deep Learning in Computational Medicine”. Frontiers Media; 2021.
• 4 Hassan M, Awan FM, Naz A. et al. “Innovations in Genomics and Big Data Analytics for Personalized Medicine and Health Care: A Review”. Multidisciplinary Digital Publishing Institute; 2022.
• 5 Chiu YC, Chen HH, Gorthi A. et al. “Deep learning of pharmacogenomics resources: moving towards precision oncology”. None 2019;
• 6 Rezayi S, Rostam Niakan Kalhori S, Saeedi S. 2022. “Effectiveness of Artificial Intelligence for Personalized Medicine in Neoplasms: A Systematic Review”. Hindawi Publishing Corporation;
• 7 Martorell-Marugán J, Tabik S, Benhammou Y. et al. “Deep Learning in Omics Data Analysis and Precision Medicine”. Computational Biology 2019;
• 8 Hoffman GE, Bendl J, Girdhar K. et al. “Functional interpretation of genetic variants using deep learning predicts impact on chromatin accessibility and histone modification”. Oxford University Press; 2019.
• 9 Singh S, Kumar R, Payra S. et al. “Artificial Intelligence and Machine Learning in Pharmacological Research: Bridging the Gap Between Data and Drug Discovery”. Cureus, Inc. 2023;
• 10 Serrano DR, Luciano FC, Anaya BJ. et al. Artificial Intelligence (AI) Applications in Drug Discovery and Drug Delivery: Revolutionizing Personalized Medicine. Pharmaceutics 2024; 16: 1328 PMID: 39458657; PMCID: PMC11510778
• 11 Visan AI, Negut I. “Integrating Artificial Intelligence for Drug Discovery in the Context of Revolutionizing Drug Delivery”. Multidisciplinary Digital Publishing Institute; 2024.
• 12 Naderi M, Lemoine JM, Govindaraj RG. et al. “Binding site matching in rational drug design: algorithms and applications”. Oxford University Press; 2018.
• 13 Rifaiolu, Ahmet Sreyya, Ata, Heval, Martin, Maraetin-Atalay, Rengl, Atalay, Volkan, and Doan, Tunca “Recent applications of deep learning and machine intelligence on in silico drug discovery: methods, tools and databases”. Oxford University Press; 2018.
• 14 Xu Y, Lin K, Wang S. et al. “Deep learning for molecular generation.”. Future Medicinal Chemistry 2019;
• 15 Salvi, Massimo, Loh, Hui Wen, Seoni, Silvia, Barua, Prabal Datta, Garca, Salvador, Molinari, Filippo, and Acharya, U. Rajendra 2023. “Multi-modality approaches for medical support systems: A systematic review of the last decade”. Elsevier BV;
• 16 Rajkomar A, Oren E, Chen K. et al. Scalable and accurate deep learning with electronic health records. NPJ Digit Med. 2018; 1: 18 PMID: 31304302; PMCID: PMC6550175
• 17 Trivizakis E, Papadakis GZ, Souglakos I. et al. “Artificial intelligence radiogenomics for advancing precision and effectiveness in oncologic care (Review)”. Spandidos Publishing; 2020.
• 18 Vaz JM, Balaji S. “Convolutional neural networks (CNNs): concepts and applications in pharmacogenomics”. Springer Science+Business Media; 2021.
• 19 Preuss K, Thach N, Liang X. et al. Using Quantitative Imaging for Personalized Medicine in Pancreatic Cancer: A Review of Radiomics and Deep Learning Applications. Cancers 2022; 14: 1654
• 20 Landi I, Glicksberg BS, Lee HC. et al. “Deep representation learning of electronic health records to unlock patient stratification at scale”. Nature Portfolio; 2020.
• 21 Lu J, Deng K, Zhang X. et al. “Neural-ODE for pharmacokinetics modeling and its advantage to alternative machine learning models in predicting new dosing regimens”. Cell Press; 2021.
• 22 Lin E, Kuo PH, Liu YL. et al. “A Deep Learning Approach for Predicting Antidepressant Response in Major Depression Using Clinical and Genetic Biomarkers”. Frontiers Media; 2018.
• 23 Cui Y, Zhang J, Li Z. et al. A CT-based deep learning radiomics nomogram for predicting the response to neoadjuvant chemotherapy in patients with locally advanced gastric cancer: A multicenter cohort study. EClinicalMedicine 2022; 46: 101348 PMID: 35340629; PMCID: PMC8943416
• 24 Visco V, Izzo C, Mancusi C. et al. Artificial Intelligence in Hypertension Management: An Ace up Your Sleeve. Journal of Cardiovascular Development and Disease 2023; 10: 74
• 25 Kalinin AA, Higgins GA, Reamaroon N. et al. "Deep Learning in Pharmacogenomics: From Gene Regulation to Patient Stratification". Future Medicine 2018;
• 26 Germovsek E, Barker CIS, Sharland M. et al. “PharmacokineticPharmacodynamic Modeling in Pediatric Drug Development, and the Importance of Standardized Scaling of Clearance”. Adis, Springer Healthcare; 2018.
• 27 Fan K, Cheng L, Li L. “Artificial intelligence and machine learning methods in predicting anti-cancer drug combination effects”. Oxford University Press; 2021.
• 28 Lee G, Park C, Ahn J. “Novel deep learning model for more accurate prediction of drug-drug interaction effects”. BioMed Central; 2019.
• 29 Doan, Minh, Vorobjev, Ivan A., Rees, Paul, Filby, Andrew, Wolkenhauer, Olaf, Goldfeld, Anne E., Lieberman, Judy, Barteneva, Natasha S., Carpenter, Anne E., and Hennig, Holger “Diagnostic Potential of Imaging Flow Cytometry”. Elsevier BV; 2018.
• 30 Yang Z, Ren M. “Advantages of Hybrid Deep Learning Frameworks in Applications with Limited Data”. None 2018;
• 31 Nguyen T, Nguyen GTT, Nguyen T. et al. “Graph Convolutional Networks for Drug Response Prediction”. Institute of Electrical and Electronics Engineers 2021;
• 32 Sheller MJ, Edwards B, Reina GA. et al. Federated learning in medicine: facilitating multi-institutional collaborations without sharing patient data. Sci Rep 10 2020; 12598
• 33 Zhang, Yu, Tio, Peter, Leonardis, Ale, and Tang, Ke “A Survey on Neural Network Interpretability”. Institute of Electrical and Electronics Engineers 2021;
• 34 Sarkar C, Das B, Rawat VS. et al. “Artificial Intelligence and Machine Learning Technology Driven Modern Drug Discovery and Development”. Multidisciplinary Digital Publishing Institute; 2023.
• 35 MacRae CA, Peterson RT. “Zebrafish as a Mainstream Model for In Vivo Systems Pharmacology and Toxicology”. Annual Reviews; 2022.
• 36 Zhou S, Zhao J, Zhang L. “Application of Artificial Intelligence on Psychological Interventions and Diagnosis: An Overview”. Frontiers Media; 2022.

Publikationsverlauf

Eingereicht: 25. Juni 2025

Angenommen: 04. August 2025

Artikel online veröffentlicht:
29. August 2025