From Algorithms to Approvals: A 2026 Perspective on AI-Driven Drug Discovery and Clinical Success

Advancements in drug discovery have significantly transformed medicine, converting previously lethal diseases into manageable treatment protocols. The process of developing and testing new pharmaceuticals has become more streamlined, enhancing medical practice. However, the creation of a new drug can take over a decade and average around $2.6 billion. Additionally, fewer than 10% of drugs that enter Phase I clinical trials manage to reach the market [1.2]. This lengthy and expensive process typically encompasses various in vitro, in silico, and in vivo experiments, which generally take about four years to complete, along with preclinical studies that cover pharmacokinetic, pharmacodynamic, and toxicological research. Drug screening entails a series of experiments and assessments aimed at evaluating potential drug candidates.

Artificial intelligence (AI), also referred to as machine intelligence, mimics human cognition, enabling computers to perform tasks linked to human thought processes during learning and problem-solving.[3] AI utilizes tools and software that allow for autonomous decision-making based on the analysis of incoming data. [4]  While artificial intelligence and machine learning are related, they are not synonymous; AI, a subset of computer science that merges engineering with statistics, employs algorithms to execute tasks and facilitate decisions and predictions. AI-based methods have also been utilized to forecast the potential toxicity of drug candidates [5].

Drug discovery has traditionally navigated a prolonged and intricate pipeline that includes target identification, hit discovery, lead optimization, preclinical studies, clinical trials, and regulatory approval. Each of these phases demands exhaustive experimentation, substantial datasets, and multidisciplinary expertise. Growing challenges arise from the increasing complexity of diseases such as cancer, neurodegenerative disorders, infectious diseases, and rare genetic conditions, all of which underscore the limitations of conventional drug development methods. Traditional approaches often depend on trial-and-error processes, making them inefficient and costly. Consequently, researchers are turning to innovative technologies to enhance the speed, precision, and success rates of pharmaceutical development.

Recent technological advancements in computational biology, bioinformatics, big data analytics, and sophisticated computing have revolutionized pharmaceutical research. The advent of vast biological databases, genomic sequencing technologies, electronic health records, and high-throughput screening techniques generates immense volumes of data, making manual management and analysis increasingly infeasible. AI presents a powerful solution to efficiently process and analyze these complex datasets, enabling researchers to uncover significant patterns and connections that might otherwise go unnoticed.

Integrated AI techniques—including machine learning (ML), deep learning (DL), natural language processing (NLP), reinforcement learning, and generative models—are steadily becoming a part of drug discovery workflows. For instance, ML algorithms can sift through historical datasets to identify characteristics linked to successful drug candidates, while DL models can distill intricate biological relationships from molecular and genomic data. Additionally, NLP facilitates the analysis of scientific literature, patents, and clinical records, expediting information gathering from extensive text sources.

The incorporation of AI into pharmaceutical research signifies a transformative shift from traditional experimental methodologies to data-driven decision-making. AI is poised to assist researchers across multiple stages of drug development, encompassing target identification, virtual screening, molecular design, toxicity prediction, biomarker discovery, patient stratification, and clinical trial optimization. By utilizing predictive analytics and automation, AI can decrease human error, cut development costs, and boost the likelihood of discovering effective therapeutic candidates.

One of the noteworthy applications of AI within contemporary drug discovery is virtual screening. Traditional laboratory methods that screen millions of chemical compounds are prohibitively expensive and time-consuming. AI-enhanced virtual screening capabilities can quickly evaluate extensive chemical libraries to identify compounds possessing desired biological properties, greatly limiting the number of compounds that require experimental evaluation and thus speeding up lead identification.

Furthermore, AI has played a pivotal role in strategies for drug repurposing, which involves leveraging existing approved drugs for new therapeutic applications. Since approved medications have known safety profiles, repurposing can significantly shorten development timelines and mitigate financial risks. During global health crises, such as outbreaks of infectious diseases, AI-supported drug repurposing tactics have proven to be particularly effective in rapidly identifying candidate molecules for therapeutic deployment.

Artificial Intelligence in the Pharmaceutical Sector-

The digitization of data in the pharmaceutical industry has surged in recent years. However, this has made it increasingly difficult to collect, analyze, and integrate information to address key clinical challenges. AI incorporates various methodological areas, including[6]. knowledge representation, problem-solving, and basic machine learning models (ML). Early implementations of neural networks, such as the Perceptron, offered promising solutions, exemplified by a notable 1992 study by Weinstein et al. that utilized neural networks to elucidate the mechanisms of cancer treatments[7]. Unsupervised machine learning can help identify disease targets and subtypes by applying feature-finding algorithms and clustering techniques. [8] 

Limitations of Current Drug Discovery Techniques-

Despite the examination of numerous potential drug molecules to identify those with necessary properties, existing methods are often costly, time-consuming, and prone to inaccuracies [9]. Additionally, they can be limited by the availability of suitable test compounds and the ability to accurately predict their bodily behavior[10]. These challenges can be addressed by various AI-based algorithms, including reinforcement learning, supervised and unsupervised learning techniques, evolutionary algorithms, and rule-based algorithms. These technologies can more precisely and efficiently predict the effectiveness and toxicity of new therapeutic molecules compared to traditional methods[11,12]. Furthermore, AI algorithms[13] can discover novel drug targets, such as specific proteins or genetic pathways involved in diseases, potentially expanding drug[14] discovery beyond the limitations of conventional approaches and leading to the development of new and more effective medications. Despite their long-standing presence, these AI methods demonstrate significant promise. While traditional pharmaceutical research methods have seen some success, they are limited by their dependence on trial-and-error approaches and their inability to accurately predict the behavior of new, potentially bioactive compounds. In contrast, methods that utilize artificial intelligence (AI) can enhance the efficiency and accuracy of drug discovery processes, leading to the development of more effective medications.

While ligand-based and structure-based methods are well-established techniques for developing unique molecular profiles with effective pharmacological properties, the process [15, 16] of computer-assisted de novo drug creation remains challenging. The integration of artificial intelligence (AI) and machine learning with single-cell biological data can yield groundbreaking outcomes in drug discovery by enhancing the prediction of biomarkers and identifying valuable disease-related targets for new drug candidates [17]. Furthermore, AI's application in drug development can optimize a medication's metabolism and excretion, thereby improving its safety and effectiveness for both humans and animals. Proper regulation of metabolism and excretion is vital for removing harmful substances from the body and preventing their accumulation, which can lead to metabolic disorders and damage to the liver and kidneys. Additionally, drug metabolism plays a significant role in issues such as multidrug resistance in cancer treatments and viral infections. Recent research has shown AI's effectiveness in predicting drug metabolism and excretion [18].

The application of machine learning (ML) in forecasting the toxicity and efficacy of pharmaceuticals is becoming increasingly important. Traditionally, assessing a compound's potential effects on the human body involves labor-intensive and time-consuming experiments, leading to lengthy and costly processes with often unclear and variable results. This is where AI techniques, such as ML, can provide significant improvements. ML algorithms can identify patterns and trends within large datasets that human researchers might overlook. For example, a recent dataset comprising known medicinal compounds and their biological activities was utilized to train a deep learning (DL) algorithm [19] , which subsequently demonstrated high accuracy in predicting the effects of new chemicals. Additionally, after extensive training with large databases of both hazardous and non-toxic substances, significant advancements have been made in preventing the toxicity of potential therapeutic agents [20].

Another crucial role of AI in drug development is the detection of drug-drug interactions, which can occur when multiple medications are administered to a patient, leading to altered effects or adverse reactions. AI techniques can analyze extensive datasets of known drug interactions to identify trends and patterns. A recent ML system has successfully predicted interactions between new medication combinations. Furthermore [21] , AI is pivotal in personalized medicine by enabling the identification of potential drug-drug interactions, thus facilitating the creation of tailored treatment plans that minimize the risk of adverse side effects.

Several successful studies in AI-based drug discovery have identified new compounds with strong potential for treating cancer. This shows that AI is an effective tool for finding new and promising treatment options. Recent studies have shown that machine learning (ML) can help identify small chemical inhibitors for the MEK22 protein, which is an important target in cancer treatment. Finding effective inhibitors for this protein has been difficult using traditional methods, but ML successfully discovered new potential inhibitors.

Similarly, machine learning has also been used to identify new inhibitors of beta-secretase, an enzyme associated with the development of Alzheimer's disease [22] . These examples demonstrate how AI can improve the prediction of a drug’s effectiveness and safety before further development. As a result, AI can speed up the drug discovery process and support the development of safer and more effective medicines.

AI plays an important role in improving medicinal chemistry by helping scientists analyze large amounts of data quickly and accurately. For AI and machine learning to work effectively, access to large datasets and proper analysis tools is necessary. Rather than replacing medicinal chemists, AI acts as a supportive tool that helps them work more efficiently. As suggested by researchers, medicinal chemists who use AI are likely to perform better than those who do not use it. AI can also help predict the safety, toxicity, and effectiveness of potential drug candidates, making the drug discovery process faster and more efficient.

AI helps improve research, development, and operational processes in several areas of chemistry, including computational chemistry, spectroscopy, material design, reaction optimization, drug discovery, and process control. It also supports clinicians through clinical decision support systems, helping them design better treatment strategies based on patient outcomes. In addition, AI is used in medical imaging to analyze MRI scans, CT scans, X-rays, and other images to detect diseases or abnormalities. According to Carolyn Meltzer, AI is not meant to replace doctors but to assist them in their work. Overall, AI has the potential to transform healthcare by improving our understanding of diseases and supporting better data-based decision-making.

Ethical Considerations of AI in the Pharmaceutical Industry-

In drug synthesis, after identifying potential molecules, the next and most difficult step is to create these compounds in the laboratory. This process is usually done through retrosynthesis [23], which means breaking down a complex molecule into simpler parts to understand how it can be made. The Synthia program, earlier called Chematica, uses computer-based rules to suggest possible methods for synthesizing important pharmaceutical compounds. It can also help scientists design materials that have never been created before and suggest alternative ways to produce patented products. Another method, called Reinforcement Learning for Structural Evolution, uses advanced deep learning models to generate and predict new molecules from the beginning, making drug discovery faster and easier.

Artificial Intelligence and Quantitative structure –

Activity Relationship (QSAR)modeling have developed significantly [29] over the last fifty years. QSAR models help researchers predict how a molecule’s structure affects its biological activity and pharmacokinetic behavior, such as toxicity and elimination from the body. To allow computers to understand molecules, their structural properties—such as shape, chemical characteristics, and pharmacophore features—are transformed into numerical values called molecular descriptors. These descriptors allow computers to analyze and compare compounds efficiently. Many modern AI techniques, including self-organizing maps and deep learning systems, originated during the rapid growth of artificial neural networks in cheminformatics during the 1990s [30,31]. Deep learning gained major recognition in 2012 after achieving excellent performance in the Merck Molecular Activity Challenge [32], demonstrating its ability to predict molecular activity accurately. However, scientists continue to debate whether deep learning always performs better than other machine [33] learning methods, such as gradient boosting models, when both use the same molecular information [34] . One major strength of deep learning is its ability to perform multitask learning [35,36]. Instead of predicting only one property at a time, multitask learning allows a model to learn several related properties simultaneously [37] . Unlike traditional multi-output methods [38,39] , multitask learning identifies relationships between different prediction tasks and uses shared information to improve performance. This is especially useful in drug development because researchers usually need to optimize many properties at the same time, such as effectiveness, safety, and metabolism. Since chemical datasets often contain incomplete information across different endpoints, multitask learning helps use available data more efficiently. The concept of predicting multiple biological outcomes from molecular descriptors existed before deep learning became popular, but modern AI approaches have made this process much more powerful and effective.

1. Prediction of Physicochemical Properties-

The success of a drug in reaching the market largely depends on its ADMET properties (Absorption, Distribution, Metabolism, Excretion, and Toxicity)  These properties are strongly affected by the physicochemical characteristics of a drug molecule [40,41] . Therefore, understanding these characteristics is very important during drug development. One important property is the ionization constant (pKa), which affects how a molecule behaves in different environments. The pKa influences important features such as water solubility and the octanol–water distribution coefficient (logD). These properties determine how well a drug dissolves and moves within the body, which can also affect the formulation and delivery of the medicine. In addition, the charge state of a compound changes at different pH levels and significantly influences its ADMET behavior. Drug molecules with suitable physicochemical properties often become good lead compounds and provide valuable ideas for designing new medicines. However, promising properties alone do not guarantee successful drug development [42]

Directly measuring the physical properties of small molecules in laboratories can be difficult and time-consuming. Therefore, accurate prediction methods are useful because they help researchers improve and optimize molecular structures until desired properties are achieved. Many computational and AI-based methods focus on predicting specific physicochemical properties such as lipophilicity [43] (fat solubility) and water solubility [44], while others can predict several properties simultaneously[45] . For example, one AI approach used Natural Language Processing (NLP) techniques on molecular SMILES representations. The model converted molecular structures into embedding vectors and used a transformer model to predict the water solubility of molecules [46].

2. Forecasting Features Related to ADMET-

Many drug candidates fail during clinical trials not because they lack effectiveness, but because they show poor ADMET characteristics. The ADME part (Absorption, Distribution, Metabolism, and Excretion) plays a major role in determining whether a drug can successfully reach its target protein inside the body and how it is transported or processed.

During the early stages of drug discovery, researchers may need to evaluate hundreds or even thousands of compounds for their ADMET properties. Testing each compound through laboratory experiments and animal studies requires a large amount of time, money, and effort To overcome this challenge, Artificial Intelligence (AI) has become widely used in drug discovery. AI models can rapidly analyze large amounts of chemical and biological data and accurately predict ADMET properties. This helps scientists identify promising drug candidates early, reduce experimental costs, save time, and improve the overall efficiency of drug development [47,48].

AI in Disease Diagnosis-

Artificial intelligence (AI) is changing the way doctors identify, treat, and manage diseases. AI systems can quickly examine huge amounts of medical information, such as patient symptoms, laboratory reports, medical records, and imaging data. By analyzing this large amount of information in a short time, AI can recognize patterns that may be difficult for humans to detect. This helps doctors identify diseases at an earlier stage. Early disease detection is very important because it allows doctors to start treatment sooner and take preventive actions before the disease becomes severe or spreads further. AI-based technologies are especially useful in detecting communicable diseases and various medical conditions with greater speed and accuracy. These systems can identify infectious organisms and diseases with high sensitivity and precision, reducing the risk of wrong diagnoses and avoiding unnecessary medical procedures. As a result, AI improves decision-making, increases treatment efficiency, and enhances overall patient outcomes Early disease detection is very important because it allows doctors to start treatment sooner and take preventive actions before the disease becomes severe or spreads further. AI-based technologies are especially useful in detecting communicable diseases and various medical conditions with greater speed and accuracy. These systems can identify infectious organisms and diseases with high sensitivity and precision, reducing the risk of wrong diagnoses and avoiding unnecessary medical procedures. As a result, AI improves decision-making, increases treatment efficiency, and enhances overall patient outcomes.

Pulmonary Hypertension-

Pulmonary hypertension is a serious cardiovascular disorder in which the blood pressure inside the pulmonary arteries becomes abnormally high. Pulmonary arteries are responsible for carrying blood from the heart to the lungs. When pressure in these arteries increases, blood flow to the lungs becomes difficult, causing strain on the heart and leading to serious health complications. Early diagnosis of pulmonary hypertension is essential because timely treatment can slow disease progression and reduce complications. Researchers developed an AI model using chest X-ray images collected from patients with different types of pulmonary hypertension as well as healthy individuals. The AI system learned to recognize subtle patterns in these images that may not be easily visible to the human eye. The model achieved an Area Under the Receiver Operating Characteristic curve (AUROC) value of 0.945, indicating excellent performance. It also showed an accuracy of 86.14% in distinguishing various forms of pulmonary hypertension. These results suggest that AI can assist doctors in making faster and more accurate diagnoses, improving patient care and treatment planning.

Alzheimer’s Disease-

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder that affects the brain. It gradually damages memory, thinking ability, learning capacity, and daily functioning. Diagnosing Alzheimer’s disease in its early stages remains challenging because symptoms often develop slowly and can resemble normal aging or other conditions. Traditional diagnosis of Alzheimer’s disease often depends on invasive tests and specialized procedures that are usually available only in selected healthcare facilities. Because of these limitations, there has been a need for a simpler, more affordable, and accessible screening method.

Recent developments in medical imaging technology, especially fluorine-18-fluorodeoxyglucose positron emission tomography (FDG-PET) brain imaging, have improved early detection of Alzheimer’s disease. AI models have been developed to analyze these brain images efficiently. The system is described as an “end-to-end” AI model because it performs all stages of diagnosis within a single framework, from image processing to final disease prediction. This approach reduces manual effort, increases diagnostic efficiency, and supports early identification of Alzheimer’s disease. Early diagnosis allows patients to receive proper care and treatment at an earlier stage, potentially improving quality of life and disease management.

Cancer-

Cancer is mainly diagnosed using clinical imaging techniques such as X-rays, CT scans, MRI scans, and microscopic examinations. These methods help doctors observe abnormal changes in body tissues and identify tumors. In recent years, artificial intelligence (AI) has become an important tool in improving cancer diagnosis by increasing speed and accuracy.

One example is an online AI-based tool called AI Dermatologist, which uses deep learning technology to detect skin cancer from photographs uploaded by users. The system carefully examines the images and studies several important features of skin lesions, including:

• Asymmetry: whether one half of the skin spot differs from the other half
• Border irregularity: whether the edges are uneven or unclear
• Color variation: presence of different colors within the same lesion
• Size: measurement of the lesion diameter
• Evolution over time: changes in shape, size, or appearance over a period

By analyzing these characteristics, the AI model can distinguish between benign (non-cancerous) and malignant (cancerous) tumors. The AI Dermatologist demonstrated a sensitivity of approximately 87% in detecting malignant cells from body images, showing its ability to accurately identify many cancer cases. Traditionally, skin cancer diagnosis depends on microscopic examination of tissue samples collected through biopsy procedures. However, AI-based systems can support doctors by providing faster preliminary assessments and reducing diagnostic workload. Researchers are continuously improving AI model structures to better analyze different patterns found in radiological and medical images.

Another major advancement supporting AI in cancer research is Next Generation Sequencing (NGS) technology. NGS can generate detailed information about an individual's entire genome, producing extremely large amounts of biological data. Processing and interpreting such data manually can take days or even weeks.

Deep learning and AI methods have created new opportunities for converting this enormous amount of information, often referred to as big data, into useful medical knowledge.

 AI applications in genomic studies include:

• Gene annotation
• Studying relationships between genes and physical characteristics
• Research on inherited disorders
• Mutation analysis
• Cancer detection
• Biomarker discovery
• Prediction of gene functions
• Variant identification

AI contributes to many stages of drug discovery including:

• Patient recruitment and monitoring
• Prediction of physicochemical properties
• Bioactivity prediction
• Toxicity prediction
• ADME property analysis
• Protein structure prediction

Although AI has many advantages, several challenges still limit its full implementation.

One major challenge is the dependence of AI systems on high-quality data availability. AI performance greatly depends on reliable and well-organized datasets. Better data collection, management systems, and user-friendly platforms are necessary to effectively manage big data characteristics such as:

• Volume: very large amounts of data
• Velocity: rapid generation of data
• Variety: different forms of data
• Volatility: continuously changing data

Without sufficient and properly curated data, extracting meaningful insights becomes difficult and can slow progress in AI-driven drug development.

Additional challenges include:

• Frequent software updates required to support new data formats and algorithms
• Shortage of trained professionals capable of operating AI-based systems
• Difficulty in predicting protein conformational changes
• Challenges in estimating binding strength between drugs and biological targets

Deep learning models requiring large amounts of high-quality data for effective performance

To overcome these limitations, researchers are exploring advanced methods such as transfer learning, which enables AI models to apply knowledge learned from one task to another related task. This approach may reduce the need for extremely large datasets and improve AI performance in future drug discovery applications.

RESULT AND DISCUSSION-

One of the major applications of AI in medicinal chemistry is predicting the toxicity and effectiveness of drug compounds. Traditionally, researchers relied on laboratory experiments that required a large amount of time, effort, and money to determine how a compound affects the human body. AI helps reduce this burden by analyzing large datasets and providing predictions quickly.

 AI improves different stages of drug discovery, including:

• Target identification: finding biological targets linked to diseases.
• Lead optimization: improving drug candidates to make them safer and more effective.
• Clinical trial planning: helping researchers design trials more efficiently.
• Virtual screening: rapidly examining thousands of compounds to identify promising candidates.
• Protein-drug interaction prediction: estimating how drugs interact with specific proteins.
• Bioactivity prediction: predicting the biological effects of molecules.

AI can create predictive models that identify compounds likely to bind strongly with target proteins, increasing the chances of discovering effective medicines. The use of AI in pharmaceutical research creates both opportunities and challenges. Programs such as AI-based structure-guided drug discovery initiatives encourage collaboration between industry experts and researchers. These collaborations combine machine learning techniques with molecular structure analysis to improve drug development methods.

However, the performance of AI systems depends heavily on the availability of large, high-quality datasets. Collecting and maintaining these datasets often requires substantial investment. Poor-quality or incomplete data can lead to inaccurate predictions and unreliable outcomes.

Several challenges still limit the full implementation of AI in the pharmaceutical industry:

1. Shortage of skilled professionals capable of managing AI systems

2. High implementation costs, especially for smaller companies

3. Concerns that AI may replace human jobs

4. Doubts regarding the reliability of AI-generated results

5. Lack of transparency due to AI's "black box" nature

Despite these concerns, AI mainly automates repetitive tasks rather than replacing human intelligence. This allows scientists and researchers to spend more time on complex analysis, creative thinking, and decision-making. Therefore, AI is more likely to support human work rather than eliminate it. To fully benefit from AI technologies, pharmaceutical companies require trained data scientists and software developers who understand both AI techniques and the goals of drug research and development.

As AI technology continues to advance, it is expected to solve many current challenges in medicine development. AI may also help reduce healthcare costs and improve treatment methods. One exciting future application is personalized medicine, where AI can help create medicines designed specifically for individual patients, including selecting the correct dosage and drug release patterns according to personal needs Overall, AI has the potential to revolutionize pharmaceutical research and transform the future of healthcare.

The successful use of artificial intelligence (AI) in healthcare and drug development depends on combining important factors and technologies according to each patient’s specific needs. Every patient is different, so treatments and medical decisions should be personalized. One major concern about AI is that people fear it may replace human jobs. Another challenge is the need for strict rules and regulations before AI can be widely used. However, the purpose of AI is not to completely replace humans; instead, it is designed to assist people and make difficult tasks easier and faster. AI can rapidly identify potential drug compounds that may become useful medicines. It can also suggest ways to create these compounds through suitable chemical synthesis pathways. Furthermore, AI can predict what type of chemical structures are needed for effective drugs and help scientists understand how drugs interact with biological targets in the body.

CONCLUSION

The pharmaceutical industry has experienced major progress after the introduction of Artificial Intelligence (AI). AI has become an important technology in drug discovery and development because it helps scientists find better treatments for serious and long-term diseases such as diabetes, Parkinson’s disease, Alzheimer’s disease, and Obsessive-Compulsive Disorder (OCD). Many of these diseases were difficult to treat effectively using traditional methods. AI can analyze huge amounts of medical and biological data quickly, helping researchers discover new solutions faster than before. The importance of AI became even more noticeable during the COVID-19 pandemic, when there was an urgent need to develop medicines and vaccines in a short period of time. AI helped speed up the process of identifying potential drug candidates and understanding disease behavior. This showed how powerful AI can be in responding to global health emergencies. In the coming years, most pharmaceutical companies are expected to work closely with AI-based organizations and technologies. Such collaborations may bring major improvements to the healthcare and pharmaceutical industries. Researchers believe that explainable or interpretable AI systems and active learning methods will become increasingly important in the next decade. These advanced tools can make AI decisions easier to understand, allowing scientists to track, examine, and better understand every stage of the drug development process. Artificial Intelligence, Machine Learning (ML), and Deep Learning are transforming the drug discovery pipeline by improving human decision-making and reducing the time needed for research. Deep learning techniques are widely used in Computer-Aided Drug Discovery (CADD), helping researchers identify important drug molecules more efficiently and accurately. This review highlights recent advancements in AI applications for drug discovery and development. However, the successful use of AI depends on several important factors, including access to high-quality and reliable data, solving ethical issues, and understanding the limitations of AI methods. To overcome these challenges, researchers are focusing on strategies such as explainable AI, data augmentation techniques, and combining AI approaches with traditional laboratory experiments. These methods can make drug discovery more effective, reliable, and beneficial for future healthcare systems.

REFERENCES

1. Mullard, A. (2014). The development cost of new drugs is $2.6 billion. Nature Reviews Drug Discovery, 13(13), 877.
2. Dowden, H., & Munro, J. (2019). Clinical success rates and trends in therapeutic focus. Nature Reviews Drug Discovery, 18(6), 495-497.
3. Niazi, S. K. (2023). The maturation of AI/ML in drug discovery, development, clinical testing, and manufacturing: Insights from the FDA. Drug Design, Development and Therapy, 17, 2691-2725.
4. Paul, D., Sanap, G., Shenoy, S., Kalyane, D., Kalia, K., et al. (2021). The role of artificial intelligence in drug discovery and development. Drug Discovery Today, 26(1), 80-93.
5. Pu, L., et al. (2019). EToxPred: A machine learning-based method to estimate the toxicity of drug candidates. BMC Pharmacology and Toxicology, 20, 1-15.
6. Ramesh, A. N., Kambhampati, C., Monson, J. R., et al. (2004). The application of artificial intelligence in medicine. Annals of the Royal College of Surgeons of England, 86, 334–338.
7. Weinstein, J. N., Kohn, K. W., Grever, M. R., et al. (1992). Neural computing methods in cancer drug development: Mechanism of action predictions. Science, 258, 447–451.
8. Chen, H., Engkvist, O., Wang, Y., et al. (2018). The emergence of deep learning in drug discovery. Drug Discovery Today, 23, 1241–1250.
9. Pu, L., et al. (2019). EToxPred: A machine learning-based approach for estimating drug candidate toxicity. BMC Pharmacology and Toxicology, 20, 1-15.
10. Gómez-Bombarelli, R., et al. (2018). Automated chemical design utilizing a data-driven continuous representation of molecules. ACS Central Science, 4, 268–276.
11. Shiammala, P. N., Duraimutharasan, N., Vaseeharan, B., Alothaim, A. S., Al-Malki, E. S., Snekaa, B., Safi, S. Z., Singh, S. K., Velmurugan, D., & Selvaraj, C. (2023). Investigating AI and machine learning in drug design: Challenges and future potential for the pharmaceutical industry. Methods. https://doi.org/10.1016/j.ymeth.2023.09.010
12. Gawehn, E., Hiss, J. A., & Schneider, G. (2016). Deep learning in drug discovery. Molecular Informatics, 35, 3–14.
13. Lysenko, A., Sharma, A., Boroevich, K. A., & Tsunoda, T. (2018). An integrative machine learning approach for predicting drug safety related to toxicity. Life Science Alliance, 1, e201800213.
14. You, J., McLeod, R. D., & Hu, P. (2019). Predicting drug-target interaction networks through deep learning models. Computational Biology and Chemistry, 80, 90–101.
15. Liu, X., IJzerman, A. P., & van Westen, G. J. P. (2021). Computational strategies for de novo drug design: Historical context, current methods, and future outlook. In Methods in Molecular Biology (Vol. 2190, pp. 139–165). Humana Press Inc.
16. Bhisetti, G., & Fang, C. (2022). AI-Driven De Novo Design: Creating Novel Synthesizable Compounds. Methods in Molecular Biology, 409-419.
17. Meyers, J., Fabian, N. B., & Brown, N. (2021). De novo molecular design and generative models. Drug Discovery Today, 26(11), 2707–2715.
18. Thi Tuyet, T. V., Tayara, H., & Chong, K. T. (2023). The role of artificial intelligence in predicting drug metabolism and excretion: Recent advances, challenges, and future directions. Pharmaceutics, 15(4), 1260.
19. Hansen, K., et al. (2015). Machine learning predictions of molecular properties: Accurate many-body potentials and nonlocality in chemical space. Journal of Physical Chemistry Letters, 6, 2326–2331.
20. Pérez Santín, E., et al. (2021). An overview of artificial intelligence applications in toxicity prediction. Wiley Interdisciplinary Reviews: Computational Molecular Science, 11, e1516.
21. Dhamodharan, G. & Mohan, C. G. developed machine learning models to forecast the efficacy of dual inhibitors for AChE and BACE1 in combating Alzheimer’s disease. Mol. Divers. 26, 1501–1517 (2022). 
22. Ezrachi, A. & Stucke, M. E. explored the implications of artificial intelligence on collusion and its potential to stifle competition. U Ill L Rev 2017; 1775. 
23. Popova, M., Isayev, O., & Tropsha, A. presented a deep reinforcement learning approach for de novo drug design. Sci Adv 2018; 4: eaap7885. 
24. Engkvist, O., Norrby, P-O., Selmi, N., et al. discussed the current state and future prospects of computational predictions in chemical reactions. Drug Discov. Today 23(6):1203–1218 (2018). 
25. Schneider, G. reviewed automated methods in drug discovery. Nat. Rev. Drug Discov. 17, 97–113 (2018). 
26. Schneider, N., Lowe, D. M., Sayle, R. A., Tarselli, M. A., & Landrum, G. A. analyzed big data from pharmaceutical patents to aid medicinal chemists. J. Med. Chem. 59, 4385–4402 (2016). 
27. Delaney, J. modeled iterative compound optimization utilizing a self-avoiding walk method. Drug Discov. Today 14(3-4), 198–207 (2009). 
28. Agrawal, P. discussed the role of artificial intelligence in drug discovery and development. Artificial Intelligence in Drug Discovery and Development. 2018; 6(2):1-2. 
29. Fu, J., Zhang, Y., Liu, J., Lian, X., Tang, J., & Zhu, F. examined pharmacometabonomics, focusing on data processing and statistical analysis. Brief Bioinform 2021; 22(5):bbab138. 
30. Basile, A. O., Yahi, A., & Tatonetti, N. P. highlighted the application of artificial intelligence in assessing drug toxicity and safety. Trends Pharmacol Sci 2019; 40(9):624–35. 
31. Meanwell, N. A. emphasized enhancing drug candidates through design by improving physicochemical properties to optimize compound safety and disposition. Chem Res Toxicol 2011; 24(9):1420–56. 
32. Lipinski, C. A. introduced the 'rule-of-five' paradigm for assessing lead and drug-like compounds. Drug Discov Today Technol 2004; 1(4):337–41. 
33. Zhang, M. Q. & Wilkinson, B. explored drug discovery beyond the 'rule-of-five'. Curr Opin Biotechnol 2007; 18(6):478–88. 
34. Manallack, D. T., Prankerd, R. J., Yuriev, E., Oprea, T. I., & Chalmers, D. K. discussed the importance of acid/base properties in drug discovery. Chem Soc Rev 2013; 42(2):485–96. 
35. Zhang, H., Xiang, M. L., Ma, C. Y., Huang, Q., Li, W., Xie, Y., et al. developed three-class classification models for logS and logP using a GA-CG-SVM approach. Mol Divers 2009; 13(2):261–8. 
36. Jorgensen, W. L. & Duffy, E. M. predicted drug solubility through Monte Carlo simulations. Bioorg Med Chem Lett 2000; 10(11):1155–8. 
37. Yang, K., Swanson, K., Jin, W., Coley, C., Eiden, P., Gao, H., et al. analyzed molecular representations for predicting properties. J Chem Inf Model 2019; 59(8):3370–88. Francoeur, P. G. & Koes, D. R. introduced SolTranNet, a machine learning tool for rapid aqueous solubility predictions. J Chem Inf Model 2021; 61(6):2530–6. 
38. Yin, J., Li, F., Zhou, Y., Mou, M., Lu, Y., Chen, K., et al. developed INTEDE, a database of drug-metabolizing enzyme interactomes. Nucleic Acids Res 2021; 49(D). 
39. Fu, T., Li, F., Zhang, Y., Yin, J., Qiu, W., Li, X., et al. released VARIDT 2.0, detailing the structural variability of drug transporters. Nucleic Acids Res 2022; 50(D1):D1417–31. 
40. Hansch, C., Maloney, P. P., Fujita, T., et al. correlated the biological activity of phenoxyacetic acids with Hammett substituent constants and partition coefficients. Nature. 1962; 194(4824):178–180. 
41. Goller, A., Kuhnke, L., Montanari, F., et al. documented Bayer’s in silico ADMET platform, highlighting two decades of advancements in machine learning. Drug Discov Today. 2020; 25(9):1702–1709. 
42. Beck, B. & Geppert, T. discussed the industrial applications of in silico ADMET studies. J. Mol. Model. 2014; 20(7). DOI:10.1007/s00894-014-2322-5. 
43. Beck, B. & Geppert, T. reiterated the industrial applications of in silico ADMET studies. J. Mol. Model. 2014; 20(7). DOI:10.1007/s00894-014-2322-5. 
44. Todeschini, R. & Consonni, V. compiled molecular descriptors for cheminformatics in two volumes: alphabetical listings and appendices. Weinheim, Germany: John Wiley & Sons; 2009. 
45. Zupan, J. & Gasteiger, J. introduced neural networks tailored for chemists. New York, NY: John Wiley & Sons, Inc.; 1993. 
46. Schneider, G. explored adaptive systems in drug design. Boca Raton, FL: CRC Press; 2002.
47. Unterthiner T, Mayr A, Klambauer G, et al. highlight the potential of deep learning in virtual screening, as presented at the deep learning workshop during NIPS in Montreal, Canada (2014).
48. Sheridan RP, Wang WM, Liaw A, et al. discuss the use of extreme gradient boosting for quantitative structure–activity relationships in the Journal of Chemical Information and Modeling (2016).