Integrative Biomedical Research

1. Introduction

AI is changing many things, but it's changing health care the most. AI can look at a lot of data, identify patterns that we might not see, and guess what will happen next. This technique helps doctors figure out what's wrong with patients, and it teaches them how to better look after people. It also helps discover new drugs and deliver them to the public faster (Mak & Pichika, 2019; Vyas, 2018).

AI stands for "artificial intelligence," which means that computers can do things that people can do. They can learn, think, and come up with answers (Rouse, 2017). AI in healthcare includes things like deep learning, natural language processing, and machine learning. People use these tools at every stage of making drugs now. They aid in target identification, molecule synthesis, drug distribution to patients, and safety monitoring post-distribution (Zhu, 2020; Chan, 2019).  It takes a long time, costs a lot of money, and doesn't work very well to discover new drugs the old way. It costs more than $2 billion and takes more than ten years to make a new drug ready for people to use (Mak & Pichika, 2019). Only about 10% of drugs tested on humans receive approval from the government This practice is mostly because biological systems

are very complicated, and it's difficult to predict what will happen after treatment. In clinical pharmacology, it is still challenging to deliver drugs to the right place, at the right time, and in the right amount (Baronzio, 2015). AI promises that it can help with many of these problems by giving us tools for personalized medicine, high-throughput screening, predictive modelling, and real-time data analysis (Brown, 2015). Recent summaries from 2024 to 2025 say that AI is involved in every step of the process, from finding something to making it to delivering it. These summaries show that translation is getting more accurate, faster, and cheaper (Marques et al., 2024; Bai et al., 2024; Gallego et al., 2021).

Finding and confirming targets is one of the best ways to use AI in drug discovery. Deep learning algorithms and natural language processing can sift through a lot of biomedical literature and omics data to uncover new molecular targets that are connected to certain diseases (Zhu, 2020; Ciallella & Zhu, 2019). By concentrating solely on viable drug targets, this focused approach conserves time. It also helps scientists discover new drugs faster. AI has also simplified the process of identifying compounds suitable for use as leads. Previously, this process relied on laborious and time-consuming experiments. Two types of AI that can predict the bioactivity, toxicity, solubility, and other pharmacokinetic properties of thousands of compounds in silico are deep neural networks and support vector machines (Zhang et al., 2017; Wang et al., 2015).

Beyond developing new compounds, AI also helps us come up with new ways to use drugs. This is a cheap way to learn how to use drugs that are already on the market in different ways. AI systems can uncover new links between drugs and diseases that older methods don't always identify. They do these tasks by combining different kinds of information, such as genomic, clinical, chemical, and pharmacological data (Pereira et al., 2016; Álvarez-Machancoses & Fernández-Martínez, 2019). Since we already know these drugs are safe, this method accelerates the process of making and getting them approved.

In the realm of drug delivery, AI is also crucial for finding new medicines and making sure people receive the right ones. AI can help us learn more about how liposomes and nanoparticles move and do their jobs in the body. This knowledge might help us discover better ways to provide people their medicine. These systems lower the risk of side effects in the whole body (Baronzio, 2015). Furthermore, people can also keep track of their health in real time with smart drug delivery pumps and wearable health monitoring devices. These gadgets help us take the right amount of medicine, which makes the treatment work much better (Blasiak, 2020).

Shifting to clinical trials, AI is also changing how clinical trials are set up and run. Many people believe that these tests slow down the process of making new drugs. People would rather not join or stay in traditional trials because they cost a lot of money and take a long time. AI tools can analyze genomics, electronic health records, and real-world evidence to identify optimal patient groups, predict trial outcomes, and monitor safety and adherence (Mak & Pichika, 2019; Firth et al., 2015). You can use machine learning to run tests that change over time. You can change the dose, the rules for who can join the trial, or the arms. Such flexibility makes both ethical integrity and statistical power better (Jain & Jain, 2015).

Post-market surveillance is equally vital. AI can record the sell data and is now used by pharmacovigilance platforms to look at EHRs, social media data, and patient reports. This not only speeds up safety notifications, but it also lets people act quickly (Zhu, 2020; Mak & Pichika, 2019).

However, there are still a lot of moral and legal questions about utilizing AI in drug research, including rule setups, biased algorithms, and unexplainable deep learning models. Most people worry about how well AI models can behave in varied scenarios when data quality and variance cannot be guaranteed (Ciallella & Zhu, 2019). Data scientists, doctors, pharmacologists, and regulators must collaborate to establish guidelines for clinical translation, validation, and reproducibility.

This article explores the exciting ways AI is shaking up drug discovery and delivery and delves into the various areas where AI plays a significant role, including pinpointing molecular targets, designing new compounds, rejuvenating old drugs, improving delivery, streamlining clinical trials, and monitoring post-market developments. Plus, it covers the cool new tech being used, real-world examples, and the important legal and ethical questions we need to figure out so AI can truly shine in healthcare.

For example, Generative adversarial networks are a new way to build molecules. Scientists can use reinforcement learning to make drugs that do more than one thing. AI-powered formulation design is a new technology that helps make drugs that can do more than one thing. For example, MOARF and other sites have used AI workflows that work together to make drugs work better, be safer, and be easier to obtain all at once (Firth et al., 2015). AI-powered platforms like CURATE.AI have demonstrated their capability to enhance personalized dosing regimens by soliciting feedback from individual patients (Blasiak, 2020). People are beginning to believe that silico medicinal chemistry methods based on deep learning and cheminformatics can replace wet-lab experiments (Brown, 2015; King, 1995).

Looking ahead, AI could help identify new drugs faster and ensure that everyone receives the best care they can. Cloud-based AI platforms, open-source data, and decentralized collaboration have made it possible for researchers in places with few It will enhance fairness, transparency, and utility This will significantly alter the process of developing new drugs.  It will make things fairer and more open.

2. AI in Healthcare as a transformative Force

AI is leading the way in the digital healthcare revolution and is changing how healthcare systems work in every country. It helps doctors figure out what's wrong with patients and how to fix it, speeds up the process of making new drugs, and makes patients' lives better. AI provides us with strong tools that help people and make the healthcare system work better. Such work is important because doctors are facing challenges with increasingly demanding patients, rising costs of medicine, and the growing need for personalized treatments.

This paper gives a full picture of how AI is being used in medicine. We'll start with what AI is and its core components, then explore its many applications in healthcare services, and finally discuss its increasing role in drug development.

2.1 Definition of AI

AI is when computers and other machines try to act like people to make reactions. This means they can learn from the past, think, and figure out the optimal way to finish tasks. AI for now is helping doctors take care of patients, keep track of records, and make the right medicines for each person.

Machine Learning is a part of AI that lets computers learn on their own from data. In reality, doctors use it a lot to look at medical images, figure out how likely someone is to become sick, and learn how diseases spread. Machine learning models can use both new and old clinical data to make predictions about how a patient will do. For example, they can tell if someone is going to have heart problems or sepsis (Liu et al., 2020; Bansal et al., 2022).

Deep learning is a more advanced type of machine learning that uses neural networks with many layers to find patterns in big sets of data. It has worked very well in radiology and pathology. It can read medical images and histological slides almost as well as a person can. Research has shown that deep learning algorithms can accurately identify skin cancers and retinal diseases, comparable to human experts (Haeberle et al., 2019; Esteva et al., 2017).

Natural Language Processing is another vital component of AI. It helps computers understand what people are saying and how to answer. NLP is used in healthcare to extract data from disorganized clinical notes, automate paperwork, and facilitate communication between digital assistants. These tools help nurses and doctors obtain important information about patients faster and do less paperwork (Gibbs et al., 2020; Davenport & Kalakota, 2019).

2.2 Applications of AI in Healthcare

AI is improving care, making things run more smoothly, and giving patients and providers more control over almost every part of modern healthcare.

In particular, AI is especially good at finding diseases in medical tests and pictures, especially when you must look at them. Radiologists can use AI algorithms to identify broken bones, tumors, infections, and brain disorders in scans. For example, DL systems have been used to identify diabetic retinopathy, breast cancer, and lung nodules with accuracy comparable to that of trained radiologists (Esteva et al., 2017). AI helps doctors figure out what's wrong more quickly and makes sure they don't make mistakes.

Beyond diagnostics, AI models can also look at an individual's medical records and other information and predict how likely it is that their health situation is going to get more severe or that they will have to go to the hospital. People with diabetes, heart disease, or other long-term illnesses can use this ability to help these individuals act rapidly and minimize their likelihood of having issues (Davenport & Kalakota, 2019; Liu et al., 2020).

AI has made surgery safer in the operating room with the help of robotic-assisted systems. These robots with AI intelligence make surgery safer, empower doctors, and minimize the chances of making a mistake. Surgeons often use them on the spine, joints, and urinary organs to speed up healing and lower the risk after surgery (Esteva et al., 2017).

AI chatbots and virtual health assistants are helping people talk to their doctors in new ways. They monitor symptoms, help with mental health, and ensure they take their medications. These tools help people receive medical care, especially in areas with few doctors and nurses. They also do things like answering questions (Gibbs et al., 2020).

At the administrative level, AI helps many hospitals work better. Natural language processing and machine learning can help you remember things, make plans, and pay your bills. This means we can save money, achieve better results, and help patients feel more confident in the care they receive (Schuhmacher et al., 2019).

Finally, AI is an enormous help for public health, too. It can track disease outbreaks, figure out how infections spread, and make vaccination plans smarter. For example, during the COVID-19 pandemic, AI was super useful for things like contact tracing, predicting when cases would increase, and analyzing how the virus was moving around the world (Gerke et al., 2020).

2.3 Relevance of AI in Pharmaceutical Sciences

AI is transforming the pharmaceutical world by making the drug development process faster and better, from the initial stages to clinical trials, manufacturing, and getting regulatory approval.

It has always been a long and costly process to find and make new medicines, but AI is changing the way drugs are made by using smart predictions to quickly figure out which drugs are most likely to work as treatments. AI platforms are getting better. Researchers have discovered innovative methods for germ eradication. Methods like examining molecular structures and biological functions help us improve preclinical studies. It can accelerate the whole process of developing drugs, from preclinical studies to clinical trials (Pankevich et al., 2014; Talat et al., 2023; Hasanzadeh et al., 2022).

Moreover, AI is like a super-smart assistant for scientists because it helps them find the right patients for drug studies much faster. It looks at people's health records and their DNA to ensure that the trials have the best chance of succeeding. Additionally, AI helps plan the studies better and keeps a careful watch on all the incoming data, making the whole clinical trial process much smoother. Furthermore, finding problems or bad things early on makes sure that people obey the rules and that the information is correct (Yun et al., 2015).

At the same time, AI is becoming important in personalized medicine, and it can examine extensive genomic and phenotypic data to identify optimal treatment. This method is very useful for treating cancer because the genetic changes that happen can be different among individuals. AI has made it easier to give people neurological disorders like Alzheimer's disease and targeted therapies like mRNA for brain-derived neurotrophic factors. The result suggests new treatment options that look appealing to us (Li et al., 2023).

In manufacturing, the drug business also uses AI in smart manufacturing. AI can help you set up automated factories that keep the quality of your products the same, cut down on waste, and make it easier to keep things getting better. You can do all of these things at the same time with sensors, robots, and real-time data analysis. A lot more people are using AI with the Quality by Design method. This method helps drug companies find important quality traits and improve their products right away (Arden et al., 2021; Grangeia et al., 2020).

Artificial intelligence systems that use data from social media, electronic health records, and published research have made it much easier to keep track of drug safety and pharmacovigilance. These systems have made it much easier to keep track of drug safety and pharmacovigilance. These systems can find early signs of bad drug reactions and send important information to drug companies and government agencies to help make sure drugs are safe to use (Schuhmacher et al., 2021; Zhao et al., 2023).

AI helps write and send documents that must be followed by law. This makes it easier to follow the rules and laws. It makes it easier to work with government agencies and makes sure that international laws are followed. AI speeds up the whole drug life cycle by making sure that all rules are followed from the start of development (Patil et al., 2023). 

We don't just think about AI in the future. It's already being used in hospitals and drugstores, which is changing how we do healthcare research. AI speeds up healthcare, makes it better, and customizes it for each patient. It's also less expensive and easier to use. As technology gets better, it's very important that we use it in a safe and moral way to keep people safe. When people and the newest technology work together, the future of medicine looks good. We need to think about how these tools will change the law, society, and morality as they get better. This is the most important thing to do to get the most out of them. The future of medicine could be very bright with the combination of human knowledge and machine intelligence.

3. AI in Drug Discovery: A Different Way to Look at Making Drugs

In the past, it was hard and expensive to bring a new drug to the market. It usually takes more than ten years and costs more than $2 billion to find new drugs the old-fashioned way. It begins with identifying the molecule and concludes with obtaining regulatory approval (Marques et al., 2024). The industry is also having a challenging time being efficient; the traditional pipeline is full of wasted resources and people leaving because clinical success rates are below 12% (Bhinder et al., 2021). AI has changed how candidates are found and checked by using huge biomedical records and the latest hardware.

3.1 Problems with Traditional Drug Discovery

3.1.1 Time and Cost Limits

When developing new drugs, the old-fashioned way is hard work because it takes a lot of wet-lab experiments to make compounds better. It usually takes approximately 10 to 15 years for a drug to go from the lab to the store (Marques et al., 2024). During this period, the process of screening, developing chemicals, and conducting multi-phase human trials wastes numerous resources. The development of a drug that works takes between 1.5 billion dollars and 2.8 billion dollars (Bhinder et al., 2021).

3.1.2 High Attrition and Low Success Rates

Problems often arise between the identification of a drug in a laboratory and its FDA approval. About 90% of people who agree to take part in clinical trials don't finish them. They usually fail because they have side effects that they weren't expecting, the treatment doesn't work, or the pharmacokinetic profiles aren't good (Chang et al., 2022). It costs plenty to be unsuccessful or not to do something at all.

3.1.3 Fragmented Data

Traditional methods rely on datasets that are broken up, like studies on animals that are done in isolation or assessments that focus on the lab. The inability to combine genomic, proteomic, clinical, and real-world evidence hinders the discovery of complex patterns important for identifying potential drug candidates (Mukherjee et al., 2024).

3.2 How AI Transforms Drug Discovery

AI is a new way to identify drugs that uses data, makes guesses, and changes when it needs to. It is changing the entire drug development process, like keeping an eye on drugs after they are sold out.

3.2.1 Smart Data Mining and Integration

AI can look at lots of data points from several different places, like genetic databases and biomedical journals. A paper in 2024 said that ML models could figure out the possibility of a factor causing cancer. It reviews molecular descriptors and chemical features (Le et al., 2024). Another study found that AI can identify pharmacogenomic links and check targets more easily (Iorio et al., 2016).

3.2.2 Recognizing Patterns and Designing New Molecules

Studies have demonstrated that deep learning models, such as convolutional neural networks and recurrent neural networks, excel in identifying patterns within chemical structures. These models can tell you how well something will dissolve, how bad it is for you, and how well it works in the body. Generative models, such as generative adversarial networks, are presently employed to expedite the optimization of molecules by synthesizing new compounds with optimal properties (Kuenzi et al., 2020). Recent surveys highlight swift advancements in foundational models and diffusion-based generators for de novo design and ADMET optimization (Marques et al., 2024; Lloyd, 2024).

3.2.3 Predictive Modeling of Toxicity and Side Effects

Cardiotoxicity and mutagenicity are two common reasons why clinical trials end. AI can make these predictions on a computer. For example, Chang et al. (2022) used AI algorithms to predict cardiotoxic responses in breast cancer patients receiving anthracycline therapy (Chang et al., 2022), and Le et al. (2024) looked into the cancer-causing potential of different compounds (Le et al., 2024). Such similar models help better identification, which saves time and money.

3.2.4 Accelerated Virtual Screening and Molecular Docking

AI lets you quickly test millions of compounds against a small number of targets. AI-based methods are much faster and more accurate than manual docking methods. Marques et al. (2024) point out that AI-driven in silico approaches could help in choosing candidates for studies, reducing the chance of wet-lab testing (Marques et al., 2024). Structure-based learning uses accurate protein modeling to make virtual screening enrichment and pose forecasting more accurate (Jumper, J., et al., 2021; Abramson et al., 2024; Marques et al., 2024).

3.2.5 AI models improve clinical trials

AI models that use electronic health records and biochemical knowledge can find people who are interested in joining a trial. It will also group them by risk factors and guess how the trial will go. Liu et al. (2019) established a model to identify the responses of patients with metastatic melanoma to PD1 blockade and demonstrated that AI could enhance the efficacy of immunotherapy (Liu et al., 2019). Johannet et al. (2021) used machine learning to predict how well immunotherapy would work and set up trials that were more flexible and less expensive (Johannet et al., 2021). Recent examinations of regulatory submissions reveal a growing adoption of AI/ML in practical applications (Liu et al., 2023; Niazi, 2023).

3.2.6 Drug Repurposing and Multi-Target Prediction

AI also helps with drug repurposing by finding new ways to use drugs that are already available. Dercle et al. (2020) utilized radiomic data and machine learning to identify optimal treatments for patients with non-small cell lung cancer (Dercle et al., 2020). This change saves time and makes things easier for regulators since these chemicals’ safety is controlled. Recent studies indicate that AI-powered pipelines are growing beyond oncology. They employ multi-omics integration and knowledge graphs, which shows the potential for future validations (Roche et al., 2023; Marques et al., 2024).

3.2.7 Enabling Personalized and Precision Medicine

In the real world, CURATE.AI is illustrating how AI infra can customize dosing plans by assessing how patients react instantaneously, which reduces side effects and elevates effectiveness (Blasiak, 2020). These models employ genomics, metabolomics, and proteomics information collected from various sources to develop treatment plans. AI has also shown promise in mimicking how psychiatric drugs work. Sheu et al. (2023) used AI to predict how different people would respond to antidepressants based on their electronic health records (Sheu et al., 2023), and Arnold et al. (2024) utilized machine learning to determine the optimal treatment for individual depression cases (Arnold et al., 2024).

3.2.8 Combining data from wearables and the real world

Wearable sensors can gather health information and use AI to guess when diseases will start or how well drugs will work. For instance, Lam et al. (2021) utilized machine learning on data obtained from activity trackers to ascertain the probability of an individual developing type 2 diabetes (Lam et al., 2021). Stankoski et al. (2021) utilized information collected by smartwatches to enhance dietary behavior analysis more accurately (Stankoski et al., 2021). These instant feedback loops help doctors change treatments much easier.

There are big changes happening in the way people look for new drugs. AI is changing how drug companies make drugs because it can handle a lot of data, find complex patterns, and imitate how living things work. AI gives us hope for a better future in drug development by fixing the problems with the old methods, which were slow, expensive, and deadly. AI will be a big part of the next generation of drug development. It will be used for things like virtual screening, predictive modeling, trial optimization, and personalized dosing.

As AI grows better, it will work better with other new fields like genomics, digital health, and quantum computing. The main goal is to not only find drugs faster but also to improve the healthcare system so that it works better and is more responsive to each patient's needs.

4. Applications of AI in Drug Discovery and Development

AI is changing the pharmaceutical industry in a big way, especially when it comes to finding and developing new drugs. It's hard to understand traditional methods, and they take a lot of time and money. AI makes it possible to do things faster, cheaper, and more accurately. It can help with anything from finding targets to improving clinical trials. Some usages are shown in Table 1 (Connor et al., 2022; Stebbing et al., 2020; Huang et al., 2021; Mesko & Topol, 2023).

Table 1: Key Applications of AI in Drug Discovery

4.1 Workflow in AI-Driven Drug Discovery and Drug Development

AI workflows can make procedures go faster and trials more likely to be successful (An & Cockrell, 2022; Roche et al., 2023; FDA, 2024) (Figure 1).

Figure 1: The AI-driven discovery and development process goes from data to targets, screening, generative design, ADMET, lead optimization, and then into preclinical and clinical stages with feedback loops.

4.2 Chemistry's Foundation Models, LLMs, and Diffusion

Recent research has progressed from task-specific models to foundational models for molecules, proteins, and texts. Large language models (LLMs) and multimodal transformers facilitate knowledge extraction, hypothesis formulation, and protocol synthesis, whereas diffusion- and transformer-based generators enhance de novo design and multitask ADMET optimization (Marques et al., 2024; Mesko & Topol, 2023; Liu et al., 2021; Lloyd, 2024). These models are using more structural priors and experimental constraints, which makes it easier to synthesize and improves the quality of hits.

For both discovery and delivery of workflows, efficiency and retrieval are very important during deployment. Structured pruning and knowledge distillation for model compression enable inference on a device or at the edge in regulated environments while ensuring accurate outcomes (Li et al., 2024a; Li et al., 2025a; Li et al., 2025b; Li et al., 2025c; Li et al., 2025d). Federated and hybrid distillation schemes improve privacy-preserving learning between institutions (Li et al., 2025e). When used with vision-language pretraining, complementary zero-shot hashing methods make it easy to find short pieces of chemistry, bioactivity, and protocols (Dong et al., 2024; Li et al., 2025f). Recent forecasting architectures that combine state-space models and diffusion transformers work better for long-term data. This can be helpful for testing and supply chains (Zeng et al., 2025) (Figure 2).

Figure 2: Ecosystem around foundation models: integrating molecule, protein/structure, and text data to enable generation, property prediction, and synthesis planning with feedback.

4.3 Investment Trends in AI for Drug Discovery

The development in the field of AI has drawn a lot of funding into the pharmaceutical industry, particularly for drug discovery. The rise of funding in AI started in 2018, which shows its potential impact. By 2025, the world plans to spend $12.5 billion in AI areas, which shows how important AI is for development and making drug discovery more accurate (Roche et al., 2023; FDA, 2024).

4.4 AI’s comparative Advantages in Drug Research

AI improves traditional methods far more effectively by rendering them faster, cheaper, and more likely to be effective. Below is a comparison of traditional and AI-based approaches (Connor et al., 2022; Ball & Dal Pan, 2022) (Table 2).

Table 2: Comparative Advantages of AI in Drug Development

New AI-driven techniques in drug development indicate a new era of rapidity and precision. AI uses big data, machine learning, and computer simulations to help it make better decisions at every step, from finding targets to running clinical trials. As AI gets smarter, it will have a bigger impact on how we make drugs in the future.

4.5 AI Contributions

AI is making a big difference in drug development in areas like the following:

• Target Identification (25%): AAI models like Graph Convolutional Networks and Deep Neural Networks are improving drug exploration. Important molecular targets can be found via going through real biological data (An & Cockrell, 2022). Google DeepMind's AlphaFold and other emerging tools can figure out how proteins fold, which is really innovative (Abramson et al., 2024).
• Hit and Lead Optimization (30%): Exscientia's work on DSP-1181 shows that virtual screening and generative models like Convolutional Neural Networks and Generative Adversarial Networks are useful. They help us predict which molecules will be active (Connor et al., 2022; Exscientia, 2023).
• Drug Repurposing (20%): AI uses clinical and molecular data to find new ways to use drugs that are already on the market. For example, Benevolent AI found that baricitinib could be used to treat COVID-19 (Stebbing et al., 2020).
• Clinical Trial Optimization (15%): Companies like Tempus and IQVIA use AI to uncover patients quickly and design trials that can change fast (Tempus, 2023; IQVIA, 2022).
• Precision Medicine (10%): Flatiron Health has performed well in cancer care, and Healx has executed well in treatments for unusual diseases (Flatiron Health, 2023; Healx, 2023).

4.6 Case Studies in AI Applications

AI is now a powerful tool in drug discovery and development, and it has helped make major breakthroughs in many areas of medicine.

Here are some notable success stories that show how AI has changed the way drugs are made (Table 3).

Table 3: Real-world Examples of AI in Drug Discovery

These cases show how AI can help with drug development at different stages. AI can look at a lot of data, identify patterns, and speed things up. This helps you identify targets and get more leads and hits.

AI helps researchers look at real-world data and find treatments that work and are appealing. This is also applicable to highly precise research and medicine. These examples show how AI is changing medicine and making it easier to find new, cheap treatments.

5. AI in Drug Delivery Systems

AI is making drug delivery systems smarter, more personalized for each patient, and better at what they do. Traditional drug delivery methods face challenges like one-size-fits-all dosing, slow results, and failing to address individual patient needs. The way new drugs are made is changed by using smart devices, adaptable algorithms and nanotechnology. AI helps to get better results and more accurate medicine (Gallego et al., 2021; Liu et al., 2020). Recent reviews said that adaptive control, smart sensors, and making drug formulations work better for different types of drugs and ways of giving them have all gotten better (Gallego et al., 2021; Barrett et al., 2023; Liu et al., 2023).

5.1 Personalized Drug Delivery

AI's ability to change the timing is a big step forward for personalized medicine. Most drug dosing models use the average for a large group of people. This approach doesn't take into account how each person absorbs the drug, their genes, or their health. AI systems, particularly those employing machine learning and reinforcement learning, can analyze extensive datasets such as genomic information, metabolic data, and clinical histories to formulate individualized dosing regimens (Liu et al., 2021; Gallego et al., 2021). New ideas for 2024 also connect programmable drugs and adaptive dosing to synthetic biology (Hill et al., 2024).

AI platforms in oncology monitor tumor markers, patient responses, and genetic predispositions to adjust chemotherapy doses in real time (Hill et al., 2024). AI algorithms can also figure out the best time to administer a drug, which can make treatment more effective through chronotherapy (Sarkar et al., 2023).

These changes are happening because the U.S. government supports the use of AI and ML in hospitals (Niazi, 2023; Liu et al., 2023).

5.2 Smart Drug Delivery Devices

AI helps with smart drug delivery devices. These devices monitor patients in real time and change medicine size based on what the patient needs (Barrett et al., 2023).

For example, AI-powered insulin pumps use data from continuous glucose monitoring to figure out when glucose levels will change and provide insulin when it is needed. AI-powered inhalers help people with asthma record their breath data, what makes their asthma worse, and how much medicine they need to take (Luo et al., 2022).

Another case is that people can eat sensors made by Proteus Digital Health. These sensors, along with AI analytics and wearable patches, monitor how drugs are consumed and related body reactions. This technology enables individuals to monitor drug effects in a closed loop (Burki, 2019).

Smart devices can also help with healthcare solutions that don't require you to be there in person. AI-based simulation platforms transformed the treatment of pulmonary hypertension during the COVID-19 pandemic by utilizing real-time patient data (Chakravarty et al., 2021). Regulatory perspectives emphasize the importance of documentation, performance assessment, and post-market education for adaptive, AI-driven devices (Niazi, 2023; Liu et al., 2023) (Figure 3).

Figure 3: In a closed loop, smart delivery works by sending data from sensors to an AI controller, which then moves an actuator. Patients can provide feedback, and there is no need to merge EHRs.

5.3 Nanotechnology and AI

Nanotechnology helps doctors administer medicine and lets them decide when and how to do it. Adding AI makes it work better to totally transform how we create and use nanoparticles. The result could mean they're more effective, safer for our bodies, and better suited for medical treatments (Gallego et al., 2021; Hao et al., 2023).

Machine learning assists us in choosing the best materials, surface coatings, and connections for nanoparticles. This makes sure they can safely target sick tissues, like tumors, without damaging healthy cells (Sexton et al., 2004). Scientists can use data from both in vitro and in vivo experiments to improve nanoparticles via deep learning algorithms with less testing (Lloyd, 2024).

AI is also vital for tracking nanoparticles' movement. It enables researchers to make predictions about the efficacy of the selected drug (Gaudelet et al., 2021). Scientists use AI and imaging technologies to learn how nanoparticles move through the body to get a better drug release strategy.

AI-based quality control for nano-drug production also checks the size and shape of nanoparticles and their packaging. These are all important steps to obtain the drugs approved by the government and verify that they work (Liu et al., 2021).

5.4 Controlled-Release and Formulation Optimization

AI can help us find the right mix by getting the right amounts of each ingredient, improving the manufacturing process, and designing the shape of the device. All of this is done to make drugs that work really well and have controlled-release profiles. Pipelines with cameras and sensors let us quickly find problems and make the changes that are needed. This makes the process much more reliable and predictable. Aerosol formulations have been used to get drugs to lungs, improve the solid dosage process, make new formulations, and predict disintegration (Lamy et al., 2018; Walsh et al., 2018; Lou et al., 2021; Momeni et al., 2024). AI-powered 3D printing can make dosing and release kinetics even more personalized (Serrano et al., 2023). Studies in mechanistic fluid dynamics indicate that localized concentration during syringe injection, coupled with shear forces, can destabilize protein injectables. People can utilize such information alongside ML screening to confirm that the device's design and formulation won’t induce aggregation (Xing et al., 2019). Simulations of tablet disintegration and dissolution grounded in physics, employing coupled Lattice Boltzmann and bonded particle models, deliver mechanistic precision that integrates effortlessly with data-driven surrogates. The outcome facilitates reliable "what-if" analyses in the selection of formulations and processes (Li, 2021; Li & Li, 2019). New hybrids, like physics-informed neural networks for multiphase LBM and graph-augmented LBM surrogates, present us useful approximations that keep the structure intact and are beneficial for optimization loops that address release and manufacturability (Li, 2025g; Li, 2025h; Li et al., 2025i) (Figure 4).

Figure 4: Controlled release and formulation optimization: a closed-loop design that uses AI, characterization, and regular updates to determine the right release and stability.

5.5 Conceptual Framework: Hybrid Data-Driven and Certifiable Control Systems

Next-generation AI drug delivery systems are distinct. They use machine learning models based on control-theoretic frameworks to verify. This hybrid method leverages control theory and neural networks, which are good at keeping track of drug transport in systems that are spread out over space (partial differential equations).

Recent advancements demonstrate that PDE-based controllers, including inverse optimal adaptive boundary control for reaction-diffusion systems and set-point regulation for Korteweg-de Vries-Burgers equations, can sustain stability and functionality even under uncertain conditions (Cai et al., 2025a; Cai et al., 2025b; Cai et al., 2025c). These certified controllers use digital twins and closed-loop delivery devices to make AI models easier to understand.

Figure 5 shows a conceptual architecture that shows how a single delivery platform can bring together AI-driven pharmacokinetic/pharmacodynamic modeling, multi-modal sensing (genomics, proteomics, imaging, wearables), and certified PDE controllers. This system is always learning from the information it has about each patient, but it never goes beyond what is formally stable. This feature lets you change the safety and optimization guarantees so that they obey the rules (Figure 5).

Figure 5: A system for distribution that uses AI and other tools. AI-driven PK/PD modeling and personalization engines employ multi-modal sensing to instruct a certified PDE controller on ensuring safe and adaptable dosing. Feedback loops are a wonderful way to learn and grow.

5.6 Predictive Analytics

AI-driven predictive analytics are changing how doctors evaluate how well medications work and what adverse effects they might have. AI systems can make educated guesses about how well some treatments will work on some patients or groups of patients by looking at data from clinical trials, real-world evidence, and multi-omics profiles (Gallego et al., 2021; Hong et al., 2023).

In the past, pharmacovigilance depended on post-marketing surveillance. AI can now identify possible safety signals before they happen. Natural language processing and other tools look through unstructured data from clinical notes, scientific papers, and social media to locate detrimental events and figure out how drugs work together (Liu et al., 2021; Liu et al., 2023). Such knowledge can help people receive warnings sooner and change labels (Schuhmacher et al., 2021; Roche et al., 2023) (Figure 6).

Figure 6: The pharmacovigilance signal detection pipeline uses many different kinds of data, such as natural language processing, extraction, transformation, loading, and machine learning, to determine signals and safety actions with feedback.

AI models also help doctors decide what to do by suggesting different treatments for each patient and finding groups of patients who are likely to respond well to certain therapies. AI has been used to identify new approaches to using old drugs to help people with COVID-19. For instance, AI biosimulation platforms for now can study how molecules interact with each other and how patients react (Chakravarty et al., 2021).

AI-powered analytics can help public health groups figure out the best ways to distribute drugs and predict how diseases will propagate. AI synthetic biology can also be used to develop therapeutic systems for monitoring diseases (Hill et al., 2024).

AI has improved drug delivery systems by making therapy safer and more responsive. Newly created nanoparticles and smart devices can predict the required dose of medicine. Such predictive analytics helps doctors lower risks. As AI gets better, doctors, AI model developers, and regulators need to work together to make sure that AI has the best possible effect on drug delivery systems (Gallego et al., 2021; Niazi, 2023; Barrett et al., 2023).

6. Discussion

AI is rapidly shifting healthcare from a reactive to a proactive and predictive system. Along with new biomedical advances like genomics, wearable sensors, and digital twin modeling, AI could offer personalized care in real time. This integration establishes a versatile environment, enabling personalized diagnosis and treatment (Marques et al., 2024; Mukherjee et al., 2024).

6.1 AI as a Catalyst in the Evolution of Precision Medicine

Precision medicine was previously focused on grouping people based on their genes and bodies. Now AI makes it dynamic, which is most obvious in oncology, where AI systems can now use molecular and clinical data to figure out whether treatments can be effective (Bhinder et al., 2021; Liu et al., 2019). Researchers have utilized AI models to forecast the response of individuals with melanoma to immunotherapy (Johannet et al., 2021). It’s also applied to evaluate the possibility of cardiac complications in breast cancer patients who receive anthracycline treatment (Chang et al., 2022).
AI-driven carcinogenicity prediction is advantageous during the preliminary stages of drug development, as it reduces the risks linked to compound pipelines (Le et al., 2024). These examples demonstrate that AI can facilitate clinical decision-making to enhance flexibility.

6.2 Genomic Integration and the Rise of Multi-Omics Intelligence

Genomic analysis is one of the hardest things to do in biomedical science because it needs a lot of data. There are millions of different versions of the human genome, and it's very complicated. We need tools to help us identify important patterns in it. AI can find polymorphisms that are linked to diseases and figure out polygenic risk scores (Marques et al., 2024; Bhinder et al., 2021).

Integrative AI models combine different omics layers, like transcriptomics and metabolomics, into a single analytical model. Iorio et al. (2016) and Kuenzi et al. (2020) have demonstrated that this form of integration can assist physicians in selecting the most effective targeted therapy and in forecasting the efficacy of various drugs in combination for cancer treatment (Iorio et al., 2016; Kuenzi et al., 2020). These changes make it easier to determine the right treatments and improve clinical trials by more effectively filtering patients.

6.3 Wearable Technologies and Real-Time Health Analytics

Making medical decisions via AI and wearable devices can be swift because they are doing inference based on real data. Monitoring physiological data such as glucose levels, heart rate variability, and sleep patterns can help predict the possibility of a disease (Lam et al., 2021; Stankoski et al., 2021).

Recent uses of wearable data include predicting the risk of Type 2 diabetes (Lam et al., 2021) and finding signs of preclinical Parkinson's disease (Schalkamp et al., 2023). Some people also suggested using AI to predict how anxiety affects blood pressure with wearable technology (Kargarandehkordi et al., 2024). They make it easier to have personalized ways to prevent problems, especially when it comes to managing long-term illnesses. Artificial intelligence and wearable technology can also address sleep issues (Garbarino & Bragazzi, 2024).

6.4 Digital Twins: Combining Simulation and Clinical Reality

Digital twins are popular AI-powered virtual biological systems. They are a new way to find out how treatments will work. It's safe to try out treatment plans without putting patients at risk, which are based on long-term genomics and imaging data.

Blasiak (2020) found that AI-driven platforms CURATE.AI could adjust patients’ medication dosage based on real-time biological data, therefore enhancing the precision (Blasiak, 2020). Radiomics-based machine learning has been used for forecasting the efficacy of medical therapy for lung cancer. There is a paper illustrating the utility of digital twins in oncology (Dercle et al., 2020). AI can simulate clinical scenarios in silico, which helps doctors make better decisions in real time and moves precision medicine forward.

6.5 Real-Time Adaptive Drug Discovery and Delivery Techniques

The traditional method of drug development consumes a lot of time and money. Fortunately, AI can sophisticatedly predict how well drugs will work and how stable their formulas will be. In reality, Deep Learning models can develop drug candidates with enhanced profiles (Gholap et al., 2024) and forecast tablet disintegration intervals (Momeni et al., 2024).

AI-powered platforms can give you real-time updates on your medications. For example, medicine is becoming more responsive through smart insulin pumps that automatically change the dose based on biosensor data (Aundhia et al., 2024). Another case is AI 3D printing, making personalized polypills to let people take various drugs at the same time (Serrano et al., 2023; Anaya et al., 2023).

AI has also helped with drug engineering by improving aerosol formulations so that they transport drugs to the lungs better (Lamy et al., 2018) and solid dosage forms by using machine learning to improve the process (Lou et al., 2021; Walsh et al., 2018). These novel concepts imply that intelligent and flexible technologies tailored to individual requirements will ensure the availability of treatments when required.

From a control-theoretic perspective, distributed-parameter dynamics govern many drugs transport and thermal-biointeraction phenomena. Recent advances in inverse optimal and adaptive boundary control for heat/reaction-diffusion systems, as well as set-point regulation for Korteweg-de Vries-Burgers (KdVB)-type PDEs, provide stability and performance guarantees under uncertainty (Cai et al., 2025c; Cai et al., 2025b; Cai et al., 2025a). Embedding such certified controllers within digital twins and closed-loop delivery devices could yield provably safe dosing and actuator policies that complement data-driven controllers.

7. Future Directions

We have not yet reached the full potential of AI to change the field of pharmaceutical sciences. New trends and research areas will shape the next generation of AI-powered drug discovery and delivery systems.

7.1 Integration of Multi-Modal Foundation Models

Large-scale foundation models trained on various biomedical datasets have enormous potential to transfer knowledge between fields and learn from only a few examples in drug design. Future research should focus on creating specialized foundation models for chemistry and pharmacology that can generalize across molecular spaces, predict novel interactions, and accelerate lead optimization through transfer learning from existing drug-target databases.

7.2 Closed-Loop Automated Discovery and Manufacturing

AI, robotics, and continuous manufacturing can create fully closed-loop systems for finding and making drugs automatically. These platforms will combine high-throughput screening, characterization, and formulation in real time, which can reduce development time a lot and quickly change things.

7.3 Explainable and Certifiable AI for Regulatory Acceptance

Regulators want AI-based choices to be safer and understandable. They want AI that everyone can agree on and understand. This means that AI systems need to be able to become certified and make promises about how well they work, how safe they are, and how fair they are. Neurosymbolic AI research integrates neural networks with symbolic reasoning. It's crucial for building models that are both understandable to human beings and capable of making accurate forecasts (Liu et al., 2020).

7.4 Federated Learning for Drug Development That Protects Privacy

Federated learning will assist schools in working together to identify new drugs without bringing all the data to a single place. This approach keeps the private data of individuals safe while enabling the models to learn from an extensive variety of datasets that are stored in various locations. Future research should focus on discovering efficient federated distillation methods that maintain the model's integrity while reducing the necessary communication.

7.5 Real-Time Adaptive Control for Drug Delivery

Using control-theoretic frameworks, like PDE-based controllers for systems that are spread out over space, will make drug delivery safe and flexible, and there will be formal guarantees of stability and performance. Using these certified controllers with data-driven models, you can make hybrid systems that are safe and simple to understand, like classical control, but also flexible, like machine learning.

7.6 Precision Medicine at Scale

Wearable tech, genomics, and digital health tech are becoming more and more common.  This means that AI systems will be able to help many people in a way that is special to them. The goal is to build infrastructure that can handle data integration, model deployment, and continuous learning while still keeping patient privacy safe and making it easy to quickly translate clinical information.

7.7 AI-Enhanced Intestinal Barrier Prediction and Drug Permeability Optimization

People's intestines need to be robust to be able to take medicine by mouth. Proteins that make tight junctions control how easily things can pass through the barrier. AI-based methods will use predictive models to figure out how strong a barrier will be and improve oral formulations so that they can be safely delivered without hurting anyone. There are enough data points about how PPAR? signaling routes and endocytic regulation change the shape of tight junctions, which can be used for AI models (Li & Ajuwon, 2021; Li et al., 2021).

AI can predict how tight junctions will change even when the related drug composition is changed, which makes it easier to make drugs better while keeping the barrier in place. Furthermore, AI models can also look at dietary compounds and formulation aids like quercetin and omega-3 fatty acids to see whether they can help protect the function of the intestinal barrier (Li et al., 2022; Li et al., 2023a; Li et al., 2023b). In the future, it will be possible to utilize AI pattern recognition to monitor the intestinal barrier's performance.

7.8 Ethical AI

When applying AI for drug development, the most important things are fairness and morality. You need to make sure that algorithms don't favor certain groups of people, that training datasets include people who aren't well-represented, and that there are ways to hold algorithms accountable. The main goal is to create a drug system that uses AI to help all patients, regardless of where they live, how much money they have, or what their genes are.

7.9 Challenges and Limitations

Researchers are using AI increasingly to study health care, but some big problems and rules need to be familiar before using it safely, ethically, and effectively. If these problems aren't resolved, they could make it much harder for AI to help people grow better and accelerate the process of making new drugs.

People should think about how to keep their personal information safe. AI usually utilizes a lot of patient data to analyze, such as genomic sequences, electronic health records, and real-time physiological signals. Putting all personal information in one place is a big risk to data security. Even if a dataset doesn't have identification information, it’s still possible to identify it via advanced AI models. When training models, companies often break data security laws such as the GDPR and HIPAA (Mak et al., 2024; Low et al., 2020).

Another big problem is that we don’t have enough well-labeled datasets. Healthcare data is always stored in various formats and locations or with biases against specific groups of people. For instance, trained models are easy to exhibit bias if they lack sufficient information of uncommon cases, which means that models don't work well in many clinical contexts (Hay et al., 2014; Vora et al., 2023). Developers need to know a lot about the field and work hard to get all needed information, which makes it even harder to make datasets that AI models can learn from (Xue et al., 2018).

Making models that are straightforward to understand is another big problem. Many AI models are actually black boxes, which means it’s always hard to explain why they make specific decisions. However, trust and responsibility are important in the healthcare industry. Lack of understanding puts patients at risk. Healthcare professionals are unlikely to adhere to AI recommendations without comprehending the underlying rationale, especially when such decisions may influence diagnoses, treatments, or prognoses (LeCun et al., 2015; Russell & Norvig, 2016). XAI is improving, but most current solutions don't provide clarifications, which are useful in clinical settings.

Legal and ethical concerns typically prohibit the use of AI for medical purposes. The regulations we currently have aren't adequate for assessing systems that evolve as they evolve over time. There are still many problems that need to be fixed, such as algorithmic drift, following people after an event, and figuring out who is responsible for AI mistakes (DiMasi et al., 2016; Fetzer, 1990). We need to think about a lot of big moral issues, such as who owns the data, fairness, equity, and the rights of patients. AI tools could make healthcare worse and make individuals less confident when too many government rules are involved (Alpaydin, 2021; Hay et al., 2014). Recent papers from lawmakers demonstrate that it is important to have rigorous data governance, like model cards (Mesko & Topol, 2023; Niazi, 2023; Liu et al., 2023).

All these problems illustrate that it is important for people from various professions to work together in developing plans about these technical, legal, and social issues when AI is utilized. Future research should focus not only on enhancing algorithms but also on improving data governance, increasing model transparency, and establishing unified regulatory frameworks.

8. Conclusion

AI has changed a lot about how drugs are made. It helps scientists find new medicines and make sure they get to people. It makes everything faster, cheaper, and more accurate, from finding the right target to giving the right dose. AlphaFold made a good guess about how proteins fold.  AI designed DSP-1181, the first clinical candidate, and drugs can be quickly repurposed. These represent a few of the most important things that have occurred. Wearables, genomics, AI, nanotechnology, and control theory could all work together to make the healthcare system flexible. We must address issues like security of data, how simple it is to understand, algorithmic bias, and rules for making this goal become a reality. The end goal is fair, accurate, and easy-to-understand treatment for all patients.

Author contributions

Y.L. and S.L. conceptualized the study and jointly supervised the work. R.T. and P.Z. reviewed AI methodologies and drug discovery applications. J.B. analyzed precision medicine and predictive analytics aspects. T.W. and P.G. compiled literature, synthesized findings, and prepared the manuscript.

Acknowledgment

The authors gratefully acknowledge the researchers and institutions whose open-access data, tools, and prior studies supported the development of this review.

References