Accelerating Drug Discovery and Development Through Artificial Intelligence: Challenges and Opportunities

Abstract

The pharmaceutical industry is transforming by incorporating Artificial Intelligence (AI) in drug discovery and development. This integration offers new possibilities to expedite the discovery of potential therapeutic agents, enhance clinical trial planning, and improve treatment customization. AI-based technologies, such as machine learning (ML), deep learning (DL), and natural language processing (NLP), facilitate more precise predictions of drug effectiveness, toxicity, and patient outcomes by examining extensive biological and chemical datasets. These innovations substantially decrease the duration and expenses associated with introducing new medications to the market. Nevertheless, despite the encouraging prospects, various obstacles persist. Concerns, including data integrity, algorithm transparency, model interpretability, and regulatory challenges, need to be tackled to ensure widespread implementation and efficacy of AI in this field. This overview examines the transformative effects of AI on drug discovery.

Keywords

Introduction

Over the past few years, there has been an exceptional increase in digitalization in the pharmaceutical segment. Be that as it may, this digitalization comes with the challenge of securing, scrutinizing, and applying that information to illuminate complex clinical issues. This spurs the utilization of AI since it can handle expansive volumes of information with improved automation. AI is a technology-based framework including different advanced devices and systems that can imitate human insights. At the same time, it does not undermine to supplant of human physical presence. AI utilizes frameworks and programs to decipher and learn from the input information to make free choices. [1,2,3,4,5]

The agent AI strategy, machine learning (ML), engages machines to learn from existing information utilizing factual approaches and make forecasts, which can be advanced classified into supervised, unsupervised, and semi-supervised learning. Profound learning (DL), a subset of ML, centers on utilizing multi-layered artificial neural organize (ANN) structures to mimic the human brain's neural systems for learning information representations, making it more capable and adaptable in taking care of complex and high-dimensional information. With the preferences of moo-fetched and quick speed, the ML approaches are revolutionizing and strengthening different stages of sedate discovery, such as target distinguishing proof, de novo medicate design, and medicate repurposing. For example, DL-based open-source devices, such as DeepDTAF and Profound Fondness, have been connected to anticipate the official partiality of drug-target intelligence (DTIs), making the chase for unused pharmaceuticals more productive. [5,6,7,]

Despite progressions in such conventional strategies, the complex, time-consuming, and exorbitant nature of sedate disclosure endures. Normally, creating a sedate from normal sources or imaginative methods can take 10–15 years and fetch over $2.5 billion. In reaction to these challenges, the sedate improvement industry, driven by the requirement to optimize costs and effectiveness, has progressively turned towards artificial intelligence (AI), especially machine learning (ML) and deep learning (DL). AI is the recreation of human insights in machines modified to think and act like people. ML includes the utilization of calculations to imitate the usefulness of the human brain to perform various tasks. DL is a more progressed subset of ML that employments multi-layered neural systems to productively distinguish complex patterns in data. This move marks an unused period in medical revelation. [8,9]

A history of counterfeit insights in healthcare: to begin with, the breakthrough of manufactured insights in healthcare came in 1950 with the improvement of turning tests. Afterward, in 1975, the to begin with inquiry about the use of computers in drugs was created, followed by the NIH to begin with central Point workshop, which stamped the significance of manufactured intelligence in healthcare. With the advancement of profound learning in the 2000s and the presentation of DeepQA in 2007, the scope of counterfeit insights in healthcare has expanded. In 2010, CAD was connected to endoscopy, to begin with, though in 2015, the to begin with Pharm Bot was created. In 2017, the to begin with FDA-approved cloud-based DL application was presented, which moreover stamped the usage of manufactured learning in healthcare. From 2018 to 2020, a few AI trials in gastroenterology were performed. A Classification of Manufactured Insights: there are seven classifications of manufactured insights, which are thinking and problem-solving, information representation, arranging and social insights, discernment, machine learning, mechanical technology: movement and control, and characteristic dialect handling, as examined by Russel and Norvig in their book “Artificial Insights: A Advanced Approach.” [10]

Fig 1: artificial intelligence in healthcare and pharmaceutical industry

Stages of the conventional sedate disclosure and improvement process:

• Target identification
• Target validation
• lead identification
• lead optimization
• Item characterization
• Detailing and development

• Preclinical research
• Investigational Unused Drug
• Clinical trials
• Modern Medication Application
• Approval

the preparation of medicate advancement clinical trials. Sedate disclosure and improvement is a costly process due to the high budgets of R&D and clinical trials. It takes nearly 12-15 a long time to create a single modern medicate particle from the time it is found when it is accessible in the advertisement for treating patients.[2] The normal toll for investigation and improvement for each solid sedate is likely to be $900 million to $2 billion. This figure incorporates the toll of the thousands of disappointments: For each 5,000-10,000 compounds that enter the examination and advancement pipeline, eventually it were one eventually achieves endorsement. The conventional sedate disclosure handle is exceptionally time-consuming."[11]

AI to Accelerate and Improve Drug Discovery and Development:

• Speeding Up Target Identification and Validation
• Enhancing Lead Identification and Optimization
• Reducing Preclinical and Clinical Trial Costs
• Optimizing Drug Formulation and Delivery
• AI in Real-time Monitoring During Clinical Trials
• Speeding Up Regulatory Approval
• Personalized Medicine
• Drug Repurposing

1. Applications of AI in Medicinal Discovery:

1. Target Distinguishing Proof and Approval: AI has affected the field of medicate revelation, especially in target recognizable proof and approval zones. This handle includes recognizing potential organic targets and illustrating their parts in infections, taken after by approving these targets to guarantee they are specifically included in an illness component and that the balance of the target is likely to have a helpful impact and plays a vital part in recognizing potential sedate tar Life 2024, 14, x FOR PEER Survey gets by analyzing the genomic, proteomic, and metabolomic information. ML calculations 5 of 38 filters through huge datasets to pinpoint the proteins or organic pathways involved in particular infections, advertising analysts’ profitable experiences for medication development.[12]

ML-based approaches, such as Kronecker regularized least squares (KronRLS), evaluate the resemblances between drugs and protein particles to decide DTBA. Correspondingly, Sim Boost utilizes relapse trees to anticipate DTBA and considers both feature-based and similarity-based interactions. Target approval is a vital step in sedate disclosure since it guarantees that an atomic target is straightforwardly included in an infection component and that the balance of the target is likely to have a restorative impact [13]. Target approval may include deciding the structure-activity relationship, the hereditary control of target qualities (knockdown or overexpression), creating a drug-resistant mutant of the assumed target, utilizing degradation-based devices to predict the impacts of the target, and observing the signaling pathways downstream of the assumed target. AI has been utilized to foresee drugs intuitively, degree the official partiality of medication, and select and approve targets [14,15,16,17]

Example: AI has been used to identify potential drug targets for complex diseases such as cancer and Alzheimer's, by processing multi-dimensional omics data and correlating gene expression patterns with disease pathways [ QiuY, et al, 2024]

2. Virtual screening of compound libraries:

AI-powered virtual screening instruments analyze the three-dimensional structures of target proteins and foresee how potential sedate atoms would associate with them. This speeds up the handling of sedate planning and permits analysts to recognize promising sedate candidates for advanced testing. [19]

Figure 2: Virtual screening ML model in the drug development process

Calculations, such as Nearest-Neighbour classifiers, RF, extraordinary learning machines, SVMs, and profound neural systems (DNNs), are utilized for VS based on amalgamation possibility and can also anticipate in vivo movement and poisonous quality. A few biopharmaceutical companies, such as Bayer, Roche, and Pfizer, have joined up with IT companies to create a stage for the disclosure of treatments in areas such as immuno-oncology and cardiovascular diseases.[20] Illustration: Profound learning models, such as convolutional neural systems (CNNs), have been utilized to screen compound libraries for potential antiviral agents against SARS-CoV-2, recognizing promising candidates within days.[21] [ Kandhare P, et al 2025]

3. Prediction of drug-target interactions:

While making a steady molecule, it is crucial to assign the correct target for productive treatment. Different proteins are included in the enhancement of the ailment, and, in a few cases, they are overexpressed. In this way, for specific targeting of illness, it is vital to anticipate the structure of the target protein to arrange the calm particle. AI can offer assistance in structure-based sedate disclosure by foreseeing the 3D protein structure since the arrange with the chemical environment of the target protein area, thus making a distinction to expect the effect of a compound on the target along with security considerations a few times as of late their mix or era [22] Walk 9, 2021 Particle sharp, a pioneer in utilizing made bits of knowledge (AI) for small -molecule sedate disclosure, these days announced it has been named to Fast Company's prestigious annually list of the World's Most Creative Companies for 2021.

Example: Atomwise was recognized for its unmatched AI arrange, AtomNet®?, as well as its developing pipeline and portfolio of organizations and joint ventures taking care of a few of the most challenging targets in small-molecule sedate development.[23]

4. De novo drug design: De novo design organizes and solidifies, making unused compounds with specific pharmacological and physicochemical characteristics from the ground up, and is diligently being energized by AI-based strategies, which are commonly utilized for nuclear period errands. Yi Tooth and colleagues have recently shown their novel method, known as the Quality Assessment-Based Unfaltering Coordinate (QADD) approach. This creative approach utilizes a multi-objective basic fortress learning approach to particularly make particles that have a few locks in characteristics. Their thinking around publicized encounters into the capability of QADD in optimizing specific nuclear properties and making particles with tall, calm potential.[24]One of the challenges we go up against is that de novo generators based on groupings routinely make invalid yields. A brief time afterward, consider proposed a course of action by utilizing post hoc post hoc correction.

Examples: quantitative structure-activity relationships (QSAR) and quantitative structure-property relationships (QSPR) models, structure-based models for warmth figures, heuristics for the calculation of physicochemical properties, and combinations thereof for multi-objective assignments. [ Schoenmaker L, et al 2023][25]

1. Optimization of Lead Compounds: AI-driven approaches in lead optimization use computational strategies and ML calculations to increase and accelerate the medicinal discovery process. These strategies tackle the control of enormous information and predictive modeling to analyze huge datasets of chemical structures, organic reactions, and test results. By learning designs and connections from this information, AI calculations can predictively direct lead optimization endeavors, prioritize promising candidates, and indeed propose novel chemical platforms with upgraded properties.[26]

Key techniques are essential for maximizing the adequacy and proficiency of lead optimization endeavors. Information integration and harmonization are foundational, guaranteeing that different datasets from exploratory measures, chemical databases, and writing are combined cohesively. Highlighting designing and representation learning play basic parts in capturing compound properties and natural exercises, with methods like graph-based representations and atomic embeddings extricating significant highlights naturally. Demonstrating determination is foremost, with different models like Profound neural systems (DNN) and gathering strategies advertising special focal points. Dynamic learning and test plan procedures iteratively select instructive compounds for testing, quickening the investigation of chemical space. Interpretability and explainability are imperative for understanding show expectations and cultivating collaborative strategies. [27,28,29]

1. Applications of AI in Drug Development:

1. Prediction of Drug Efficacy and Safety: Drug safety is a major challenge in bringing new drugs to market. Unexpected toxicities are a major source of attrition during clinical trials, and post-marketing safety concerns cause unnecessary morbidity and mortality. Adverse events (AEs), or adverse drug reactions (ADRs) when causality is demonstrated, are unexpected effects occurring from a normal dosage of the drug.[30]

• Pre-clinical Medication Security: AI strategies have appeared to play a critical part in pre-market sedate safety, particularly in the area of toxicity assessment. Medicinal quality assurance is a primary step in sedate plans and includes identifying the AEs of chemicals on people, plants, animals, and the environment [31]. Pre-clinical assessments are needed to prevent harmful drugs from coming to clinical trials. Despite this, tall poisonous quality is still a major contributor to sedate disappointment bookkeeping for two-thirds of post-market sedate withdrawals [32] and one-fifth of disappointments amid clinical trials [33]. In this way, precise poisonous quality gauges are essential for guaranteeing medication security and can help decrease the fetched and marketing time of bringing modern drugs to advertise. Creature ponders have truly been the most ordinary approach taken to survey poisonous quality.

2. Plan and Optimization of Clinical Trials: As AI cements its presence, the clinical trial scene is balanced for change. Victory lies in recognizing destinations with vital agents able to pull in high-quality subject assembly pattern clinical trial necessities. AI models, beneath the heading of CRO [Transformation Rate Optimization] Pro specialists, can fastidiously analyze information and rank outcomes, teach, destinations, examiners, nations, and geos, enabling the trial support to zero in on the right destinations. This makes a difference in the consistent engagement and conduct of the trial. By grasping AI-enabled capabilities, biopharma companies.

• Past enrolment measurements: AI models can rapidly scrutinize chronicled data focuses related to persistent enrolment in past clinical trials. This examination predicts enrolment rates for future trials and makes a difference in optimizing the arrangement and asset allocation.
• Persistent get-to information: By distinguishing potential obstructions to enrolment and recommending ways to make strides quiet get to by utilizing data almost patients’ capacity to get to trial destinations, their well-being status, and other pertinent factors.
• Information accessible on hardware: Progressed AI arrangements can offer assistance to guarantee that trials are well-equipped and foresee and avoid potential equipment-related issues by getting data about the restorative hardware utilized in trials, such as its availability, usefulness, and usage.[34]

Example: AI has been utilized to optimize the clinical trial plan for oncology drugs by foreseeing understanding reactions based on hereditary profiles and malady subtypes [Quazi S, et al, 2022][35]

3. Patient stratification for personalized medicine: inquire has illustrated that understanding stratification models, driven by AI/ML modeling, can be extended to other viral diseases, and might be utilized to accurately recognize the signs of clinical biomarkers, coming about in more exact analysis and treatment alternatives in the setting of personalized medicine.[36]

The thought behind personalized medication is to tailor medications for the anticipation and treatment of maladies to the person, or maybe than utilizing a one-size-fits-all approach. The total sequencing of the human genome in 2003 encouraged the assisted development of personalized pharmaceuticals, moving past the genome into the whole range of atomic pharmaceuticals. Progresses in innovation, such as DNA proteomics, imaging conventions, and remote well-being observing gadgets, have contributed to critical enhancements in personalized pharmaceuticals. Be that as it may, there are still challenges to overcome, such as the requirement for more pertinent models based on human cell societies and the personalization of treatment. [37,38]

The integration of AI-derived experiences into scheduled persistent care has suggestions for healthcare suppliers, pharmaceutical companies, and administrative bodies. Healthcare suppliers can utilize AI to achieve persistent results, decrease healthcare costs, and upgrade understanding of security. Pharmaceutical companies can utilize AI to speed up medication discovery, approve medication targets quickly, and enhance treatments. Administrative bodies can utilize AI to assess medication security and adequacy, streamline medication endorsement forms, and guarantee understanding of safety [46]. AI-powered pharmacogenomics is changing clinical decision-making by giving personalized treatment and overseeing care. AI-derived experiences can be coordinated into scheduled quiet care to move forward with sedation choice, dose, and security. The integration of AI-derived bits of knowledge into scheduled primary care has suggestions for healthcare suppliers, pharmaceutical companies, and administrative bodies.[39]

4. Real-World Prove Examination: Real-world proof is pivotal to understanding the dissemination of unused oncologic treatments, checking cancer results, and recognizing unforeseen toxicities. In-home, proof is challenging to collect quickly and comprehensively, frequently requiring costly and time-consuming manual case finding and explanation of clinical content. In this Survey, we outline later advancements in the utilization of fake insights to collect and analyze real-world evidence in oncology. Artificial insights (AI) approaches are progressively utilized to effectively phenotype patients and tumors at an expansive scale. These instruments, too, may give novel natural bits of knowledge and move forward hazard forecasts through multimodal integration. of radiographic, obsessive, and genomic datasets. Custom dialect-preparing pipelines and expansive dialect models hold extraordinary guarantees for clinical prediction.[40]

FUTURE DIRECTIONS:

1. Integration of Multi-Omics Information for Comprehensive Medication Revelation:  To comprehend complex organic forms comprehensively, it is essential to take an integrator approach that combines multi-omics information to highlight the interrelationships of the included biomolecules and their capacities. With the appearance of high-throughput strategies and the accessibility of multi-omics information produced from an expansive set of tests, a few promising instruments and strategies have been created for information integration and translation. In this survey, we collected the instruments and strategies that receive an integrator approach to analyze numerous omics information and summarized their capacity to address applications such as illness subtyping, biomarker forecasting, and determining bits of knowledge from the information. We give the strategy, utilize cases, and restrictions of these devices; a brief account of multi-omics information stores and visualization entries; and challenges related to multi-omics information integration. Examples: multi-omics, information integration, illness subtyping, biomarker expectation, and information repositories.[41]
2. Made strides in AI Models for foreseeing the Sedate Digestion system and Toxicity:

In later a long time, artificial intelligence (AI) has risen as a capable instrument for anticipating medicate digestive system and diseases, advertising the potential to speed up sedate development and advance clinical victory rates. This survey highlights later progress in AI-based medicate diagnosis systems and disease prediction, including deep learning and machine learning calculations. give a list of open information sources and free expectation apparatuses for the investigative community.[42]

3. AI-Driven Automation of Laboratory Processes:

AI-driven automation of laboratory processes streamlines drug discovery and development by accelerating tasks, optimizing workflows, and improving efficiency, leading to faster timelines and reduced costs. 

4. Progressed Characteristic Tongue Organizing for Mining Dependable Literature:

Natural Lingo Organizing (NLP) and its broader recommendations, challenges, and future directions. With the ever-increasing volume of substance data passed on each day from organized sources, filtering related and useful information is getting more complex. Standard manual strategies for supervising and analyzing composed information are troublesome and slight to errors, underscoring the requirement for effective automated choices. The headways in Standard Tongue Orchestrating (NLP), particularly in transformer-based models and vital learning methods, have laid out the imperative potential in advancing the precision and consistency of unmistakable NLP applications. This work presents a novel approach that combines capable review strategies with progressed NLP approaches to redesign the common reasonability of NLP systems. The proposed methodology guarantees an organized and clear composing review process, resulting in more edifying and reasonably related outcomes. the mining of colossal volumes of coherent composing to recognize essential considerations, clinical trials, and drug-related information in real time.[43]

5. Improvement of Interpretable AI Models for Administrative Compliance: The expanding complexity of compliance and administrative systems for businesses requires imaginative arrangements for overseeing and translating expansive volumes of information. Reasonable Fake Insights (XAI) offers a promising approach by giving straightforward and interpretable AI models, in high-stakes spaces like finance, healthcare, and legal segments, where responsibility and belief are foremost. XAI addresses these challenges by making the decision-making prepare more straightforward, empowering partners to understand the rationale behind AI-driven activities. suggestions and guaranteeing compliance with laws and approaches. In addition, the integration of XAI into compliance models upgrades audibility and traceability, giving controllers and reviewers the tools to approve and confirm adherence to benchmarks. This straightforwardness is significant for building belief in AI frameworks and cultivating collaboration between decision-makers and AI tools.[44]

Limitations and Challenges:

1. Information Quality and Availability: Data quality challenges in pharmaceuticals include ensuring the exactness and consistency of information collected from different sources. The industry bargains with enormous datasets, making information management and integration complex. Administrative compliance requires thorough information approval, but keeping up with information judgment can be challenging due to advancing guidelines. Additionally, human mistakes in the information section and potential inclinations in information collection can influence the unwavering quality of investigative discoveries, affecting stable improvement, quiet security, and administrative approvals.

[https://www.elucidata.io/blog/the-challenges-of-data-quality-in-drug-discovery-managing-complexity-for-better-innovations]

2. Need for Standardization in Information Designs and Protocols: Data standardization refers to the handling of changing information from distinctive groups and sources into a standardized arrangement. This consistency is vital for exact information investigation and reporting, as it eliminates disparities and guarantees that information from diverse frameworks can be effortlessly compared and coordinated. AI-powered information standardization takes this a step, encouraged by computerizing the process, lessening manual exertion, and altogether moving forward the speed and precision of information standardization tasks. Inconsistent information structures and the nonappearance of standardization between databases and inquiry about stages make major obstacles in joining AI models into static disclosure pipelines. Sedate disclosure involves a part of heterogeneous information, such as genomic information, clinical trial results, chemical structures, and pharmacological information, all of which are regularly stored in changed formats.

[https://www.espire.com/blog/posts/ai-powered-data-standardization-the-ultimate-guide]

3. Bias in Artificial Intelligence Systems: The nonappearance of inclination, or equally the presence of decency, is basic to guaranteeing that AI frameworks work morally and even-handedly. In information science, predisposition alludes to any efficient mistake or deviation from the genuine value in information collection, arrangement, or investigation. This can emerge due to different components such as inadequate information, skewed examination strategies, or mistakes in information recording. [45]

CONCLUSION

Artificial Intelligence is poised to transform the drug discovery and development space by providing faster, less expensive, and more precise alternatives to conventional methods. Although AI holds many opportunities for enhancing drug design, clinical trials, and patient outcomes, some challenges persist, especially those related to data quality, regulatory compliance, and ethics. To realize the full potential of AI in pharma, sustained research, interdisciplinary collaboration, and close attention to these challenges will be required. As AI keeps developing, its influence on medicine and the pharmaceutical sector is going to be revolutionary, marking the beginning of a new age of precision medicine and novel drug treatments.

REFERENCES

1. Paul D, Sanap G, Shenoy S, Kalyane D, Kalia K, Tekade RK. Artificial intelligence in drug discovery and development. Drug Discovery Today. 2021 Jan;26(1):80-93. doi: 10.1016/j.drudis.2020.10.010.
2. 1 Ramesh, A. et al. (2004) Artificial intelligence in medicine. Ann. R. Coll. Surg. Engl. 86, 334–338.
3. Miles, J. and Walker, A. (2006) The potential application of artificial intelligence in transport. IEE Proc.-Intell. Transport Syst. 153, 183–198.
4. Yang, Y. and Siau, K. (2018) A Qualitative Research on Marketing and Sales in the Artificial Intelligence Age. MWAIS.
5. Wirtz, B.W. et al. (2019) Artificial intelligence and the public sector—applications and challenges. Int. J. Public Adm. 42, 596–615.
6. Qi X, Zhao Y, Qi Z, Hou S, Chen J. Machine learning empowering drug discovery: applications, opportunities and challenges. Molecules. 2024 Apr;29(4):903. doi: 10.3390/molecules29040903.
7. Deng, J.; Yang, Z.; Ojima, I.; Samaras, D.; Wang, F. Artificial intelligence in drug discovery: Applications and techniques. Brief. Bioinform 2022, 23, bbab430. [CrossRef] [PubMed]
8. Mak, K.K.; Pichika, M.R. Artificial intelligence in drug development: Present status and prospects. Drug Discov. Today, 2019, 24, 773–780. [CrossRef].
9. Chen, H.; Engkvist, O.; Wang, Y.; Olivecrona, M.; Blaschke, T. The rise of deep learning in drug discovery. Drug Discov. Today, 2018, 23, 1241–1250. [CrossRef]
10. Gupta R, Srivastava D, Sahu M, Tiwari S, Ambasta RK, Kumar P. Artificial intelligence to deep learning: machine intelligence approach for drug discovery. Mol Divers. 2021;25(3):1315-1360. doi: 10.1007/s11030-021-10217-3.
11. Deore AB, Dhumane JR, Wagh R, Sonawane R. The stages of the drug discovery and development process. Asian J Pharm Res Dev. 2019;7(6):62-67. doi: 10.22270/ajprd. v7i6.616.
12. Bhatt, T.K.; Nimesh, S. The Design and Development of Novel Drugs and Vaccines: Principles and Protocols; Academic Press: Cambridge, MA, USA,2021; ISBN 978-0-12-821475-6.
13. Visan AI, Negut I. Integrating artificial intelligence for drug discovery in the context of revolutionizing drug delivery. Life. 2024;14(2):233. doi: 10.3390/life14020233.
14. Paul, D.; Sanap, G.; Shenoy, S.; Kalyane, D.; Kalia, K.; Tekade, R.K. Artificial intelligence in drug discovery and development. Drug Discov. Today, 2021, 26, 80–93. [CrossRef]
15. Mathai, N.; Chen, Y.; Kirchmair, J. Validation strategies for target prediction methods. Brief. Bioinform. 2020, 21, 791–802. [CrossRef]
16. Gupta, R.; Srivastava, D.; Sahu, M.; Tiwari, S.; Ambasta, R.K.; Kumar, P. Artificial intelligence to deep learning: Machine intelligence approach for drug discovery. Mol. Divers. 2021, 25, 1315–1360. [CrossRef]
17. Yaseen, B.T.; Kurnaz, S. Drug–target interaction prediction using artificial intelligence. Appl. Nanosci. 2023, 13, 3335–3345. [CrossRef]
18. Qiu Y, Cheng F. Artificial intelligence for drug discovery and development in Alzheimer’s disease. Curr Opin Struct Biol. 2024; 85:102776. doi: 10.1016/j.sbi.2024.102776.
19. Rifaioglu, A.S.; Atas, H.; Martin, M.J.; Cetin-Atalay, R.; Atalay, V.; Do?gan, T. Recent applications of deep learning and machine intelligence on in silico drug discovery: Methods, tools and databases. Brief. Bioinform. 2019, 20, 1878–1912. [CrossRef] [PubMed]
20. Mak, K.-K. and Pichika, M.R. (2019) Artificial intelligence in drug development: present status and future prospects. Drug Discovery Today 24, 773–780
21. Kandhare P, Kurlekar M, Deshpande T, Pawar A. A review on revolutionizing healthcare technologies with AI and ML applications in pharmaceutical sciences. Drugs Drug Candidates. 2025;4(1):9. doi: 10.3390/ddc4010009.
22. Wan, F. and Zeng, J. (2016) Deep learning with feature embedding for compound– protein interaction prediction. bioRxiv 2016, 086033.

https://www.atomwise.com/2021/03/09/atomwise-named-to-fast-companys-annual-list-of-the-worlds-most-innovative-companies-for 2021/#:~:text=%2D%2D%20March%209%2C%202021%20%2D%2D,Most%20Innovative%20Companies%20for%202021.

23. Fang Y, Pan X, Shen HB. De novo drug design by iterative multiobjective deep reinforcement learning with graph-based molecular quality assessment. Bioinformatics. 2023;39(4):157. doi:10. 1093/bioinformatics/btad157.
24. Schoenmaker L, B equignon OJM, Jespers W, et al. UnCorrupt SMILES: a novel approach to de novo design. J Cheminform. 2023;15(1):22. doi:10.1186/s13321-023-00696-x.
25. de Souza Neto, L.R., et al., In silico Strategies to Support Fragment-to-Lead Optimization in Drug Discovery. Front Chem, 2020. 8: p. 93.
26. Narayanan, H., et al., Design of Biopharmaceutical Formulations Accelerated by Machine Learning. Mol Pharm, 2021. 18(10): p. 3843-3853.
27. Hughes, J.H., J. Long-Boyle, and R.J. Keizer, Maximum a posteriori Bayesian methods out-perform non-compartmental analysis for busulfan precision dosing. J Pharmacokinet Pharmacodyn, 2024.
28. Niazi, S. K., Mariam, Z., & Magoola, M. (2024). Optimizing Lead Compounds: The Role of Artificial Intelligence in Drug Discovery.
29. https://doi.org/10.20944/preprints202404.0055.
30. Basile AO, Yahi A, Tatonetti NP. Artificial Intelligence for Drug Toxicity and Safety. Trends Pharmacol Sci. 2019 Sep;40(9):624-635.
31. . Raies AB, Bajic VB, In silico toxicology: computational methods for the prediction of chemical toxicity, Wiley Interdisciplinary Reviews: Computational Molecular Science 6 (2) (2016) 147– 172. [PubMed: 27066112]
32. Onakpoya IJ, Heneghan CJ, Aronson JK, Worldwide withdrawal of medicinal products because of adverse drug reactions: a systematic review and analysis, Critical reviews in toxicology 46 (6) (2016) 477–489. [PubMed: 26941185]
33. Segall MD, Barber C, addressing toxicity risk when designing and selecting compounds in early drug discovery, Drug discovery today 19 (5) (2014) 688–693. [PubMed: 24451294].
34. Dr Sunder,s, Artificial intelligence is playing an important role in optimizing clinical trials, including site identification and selection May 27, (2024) 1-3.
35. Quazi S. Artificial intelligence and machine learning in precision and genomic medicine. Med Oncol. 2022;39(8):1-10. doi: 10.1007/s12032-022-01711-1.
36. Bello B, Bundey YN, Bhave R, Khotimchenko M, Baran SW, Chakravarty K, et al. Integrating AI/ML models for patient stratification leveraging omics dataset and clinical biomarkers from COVID-19 patients: A promising approach to personalized medicine. Int J Mol Sci. 2023;24(7):6250. doi: 10.3390/ijms24076250.
37. Rykov MY. [The evolution of personalized medicine: publications review]. Probl Sotsialnoi Gig Zdravookhranenniiai Istor Med. 2022;30(6):1211–1219.
38. . Visvikis-Siest S, et al. Milestones in personalized medicine: from ancient times to nowadays—the provocation of COVID-19. Front Genet. 2020; 11:569175.
39. Taherdoost H, Ghofrani A. AI’s role in revolutionizing personalized medicine by reshaping pharmacogenomics and drug therapy. In: Intelligent Pharmacy. KeAi Publishing Communications Ltd; 2024. doi: 10.1016/j.ipha.2024.08.005.
40. Bryant AK, Zamora-Resendiz R, Dai X, Morrow D, Lin Y, Jungles KM, et al. Artificial intelligence to unlock real-world evidence in clinical oncology: A primer on recent advances. Cancer Med. 2024;13(12):4250-4267. doi: 10.1002/cam4.7253.
41. Subramanian I, Verma S, Kumar S, Jere A, Anamika K. Multi-omics data integration, interpretation, and its application. Bioinformatics Biol Insights. 2020; 14:1177932219899051. doi: 10.1177/1177932219899051.
42. Tran TTT, Tayara H, Chong KT. Artificial intelligence in drug metabolism and excretion prediction: recent advances, challenges, and future perspectives. Pharmaceutics. 2023;15(4):1260. doi: 10.3390/pharmaceutics15041260.
43. Supriyono W, Wibawa AP, Suyono, Kurniawan F. Advancements in natural language processing: Implications, challenges, and future directions. Telemat Informatics Rep. 2024; 16:100173. doi: 10.1016/j.teler.2024.100173.
44. Martin C, DeStefano K, Haran H, Zink S, Dai J, Ahmed D, Razzak A, Lin K, Kogler A, Waller J, Kazmi K, Umair M. The ethical considerations include inclusion and biases, data protection, and proper implementation of AI in radiology and potential implications. In: Intelligence-Based Medicine. Vol. 6. Elsevier B.V.; 2022. doi: 10.1016/j.ibmed.2022.100073.
45. Hanna MG, Pantanowitz L, Jackson B, Palmer O, Visweswaran S, Pantanowitz J, Deebajah M, Rashidi HH. Ethical and bias considerations in artificial intelligence/machine learning. In: Modern Pathology. Vol. 38, Issue 3. Elsevier B.V.; 2025. doi: 10.1016/j.modpat.2024.100686.