Artificial Intelligence in Drug Discovery and Pharmacology

Abstract

Artificial Intelligence (AI) has emerged as a transformative force in drug discovery and pharmacology, offering unprecedented capabilities to accelerate and optimize the traditionally lengthy, expensive, and complex drug development pipeline. By leveraging machine learning (ML), deep learning (DL), and natural language processing (NLP), AI enables rapid identification of drug targets, virtual screening of lead compounds, drug repurposing, and early prediction of pharmacokinetic and pharmacodynamic properties. These techniques significantly reduce the time and cost associated with bringing new therapies to market while enhancing accuracy and efficacy. In pharmacology, AI aids in modeling drug interactions, understanding pharmacogenomic profiles, and predicting adverse drug reactions (ADRs) through real-time data from electronic health records (EHRs) and post-marketing surveillance. Tools such as AlphaFold for protein structure prediction, DeepChem for cheminformatics, and commercial platforms like Schrödinger and Insilico Medicine exemplify the integration of AI into both preclinical and clinical stages of drug development. Despite the promise, challenges such as data heterogeneity, lack of interpretability in deep models, and regulatory ambiguities remain significant hurdles. Ethical concerns surrounding data privacy and algorithmic bias also necessitate cautious implementation. However, ongoing advancements in explainable AI, federated learning, and quantum computing are poised to address many of these limitations. This review provides a comprehensive overview of current AI applications in drug discovery and pharmacology, discusses emerging tools and platforms, and explores future directions aimed at realizing precision medicine. The interdisciplinary collaboration between AI experts, pharmacologists, and regulatory bodies will be key to unlocking the full potential of AI in therapeutics.

Keywords

Artificial Intelligence, Drug Discovery, Pharmacology, Machine Learning, Precision Medicine.

Introduction

The global drug discovery landscape is undergoing a paradigm shift driven by the increasing integration of Artificial Intelligence (AI) technologies. Traditionally, drug development is an extensive and resource-intensive process, often taking over 10–15 years and costing upwards of $2.6 billion to bring a single new chemical entity to market [1]. Despite these efforts, high attrition rates in clinical trials due to inefficacy or toxicity pose a persistent challenge, highlighting the urgent need for more efficient and predictive approaches. Artificial Intelligence, particularly machine learning (ML) and deep learning (DL), offers powerful tools for learning from large-scale biological, chemical, and clinical data to identify hidden patterns and actionable insights. These technologies are transforming various stages of the drug discovery pipeline—ranging from target identification, virtual screening, and de novo molecule generation to predicting pharmacokinetic/pharmacodynamic (PK/PD) behavior and adverse drug reactions (ADRs) [2,3]. Furthermore, AI-driven techniques such as natural language processing (NLP) facilitate mining of vast biomedical literature, patents, and clinical records to uncover novel drug-target-disease relationships [4]. In pharmacology, AI is proving instrumental in optimizing dosage regimens, modeling drug-drug interactions, and individualizing therapies through pharmacogenomics [5]. Platforms such as DeepMind's AlphaFold have revolutionized structural biology by accurately predicting protein structures, enabling rational drug design at an unprecedented scale [6]. Several biopharmaceutical companies and startups are now leveraging AI for high-throughput virtual screening, drug repurposing, and preclinical toxicity prediction [7,8]. Notable examples include BenevolentAI, Insilico Medicine, and Exscientia, all of which have developed AI-generated compounds that have progressed to clinical trials [9]. The COVID-19 pandemic has further accelerated AI adoption, with platforms identifying repurposable drugs in record time [10].

Despite these advances, integrating AI in drug development presents challenges, such as the need for high-quality, interoperable datasets, model interpretability, regulatory hurdles, and ethical considerations related to patient data. Nonetheless, AI is poised to become a cornerstone in modern drug discovery and pharmacology, with the potential to shorten development cycles, reduce costs, and improve therapeutic outcomes.

2. AI TECHNOLOGIES IN DRUG DISCOVERY

2.1 Machine Learning (ML)

Machine Learning (ML) plays a pivotal role in modern drug discovery by enabling the analysis of complex, high-dimensional biological and chemical datasets. Supervised learning is commonly applied in predicting molecular properties, binding affinities, and clinical outcomes based on labeled data. Unsupervised learning helps in clustering compounds and identifying hidden patterns in omics data. Reinforcement learning (RL) is emerging as a method to optimize de novo molecular design by simulating trial-and-error strategies in molecular spaces [11,12].

2.2 Deep Learning (DL)

Deep Learning, a subfield of ML, utilizes neural networks with many layers to learn complex representations of data. Convolutional Neural Networks (CNNs) are widely used for analyzing biomedical images, such as histopathological slides, to identify cancer subtypes and drug responses. Recurrent Neural Networks (RNNs), including long short-term memory (LSTM) networks, are effective in processing sequential data such as SMILES (Simplified Molecular Input Line Entry System) strings and genomic sequences. DL models significantly outperform traditional QSAR models in prediction tasks [13,14].

2.3 Natural Language Processing (NLP)

Natural Language Processing (NLP) is employed to mine biomedical literature, patents, clinical trial reports, and EHRs. By extracting relationships among drugs, genes, proteins, and diseases, NLP helps in building knowledge graphs and drug-disease association maps. Named entity recognition (NER), part-of-speech tagging, and dependency parsing are common techniques used in biomedical NLP pipelines. Recent models like BioBERT and SciSpacy are trained specifically on biomedical corpora [15,16].

2.4 Generative Models

Generative models such as Generative Adversarial Networks (GANs) and Variational Autoencoders (VAEs) have revolutionized the field of de novo drug design. These models generate novel chemical structures by learning latent representations of molecular data. GANs use a generator-discriminator setup to create realistic molecular graphs or SMILES strings, while VAEs map molecules into a continuous latent space for property optimization. These methods can generate drug-like molecules with specific target affinities or ADMET profiles [17,18].

3. Key Applications in Drug Discovery

3.1 Target Identification and Validation

Target identification is the foundational step in drug discovery. AI analyzes complex biological datasets—such as genomics, transcriptomics, and proteomics—to uncover disease-associated genes and proteins. Network-based algorithms and machine learning methods like Random Forests, Support Vector Machines (SVM), and deep learning extract actionable insights from high-dimensional omics data. AI can also integrate knowledge graphs and protein–protein interaction networks to prioritize and validate therapeutic targets [19,20]

3.2 Lead Compound Discovery and Optimization

AI accelerates hit-to-lead and lead optimization stages by predicting how candidate compounds interact with biological targets. Virtual screening (VS) powered by ML and DL can rapidly evaluate thousands to millions of compounds for desirable binding affinity, selectivity, and physicochemical properties. Tools such as AtomNet and DeepDock deploy convolutional neural networks (CNNs) to assess 3D molecular docking scores and suggest lead molecules [21,22].

3.3 Drug Repurposing

AI significantly shortens the drug development timeline through drug repurposing—identifying new uses for existing drugs. Using machine learning on molecular profiles, clinical data, and literature, AI uncovers hidden relationships among drugs, targets, and diseases. Notably, Benevolent AI used NLP and knowledge graph-based approaches to propose baricitinib as a COVID-19 therapeutic, which later entered clinical trials [23,24].

3.4 ADMET Prediction

Predicting Absorption, Distribution, Metabolism, Excretion, and Toxicity (ADMET) early in drug development reduces costly clinical failures. AI models trained on large annotated datasets can forecast these properties with high precision. Deep neural networks, ensemble models, and transfer learning approaches have improved the prediction of hepatotoxicity, blood-brain barrier permeability, and cytochrome P450 interactions [25,26].

3.5 Clinical Trial Design and Patient Stratification

AI enhances clinical trial success by identifying biomarkers, optimal endpoints, and responsive patient subgroups. Machine learning can stratify patients based on genetic or phenotypic data, allowing for precision medicine approaches. Predictive models also optimize trial design by simulating outcomes and estimating statistical power, helping reduce time and cost. AI has been integrated into adaptive trials to modify protocols in real time [27,28].

4. 4. Applications in Pharmacology

4.1 Pharmacokinetic and Pharmacodynamic Modeling (PK/PD)

Pharmacokinetics (PK) and pharmacodynamics (PD) are essential for understanding how drugs behave in the body and how they elicit therapeutic effects. AI, especially machine learning algorithms, enhances PK/PD modeling by predicting drug concentration-time profiles, inter-individual variability, and dose-response relationships. Unlike traditional compartmental models, AI-based models can account for nonlinear dynamics, complex covariate effects, and large-scale population variability (29). For instance, deep learning models can simulate virtual populations to forecast plasma drug levels under different dosing scenarios (30). This enables optimization of dosage regimens across diverse patient subgroups, improving therapeutic outcomes and reducing toxicity risks.

4.2 Pharmacogenomics

Pharmacogenomics investigates how genetic differences influence drug response. AI accelerates this process by integrating genomic, transcriptomic, and clinical data to identify genotype-phenotype associations. Techniques such as Random Forests, support vector machines (SVM), and neural networks are used to classify responders and non-responders to specific treatments (31). AI has facilitated the discovery of pharmacogenomic biomarkers, such as CYP2C9 and VKORC1 variants for warfarin dosing, supporting personalized therapy (32). Moreover, AI-based models can suggest genetic signatures associated with adverse drug reactions or altered metabolism, laying the groundwork for precision medicine.

4.3 Pharmacovigilance

Traditional pharmacovigilance relies on spontaneous reporting systems, which are often slow and underreported. AI, particularly natural language processing (NLP) and machine learning (ML), transforms pharmacovigilance by enabling real-time analysis of adverse drug events (ADEs) from diverse sources such as electronic health records (EHRs), clinical notes, social media, and regulatory databases (33). NLP can extract drug-event relationships from unstructured texts, while ML classifies and predicts the severity and frequency of side effects (34). Platforms like MedWatcher Social and FDA’s Sentinel system have incorporated AI-based monitoring, improving early detection of safety signals and public health responses (35).

5. PROMINENT TOOLS AND PLATFORMS

DeepChem

DeepChem is an open-source Python-based library that provides tools for applying deep learning to drug discovery, materials science, and quantum chemistry. It supports molecular featurization, graph convolutional models, and multitask learning, enabling rapid development of predictive models for molecular properties, bioactivity, and ADMET characteristics (36). Its modular structure allows researchers to implement state-of-the-art algorithms and benchmark them across multiple datasets.

AtomNet

AtomNet is a pioneering deep convolutional neural network (CNN) architecture developed by Atomwise for structure-based drug design. It predicts binding affinities by analyzing the spatial configuration of atoms within target-ligand complexes. AtomNet has been used in virtual screening campaigns for diseases including Ebola and multiple cancers, demonstrating high-throughput prediction of molecular interactions with protein binding sites (37).

AlphaFold

Developed by DeepMind, AlphaFold is a revolutionary deep learning system that predicts protein 3D structures from amino acid sequences with atomic-level accuracy. Alpha Fold has solved structures for nearly every known human protein, greatly enhancing structure-based drug discovery by providing structural information that was previously inaccessible via experimental methods (38). Its open-source implementation, AlphaFold2, has been widely adopted by both academia and industry.

Commercial Platforms: Schrödinger, BioXcel, Insilico Medicine

Several commercial platforms have successfully integrated AI into pharmaceutical R&D:

• Schrödinger combines physics-based modeling with ML algorithms to accelerate drug lead identification and optimization. Its platform is used for virtual screening, molecular docking, and ADMET prediction (39).
• BioXcel Therapeutics uses AI to repurpose and develop drugs for neuroscience and immuno-oncology. Their AI engine analyzes multi-modal data to identify new therapeutic opportunities (40).
• Insilico Medicine utilizes generative adversarial networks (GANs) and reinforcement learning to design novel molecules. Its end-to-end AI drug discovery platform has led to the identification of clinical candidates in record time (41).

6. CASE STUDIES

6.1 AlphaFold 2 by DeepMind (2021)

In a landmark achievement, DeepMind’s AlphaFold 2 solved the protein structure prediction problem with near-experimental accuracy. Leveraging advanced deep learning techniques such as attention mechanisms and geometric reasoning, AlphaFold 2 outperformed all other methods at the 14th Critical Assessment of protein Structure Prediction (CASP14) (42). The ability to predict three-dimensional protein structures solely from amino acid sequences has significantly accelerated target identification in drug discovery by enabling structural understanding of previously uncharacterized proteins (43). The freely available AlphaFold Protein Structure Database provides access to structures for nearly all human proteins, facilitating both basic biology and drug development efforts.

6.2 DSP-1181: AI-Designed Drug from Exscientia and Sumitomo Dainippon Pharma

DSP-1181 is the first drug molecule designed using artificial intelligence to enter human clinical trials. Developed collaboratively by Exscientia and Sumitomo Dainippon Pharma, DSP-1181 is a serotonin 5-HT1A receptor agonist intended to treat obsessive-compulsive disorder (OCD). Using AI, the team reduced the drug design cycle to less than 12 months—far shorter than traditional timelines (44). Exscientia’s AI platform integrated data on chemistry, pharmacology, and clinical outcomes to generate and optimize the molecular structure, setting a precedent for AI-enabled de novo drug design and candidate selection.

6.3 IBM Watson for Drug Discovery

IBM Watson applied natural language processing (NLP) and machine learning to mine vast quantities of biomedical literature and databases. Its system was used to identify novel drug-disease associations by correlating disparate pieces of biomedical information. For instance, it helped uncover potential drug candidates for glioblastoma by analyzing published literature and inferring connections between genes, pathways, and drugs that human researchers might miss (45). Watson’s ability to synthesize knowledge from unstructured data sources has influenced research pipelines in drug repurposing and target discovery.

7. CHALLENGES AND LIMITATIONS

7.1 Data Quality and Integration

A fundamental challenge in AI-driven drug discovery is the quality and heterogeneity of biomedical data. Data often come from disparate sources—genomic databases, electronic health records (EHRs), clinical trials, and literature—each with its own structure, format, and bias (46). Many datasets are incomplete, unstructured, or mislabeled, which can significantly degrade model performance and reproducibility. The integration of multi-omics data and longitudinal patient information remains an ongoing technical and computational hurdle (47).

7.2 Model Interpretability

While deep learning models, such as convolutional and recurrent neural networks, provide high predictive accuracy, they often function as “black boxes,” lacking transparent reasoning pathways (48). This opacity raises concerns in healthcare, where decisions must be justified to regulators, clinicians, and patients. Tools like SHAP (SHapley Additive exPlanations) and LIME (Local Interpretable Model-agnostic Explanations) are being developed to enhance interpretability, but widespread implementation remains limited (49).

7.3 Regulatory Barriers

There is currently a lack of clear regulatory frameworks for the approval of AI-generated drug candidates. Regulatory agencies like the FDA and EMA are still adapting to the rapidly evolving AI landscape, struggling to define validation standards, safety thresholds, and responsibility for AI-driven decisions (50). This uncertainty can delay or complicate the adoption of AI tools in official drug pipelines.

7.4 Ethical Issues

Ethical considerations include data privacy, especially when using sensitive genetic or patient health information. Moreover, algorithmic bias—stemming from non-representative training data—can lead to unequal outcomes across demographic groups (51). Another major issue is equitable access to AI-powered drug discovery technologies, which are often concentrated in wealthier institutions or countries, potentially widening global health disparities (52).

8. FUTURE PROSPECTS

• Federated Learning: Enables decentralized training across institutions while preserving data privacy.
• Explainable AI (XAI): Making AI decisions interpretable to enhance trust and transparency.
• Integration with Quantum Computing: Promises exponential speed-ups in molecular simulations.
• AI in Polypharmacology: Designing multi-target drugs for complex diseases like cancer and Alzheimer’s.

9. CONCLUSION

Artificial intelligence has ushered in a new era of efficiency and innovation in drug discovery and pharmacology. While challenges remain, the convergence of AI with omics, real-world data, and high-throughput screening is laying the foundation for a more precise, cost-effective, and patient-centric drug development paradigm. A multidisciplinary approach, combining domain expertise with robust AI methodologies, will be pivotal in realizing the full potential of this technological revolution.

REFERENCES

1. 1. 1. 1. DiMasi JA, Grabowski HG, Hansen RW. Innovation in the pharmaceutical industry: New estimates of R&D costs. J Health Econ. 2016;47:20–33.
         2. Vamathevan J, Clark D, Czodrowski P, et al. Applications of machine learning in drug discovery and development. Nat Rev Drug Discov. 2019;18(6):463–477.
         3. Chen H, Engkvist O, Wang Y, Olivecrona M, Blaschke T. The rise of deep learning in drug discovery. Drug Discov Today. 2018;23(6):1241–1250.
         4. Zhang YF, Chen Y, Li W, Wang J, Tang Y. Applications of natural language processing in drug development. Drug Discov Today. 2021;26(2):484–495.
         5. Kalinin AA, Athey BD, Dinov ID. Biomedical data science and engineering for personalized medicine. J Am Med Inform Assoc. 2018;25(3):274–278.
         6. Jumper J, Evans R, Pritzel A, et al. Highly accurate protein structure prediction with AlphaFold. Nature. 2021;596(7873):583–589.
         7. Zhavoronkov A, Ivanenkov YA, Aliper A, et al. Deep learning enables rapid identification of potent DDR1 kinase inhibitors. Nat Biotechnol. 2019;37(9):1038–1040.
         8. Mak K-K, Pichika MR. Artificial intelligence in drug development: present status and future prospects. Drug Discov Today. 2019;24(3):773–780.
         9. Hopkins AL, Groom CR, Alex A. Ligand efficiency: a useful metric for lead selection. Drug Discov Today. 2004;9(10):430–431.
         10. Beck BR, Shin B, Choi Y, Park S, Kang K. Predicting commercially available antiviral drugs that may act on the novel coronavirus (SARS-CoV-2) through a drug-target interaction deep learning model. Comput Struct Biotechnol J. 2020;18:784–790.
         11. Chen H, Engkvist O, Wang Y, Olivecrona M, Blaschke T. The rise of deep learning in drug discovery. Drug Discov Today. 2018;23(6):1241–1250.Grys BT, Lo DS, Sahin N, et al. Machine
         12. and computer vision approaches for phenotypic profiling. J Cell Biol. 2017;216(1):65–71.
         13. Ching T, Himmelstein DS, Beaulieu-Jones BK, et al. Opportunities and obstacles for deep learning in biology and medicine. J R Soc Interface. 2018;15(141):20170387.
         14. Öztürk H, Özgür A, Ozkirimli E. DeepDTA: Deep drug–target binding affinity prediction. Bioinformatics. 2018;34(17):i821–i829.
         15. Lee J, Yoon W, Kim S, et al. BioBERT: A pre-trained biomedical language representation model for biomedical text mining. Bioinformatics. 2020;36(4):1234–1240.
         16. Wang Q, Mao Z, Wang B, Guo L. Knowledge graph embedding: A survey of approaches and applications. IEEE Trans Knowl Data Eng. 2017;29(12):2724–2743.
         17. Kadurin A, Nikolenko S, Khrabrov K, et al. druGAN: An advanced generative adversarial autoencoder model for de novo generation of new molecules with desired molecular properties in silico. Mol Pharm. 2017;14(9):3098–3104.
         18. Gómez-Bombarelli R, Wei JN, Duvenaud D, et al. Automatic chemical design using a data-driven continuous representation of molecules. ACS Cent Sci. 2018;4(2):268–276.
         19. Li Y, Wang J, Wang L, et al. Deep learning in bioinformatics: Introduction, application, and perspective in the big data era. Methods. 2019;166:4–21.
         20. Zeng X, Song X, Ma T, et al. Repurpose open data to discover therapeutics for COVID-19 using deep learning. J Proteome Res. 2020;19(11):4624–4636.
         21. Gomes J, Ramsundar B, Feinberg EN, et al. Atomic convolutional networks for predicting protein–ligand binding affinity. arXiv preprint. 2017. [arXiv:1703.10603]
         22. Pereira JC, Caffarena ER, dos Santos CN. Boosting docking-based virtual screening with deep learning. J Chem Inf Model. 2016;56(12):2495–2506.
         23. Brown AS, Patel CJ. A standard database for drug repositioning. Sci Data. 2017;4:170029.
         24. Richardson P, Griffin I, Tucker C, et al. Baricitinib as potential treatment for 2019-nCoV acute respiratory disease. Lancet. 2020;395(10223):e30–e31.
         25. Zhang W, Liu F, Chen Y, et al. Predicting human intestinal absorption with ensemble learning based QSAR modeling. J Comput Chem. 2019;40(2):230–239.
         26. Cheng F, Li W, Zhou Y, et al. admetSAR: A comprehensive source and free tool for assessment of chemical ADMET properties. J Chem Inf Model. 2012;52(11):3099–3105.
         27. Weng C, Shah NH, Hripcsak G. Deep phenotyping: Embracing complexity and temporality—Towards scalability, portability, and interoperability. J Biomed Inform. 2020;105:103433.
         28. Liu R, Rizzo S, Whalen DM, et al. Evaluating eligibility criteria of oncology trials using real-world data and AI. Nature. 2021;592(7855):629–633.
         29. Polak S, Wyska E, Pendri A, et al. Application of machine learning in PK/PD modeling: Current status and future perspectives. CPT Pharmacometrics Syst Pharmacol. 2020;9(10):587–597.
         30. Tsamandouras N, Rostami-Hodjegan A, Aarons L. Combining the ‘bottom up’ and ‘top down’ approaches in pharmacokinetic modelling: Fitting PBPK models to observed clinical data. Br J Clin Pharmacol. 2015;79(1):48–55.
         31. Li B, Tang J, Yang Q, et al. An ensemble approach to predict gene–drug interactions. Bioinformatics. 2019;35(14):2391–2399.
         32. Luzum JA, Pakyz RE, Elsey AR, et al. The pharmacogenomics research network translational pharmacogenetics program: Outcomes and metrics of pharmacogenetic implementations across diverse healthcare systems. Clin Pharmacol Ther. 2017;102(3):502–510.
         33. Harpaz R, DuMouchel W, Shah NH, et al. Novel data-mining methodologies for adverse drug event discovery and analysis. Clin Pharmacol Ther. 2012;91(6):1010–1021.
         34. Wang Y, Wang L, Rastegar-Mojarad M, et al. Clinical information extraction applications: A literature review. J Biomed Inform. 2018;77:34–49.
         35. Freifeld CC, Brownstein JS, Menone CM, et al. Digital drug safety surveillance: Monitoring pharmaceutical products in Twitter. Drug Saf. 2014;37(5):343–350.
         36. Ramsundar B, Eastman P, Walters P, et al. Deep Learning for the Life Sciences: Applying Deep Learning to Genomics, Microscopy, Drug Discovery, and More. O’Reilly Media; 2019.
         37.   Wallach I, Dzamba M, Heifets A. AtomNet: A deep convolutional neural network for bioactivity prediction in structure-based drug discovery. arXiv preprint. 2015. [arXiv:1510.02855]
         38.   Jumper J, Evans R, Pritzel A, et al. Highly accurate protein structure prediction with AlphaFold. Nature. 2021;596(7873):583–589.
         39.   Friesner RA, Murphy RB, Repasky MP, et al. Extra precision glide: Docking and scoring incorporating a model of hydrophobic enclosure for protein–ligand complexes. J Med Chem. 2006;49(21):6177–6196.
         40. Krishnan SP, Gordon GJ, Hossain I, et al. BioXcel Therapeutics: Applying AI to drug repurposing and development. Drug Discov Today. 2020;25(8):1422–1427.
         41. Zhavoronkov A, Ivanenkov YA, Aliper A, et al. Deep learning enables rapid identification of potent DDR1 kinase inhibitors. Nat Biotechnol. 2019;37(9):1038–1040.
         42. Jumper J, Evans R, Pritzel A, et al. Highly accurate protein structure prediction with AlphaFold. Nature. 2021;596(7873):583–589.
         43. Tunyasuvunakool K, Adler J, Wu Z, et al. Highly accurate protein structure prediction for the human proteome. Nature. 2021;596(7873):590–596.
         44. Mak K-K, Pichika MR. Artificial intelligence in drug development: present status and future prospects. Drug Discov Today. 2019;24(3):773–780.Ferrucci D, Levas A, Bagchi S, et al. Watson: Beyond Jeopardy! Artif Intell. 2013;199:93–105.
         45. Beam AL, Kohane IS. Big data and machine learning in health care. JAMA. 2018;319(13):1317–1318.
         46.  Wang L, Reddy CK. Machine learning for survival analysis: A survey. ACM Comput Surv. 2020;51(6):1–36.
         47. Rudin C. Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead. Nat Mach Intell. 2019;1(5):206–215.
         48. Ribeiro MT, Singh S, Guestrin C. “Why should I trust you?” Explaining the predictions of any classifier. In: Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. 2016:1135–1144.
         49. Topol EJ. High-performance medicine: The convergence of human and artificial intelligence. Nat Med. 2019;25(1):44–56.
         50. Obermeyer Z, Powers B, Vogeli C, Mullainathan S. Dissecting racial bias in an algorithm used to manage the health of populations. Science. 2019;366(6464):447–453.
         51. Floridi L, Cowls J, Beltrametti M, et al. AI4People—An ethical framework for a good AI society: Opportunities, risks, principles, and recommendations. Minds Mach. 2018;28(4):689–707.