INTRODUCTION:

The scientific field of artificial intelligence (AI) is concerned with intelligent machine learning, particularly with intelligent computer programs that produce outcomes a kind to those of human attention processes. A more modern and exciting method for drug research and design procedures is artificial intelligence (AI). The field of pharmaceutical research has made substantial use of machine learning, one branch of artificial intelligence. Because data is now readily available and stored in machines, scientists and researchers are more capable and skilled than before. The development and use of fundamental algorithms are a result of the advancement and use of machine learning. At a 1956 Dartmouth Conference, John McCarthy coined the term "artificial intelligence" to refer to the study of creating intelligent robots¹.

The worry of the prospect of unemployment is often linked to the advancement and invention of AI applications. But nearly every breakthrough in the use of AI technology is being hailed due to the industry's faith, which greatly increases the technology's effectiveness. Artificial intelligence (AI) has emerged as a critical component of industry, with practical applications across a wide range of technical and scientific domains. There has already been a transition from hype to hope in the emerging push to adopt the applications of AI technology in pharmacy, including drug delivery formulation development, drug discovery, and other healthcare applications. Utilizing AI models also enables the prediction of in vivo reactions, therapeutic pharmacokinetic characteristics, appropriate dosage, and other factors. Given the significance of drug pharmacokinetic prediction, the use of in silico models in drug research contributes to its affordability and efficacy. The advancements in AI technology fall into two main categories. The first one consists of traditional computing approaches, such as expert systems, which can mimic human experiences and provide examples of the conclusions drawn from the principles. The second one consists of artificial neural network (ANN)-based technologies that can simulate how the brain functions. Specifically, different artificial neural networks (ANNs) such as recurrent neural networks (RNNs) and deep neural networks (DNNs) govern how AI technology advances. DNN problems demonstrate higher predictivity than the baseline machine learning techniques in the NIH Tox21 and Merck Kaggle challenges. Machine learning uses appropriate statistical techniques that can learn both with and without explicit programming. De novo design also encourages the development of novel medicinal compounds with respect to ideal or desired properties. The uses of AI in pharmacy are covered in the current review paper, with a focus on polypharmacology, hospital pharmacy, drug delivery formulation development, and drug discovery¹.

· Classification of Artificial Intelligence:

Figure 1: Classification of Artificial Intelligence.

· Based on their caliber, AI system is classified as follows:

1. Artificial narrow intelligence (ANI): Also known as weak intelligence, is a system that is trained and developed to carry out a certain activity, including traffic signals, chess, driving, or facial recognition. For instance, tagging in social media and Apple SIRI, a virtual personal assistant.^2,3

2. Artificial General Intelligence (AGI): Strong AI, often known as Artificial General Intelligence (AGI) Another name for it is Human-Level AI. It has the power to simplify human intelligence. Because of this, it can solve problems when faced with new tasks. AGI is capable of all human functions.^2,3

3. Artificial Super Intelligence (ASI): This is mental capacity that surpasses that of intelligent humans in all domains, including science and the arts, mathematics, space exploration, and sketching. It can be as little as that of a human being or as much as a trillion times more intelligent than that of a machine.^2,3

· Based on the technology's presence and absence, an AI scientist categorized it. They are listed in the following order:

Type 1: An AI system of this kind is referred to as a reactive machine. For instance, the IBM chess algorithm Deep Blue defeated Garry Kasparov in the 1990s. It lacks the memory to draw on prior experiences, but it can recognize checkers on a chessboard and make predictions. It is useless in other circumstances and was intended just for specific uses. AlphaGo from Google is another example.^4,5

Type 2: A system with limited memory is this kind of artificial intelligence. This approach can apply lessons learned from the past to current and upcoming issues. Only this strategy is used in the design of various decision-making functions in autonomous cars. The acts that take place in the future, such an automobile changing lanes, are documented using the observations that were made. The observations are not retained in the mind indefinitely.^4,5

Type 3: The term "theory of mind" refers to this kind of AI system. It implies that every individual has thoughts, goals, and desires that influence the choices they make. This AI doesn't exist.^4,5

Type 4: This category includes self-awareness. The AI systems are sentient and have a feeling of self. If the machine is self-aware, it recognizes the situation and makes use of the concepts found in other people's minds. This AI doesn't exist.^4,5

Figure 2: Types of AI.

· Advantages of AI technology:

1. Error minimization: AI helps to more precisely raise accuracy and reduce errors. Intelligent robots are dispatched to explore space because their metal bodies are durable and they can withstand the harsh atmosphere there.^6,⁷.

2. Difficult exploration: The mining industry is one area where AI shows promise. The field of fuel exploration also makes use of it. Artificial intelligence (AI) systems have the ability to overcome human error and conduct ocean research.^6,⁷

3. Repeated tasks: People can typically handle one repetitive task at a time. Comparatively speaking, machines can do tasks that need multiple tasks at once and evaluate data faster than humans. It is possible to modify different machine characteristics, such as speed and time, based on specific needs.^6,⁷

4. Medical applications: Generally speaking, doctors can utilize AI programs to evaluate patients' conditions and examine side effects and other health hazards related to medication. Through the use of AI programs, such as different artificial surgical simulators (such as those that simulate the heart, gastrointestinal tract, brain, etc.), trainee surgeons can learn new techniques.^6,⁷

5. No risk: There is always a great possibility that the employees working in hazardous environments, such as fire stations, will be harmed. In the event of an accident, the machine learning systems can be used to repair damaged components.^6,⁷

6. Acts as aids: AI technology has taken on a new role by providing round-the-clock assistance to both elderly and children. It can serve as a resource for everyone's education.^6,⁷

7. Limitless functions: There are no limitations for machines. Emotionless machines are capable of greater efficiency and accuracy than humans in all aspects of production.^6,⁷

· Disadvantages of Artificial intelligence:

1. Expensive: AI adoption results in significant financial outlays. Complex machine design, upkeep, and repair are incredibly economical. The research and development division needs a lot of time to design a single AI machine. The software on an AI machine must be updated frequently. Reinstallations and machine recovery take a long time and cost a lot of money.^8,⁹

2. No human replication: Artificial intelligence (AI)-enabled robots are said to possess human-like thought processes and emotionlessness, which confers some benefits such as enhanced task accuracy and impartial performance. When unknown issues emerge, robots are unable to make decisions and may give inaccurate information.^8,⁹

3. Experience does not improve human resources: Experience does improve human resources. On the other hand, AI-powered robots are incapable of learning from experience. They are unable to distinguish between the hard-working and non-working individuals.^8,⁹

4. Lack of inventiveness: AI-powered machines lack both emotional intelligence and sensitivity. People are able to see, hear, feel, and think. They are able to employ both their imagination and reasoning. Machines are not capable of achieving these features.^8,⁹

5. Unemployment: The extensive application of AI technology across all industries could lead to a significant increase in joblessness. Due to unfavorable unemployment, employees may become less creative and prone to bad work practices.^8,⁹

· Application of Artificial Intelligence:

1. Drug Discovery and Design: The features or activities of novel drug candidates can be predicted using supervised learning techniques. The model may identify patterns and connections between desired outcomes and molecular properties by training on a dataset of known substances and the actions that go along with them. This helps in drug discovery and design by making it possible to anticipate the activity, potency, or toxicity of novel molecules.^10,11

2. Predictive Maintenance and Quality Control: Supervised learning can be applied to pharmaceutical manufacturing to support quality control and predictive maintenance. The model can be trained to forecast equipment failure, product quality deviations, or process anomalies through training on data from production processes, equipment sensor data, or quality testing findings. This enables proactive maintenance and quality assurance.^10,11

3. Drug Target Identification: By examining biological data, supervised learning algorithms can assist in the identification of possible drug targets. The model can learn patterns and indicate possible targets for more research by being trained on data pertaining to genomic, proteomic, or transcriptomic traits and their relationship to treatment response or illness progression.^10,11

4. Disease Diagnosis and Prognosis: Based on medical data, supervised learning models can be used to forecast patient outcomes or diagnose illnesses. The model is able to predict the course of the disease or the response to treatment by training on labeled datasets that comprise patient characteristics, clinical data, and disease outcomes.^10,11

5. Adverse Event Detection: Pharmacovigilance data can be used to identify and categorize adverse events linked to medications by using supervised learning algorithms. The model can assist in the identification and characterization of adverse events by learning to spot patterns and potential safety signals by training on labeled adverse event reports.^10,11

6. Predictive Modeling for Clinical Trials: Clinical trial outcomes can be predicted through the use of supervised learning. The model can be trained to predict patient response, treatment efficacy, or safety outcomes using historical clinical trial data, which includes patient characteristics, treatment interventions, and trial outcomes. This data can improve patient selection and serve as a guide for trial design.^10,11

Figure 3: Application of AI.

· AI in targeted genomic therapy and diagnosis:

AI is used in hospital-based health care systems in a variety of ways, including the organization of dosage forms for specific patients and the selection of appropriate or practical administration routes or treatment plans.¹²

a. Maintaining of medical records: It might be difficult to keep up with patients' medical records. By using the AI system, data collection, storage, normalization, and tracking are made simpler. The Google Deep Mind health initiative facilitates the expeditious excavation of medical records. Thus, this project is helpful in providing faster and better healthcare. This project helps to improve eye treatment at the Moorefield’s Eye Hospital NHS.^13,14

b. Treatment plan designing: AI technology makes it feasible to create treatment programs that are both effective and efficient. An artificial intelligence (AI) system is required to take control of the situation when a patient develops a severe condition and choosing an appropriate treatment plan becomes challenging. The treatment plan provided by this technology takes into account all of the prior data and reports, clinical expertise, etc. The software as a service, IBM Watson for Oncology, is a cognitive computing decision support system that compares patient data to thousands of past cases and insights gained from working with Memorial Sloan Kettering Cancer Center physicians for thousands of hours. It then offers treatment options to oncology clinicians so they can make well-informed decisions. The literature that Memorial Sloan Kettering has selected to support these therapeutic alternatives includes more than 200 textbooks, over 300 medical periodicals, and about 15 million pages of text.^13,14

c. Assisting in repetitive tasks: AI technology also helps with some repetitious activities, such analyzing radiography, ECHO, ECG, and X-ray imaging to identify and detect diseases or abnormalities. IBM introduced Medical Sieve, an algorithm that functions as a "cognitive assistant" with strong analytical and reasoning skills. In order to combine deep learning with medical data and enhance the patient's condition, a medical startup is required. For every bodily component, there is a separate computer program that is employed in a certain illness state. For practically any kind of imaging analysis, including X-ray, CT, ECHO, ECG, etc., deep learning can be used.^13,14

d. Health support and medication assistance: AI technology has been shown to be effective in recent years for both pharmaceutical assistance and health support services. Molly, a virtual nurse created by start-ups, is greeted with a friendly face and voice. Its goal is to support patients with their chronic ailments during doctor appointments and assist them in directing their own treatment. An program called AI Cure, which works with a smartphone's webcam, keeps track of patients and helps them manage their ailments. Patients who take part in clinical trials and those with severe drug conditions can both benefit from this app.^13,14

e. Accuracy of medicine: AI has a positive effect on genetic development and genomics. Using patterns found in genetic data and medical records, Deep Genomics, an AI system, can be used to find mutations and their connections to diseases. This technique provides physicians with information on what happens inside a cell when genetic variation modifies DNA. Craig Venter, the creator of the human genome project, creates an algorithm that uses a patient's DNA to provide physical traits. When vascular illnesses and cancer are still in their early stages, "Human Longevity" AI technology can be used to pinpoint their precise location trials.^13,14

f. Drug creation: Pharmaceuticals require billions of rupees and more than ten years to manufacture or create. The AI program "Atomwise," which makes use of supercomputers, is helpful in determining treatments from the molecular structure database. It launched a virtual search campaign for an Ebola virus treatment that is both safe and effective using already available medications. Two medications that led to an Ebola infection were found using technology. In contrast to months or years when analysis was done by hand, this study was finished in a single day. Big data was created by a Boston-based Biopharma company for patient management. It stores information to determine the causes of some patients' illness survival. They distinguished between air conditions that are conducive to health and those that are conducive to sickness using biological data from patients and artificial intelligence. It supports applications for problem-solving, healthcare, and medication development and creation.^13,14

g. AI helps people in the health care system: - In 2016, one of the ten most promising technologies was the "open AI ecosystem." Compiling and contrasting the data from social awareness algorithms is helpful. A great deal of data, including treatment history and patient medical history from childhood until that age, is documented in the healthcare system. Ecosystems can analyze this massive amount of data and provide recommendations regarding the patient's lifestyle and behaviors.^13,14

h. Healthcare system analysis: Data retrieval is made simple in the healthcare system if all of the data is computerized. Ninety-seven percent of Dutch invoices which include treatment information, doctor and hospital names are kept on file digitally. As a result, these are easily retrievable. Zorgprisma Publiek, a nearby business, uses IBM Watson cloud technology to analyze the bills. In the event of an accident, it detects it right away and reacts appropriately. As a result, it enhances and prevents hospitalization of patients.^13,14

· AI in Drug Discovery:

Lack of suitable technology makes it difficult to build many drug compounds from a chemical space; however, this can be fixed by utilizing AI to speed up the drug development process. The various parameters' forecasting activities, such as log P or log D, which can anticipate and generate predictions through computations and justify the biological safety, efficacy, and adverse effects, including the pharmacokinetics of the significant molecule, are influenced by the quantitative structure–activity relationship. The vast area necessitates the delocalization of molecules based on their three-dimensional distribution and attributes. It is best to gather all previous data from many domains, such as PubChem, ChemBank, DrugBank, and ChemDB, addressing the selectivity and molecular positioning for demonstrating the bioactivity. Virtual screening is done using a variety of in silico techniques, which typically yield better analysis, faster elimination, and a wider selection. When choosing a lead compound to bind with and produce activities, drug design algorithms take into account the physical, chemical, and toxicological profiles. The biological activity and effectiveness can be enhanced by several physicochemical qualities. With AI-based QSAR techniques, QSAR is optimized for the possible use of the drug candidate. Controlling the biological activity that has been identified and developed may take ten years if the conventional methods for getting statistical differences are adhered to. When developing a new medication, factors such as inherent permeability, degree of ionization, partition coefficient, and solubility influence target receptor binding. To predict the binding qualities, algorithms use molecular descriptors like the Simplified Molecular Input Line Entry System (SMILES). The six physicochemical properties, referred to as the Estimation Program Interface Suite, are typically determined using a quantitative structure–property relationship (QSPR). The lipophilicity and solubility of different substances have been predicted using deep learning and neural networks based on the ADMET predictor and ALGOPS program. A lot of undirected graphs are used in solubility prediction. When predicting a new chemical entity, factors such as surface area, mass, hydrogen count, refractivity, volume, log P, surface area, sum of the indices, solubility index, and rotatable bonds are taken into account.^15,16

· AI in Digital Therapy/Personalized Treatment:

AI has the ability to identify important relationships in the raw datasheets that may be applied to the disease's diagnosis, mitigation, and therapy. Numerous more recent methods utilized in this developing field of computational knowledge have the potential to be employed in nearly every area of medical study. The difficulty of gathering, evaluating, and using a wealth of knowledge must be met in order to tackle the complicated clinical problems. The advancement of AI in medicine has aided medical professionals in resolving challenging patient issues. Healthcare professionals can receive assistance in manipulating data from systems like artificial neural networks (ANNs), evolutionary computing, fuzzy expert systems, and hybrid intelligent systems. The organic nervous system serves as the foundation for the artificial neural network (ANN). Neural networks are a type of interconnected computer processor network that can process input in parallel using computations. A binary threshold function was used in the development of the first artificial neuron. The most widely used model with many layers—input, middle, and output—was the multilayer feed-forward perceptron. Every neuron is linked by connections with a numerical weight.^17,18

· Utilizing AI to Forecast Epidemics and Pandemics:

Pandemics have no boundaries and can result in both morbidity and death. Numerous pandemic outbreaks have occurred worldwide, including the COVID-19 outbreak, the Spanish flu, the Black Death, cholera, influenzas, and AIDS. These outbreaks have the potential to disrupt economies and societies. Early diagnosis and effective treatment of the illness are critically dependent on each other, which lessens the toll that it has on people's health as well as on the political, social, and economic institutions. One of the main tools for achieving early detection is surveillance. Extensive resources, labor, and time are required for active surveillance. It is difficult to forecast epidemics and pandemics in real life. But because to recent developments, it is now possible to examine the spread of terrible diseases. The greatest way to accomplish surveillance while using resources efficiently is with artificial intelligence (AI). Deep learning and machine learning are being used in several healthcare sectors and are proven to be more efficient than human resources. Because of their complexity, developing epidemiological models continues to be difficult. Recently, models for predicting outbreaks have been developed using machine learning. AI is being applied in pandemic and epidemic detection, prevention, response, and recovery. Prediction, surveillance, and information are starting to be widely employed in prevention, particularly in light of the most recent COVID-19 outbreak. Because of its fluctuating epidemic peaks, recurrent peaking, etc., influenza epidemic forecast is never easy. Even in regions with variable seasonal influenza, an accurate forecast is achievable with the integration of the SAAIM (self-adaptive AI model). For instance, machine learning and ensemble techniques have been employed in Taiwan to accurately estimate seasonal influenza cases. The influenza forecasting output precision of the machine learning feed-forward propagation neural network model (MSDII-FFNN) is 90%. Australia and the USA have used machine learning anonymized mobility map (AMM) to forecast influenza cases. By combining data from smartphones, AMM is able to forecast epidemics by taking into account human mobility across state lines. Ebola remains a threat in Africa. Numerous methods have been used to forecast Ebola, including a hybrid neural network created by Umang Soni et al. that exhibits 100% accuracy when random forest is used as a classification method. Reliable findings in predicting the propagation have been obtained by integrating machine learning with experimental models involving artificial societies. For instance, the course of the Ebola outbreak has been anticipated based on research done on a simulated version of Beijing. Due to the lack of trustworthy forecasts, allocating surveillance resources during the 2015 Zika outbreak was extremely difficult. Later, to predict the spread, a dynamic neural network model was employed. This adaptable prediction model framework proved to be dependable and valuable throughout the early stages of the pandemic. In the Zika project, mobile applications were utilized to track the number of mosquitoes, and artificial intelligence neural networks were used to detect cases early. The results of vaccine-derived poliovirus (VDPV) monitoring have drawn interest. A VDPV epidemic can be predicted by combining the whale optimization algorithm (WOA) with random vector functional link (RVFL) networks in hybrid machine learning. Machine learning (ML) offers the ability to identify probable candidates for pre-exposure prophylaxis in HIV/AIDS prevention strategies. In tropical and sub-tropical regions, dengue fever is common. Support vector regression (SVR), a machine learning technique, can track and predict dengue outbreaks in China with very little error. The best predictor for dengue in Malaysia was the ML Support Vector Model (SVM) with a linear kernel; Bayesian network machine learning techniques were used to predict dengue outbreaks. With TB suspicious data, the ANN is integrated for quick diagnosis; its total efficacy exceeds 94%. This will facilitate the prompt implementation of some control measures and aid in the detection of the disease's overall spread. The tuberculosis AI (TB-AI) CNN model detected tuberculosis bacillus with 97.94% sensitivity by utilizing deep learning and machine learning. With seven psychological symptoms of yellow fever, the Multilayer Perceptron Neural Network Classifier (MPNN) was proposed as a diagnostic tool and was able to reach 88% prediction accuracy. The COVID-19 outbreak took the globe by surprise. The COVID-19 was predicted using modified stacked auto-encoder modeling inspired by AI. Using the fuzzy rule in conjunction with deep learning Composite Monte Carlo (CMC) was beneficial for making decisions and forecasting the COVID-19 epidemic. To forecast with very little error, a polynomial neural network with corrective feedback (PNN + CF) is employed. China uses CNN, a deep neural network with accurate prediction efficacy. In Switzerland, COVID-19 predictions are made using Big Data and the AI model Enerpol. Statistical and deep learning systems, including feed-forward neural networks (FNN), multilayer perception (MLP), autoregressive integrated moving averages (ARIMA), and long short-term memory (LSTM), were merged to study the dynamical behavior of COVID-19. The generated data may serve as a helpful point of reference for the COVID-19 forecast.^19,20,21

· AI Validation in Medicine:

AI has demonstrated promise in forecasting the medications for which various algorithms have received approval. Nevertheless, because newly created algorithms must satisfy the clinical criteria set forth by regulatory and professional authorities, there is a very low approval rate for new algorithms. For the advanced algorithms, testing and external validation are required procedures. In addition to the TRIPOD Checklist, this is the current model for the analytical method of multi-variable prediction in medicine. There are a few other well-known techniques. The U.S. Food and Drug Administration (FDA) authorized the WAVE Clinical Platform in 2018, which combines patient data with real-time vital sign data. The algorithm identified the hospital's in-house patients who were susceptible to unstable vital signs. Regarded as the first AI algorithm product, WAVE was the main predictive investigation program utilized for electronic health records (EHR). The FDA approved it based on the most likely evidence. However, compared to the analysis of diagnostic imaging, the FDA has approved fewer advanced algorithms for use in clinical practice. The algorithms get many variables as inputs. Thus, unlike a gadget or medication, algorithms are dynamic because they can change their predicted results in response to new inputs of data. As a result, the FDA has established new regulatory agendas for novel diagnostic techniques. To authenticate biomarkers for medication development and testing, the FDA established the Biomarker Qualification Program. The five notable indicators—oxygen saturation, blood pressure, heart rate, temperature, and respiration rate—are utilized by the various health systems and are part of the WAVE algorithm. There are various sophisticated machine learning methods for imaging parameters or EHR, but they are highly specialized and cannot be used to other EHRs. The clinical staff's capacity to react to a prediction algorithm's outcome is further hampered by the diverse user interface and various operational constraints in a distinct clinic setting. Therefore, the regulatory authorities' identification of the EHR inputs can improve the predicted outputs. When developing algorithms, the input variables must to be quite specific in order to obtain dependable results for the entire company. Regulators ought to be open about the private rights and intellectual property of the algorithm developers. Instead of being dependent on data from a single institution, the algorithms' data should come from a multiethnic community. As a result, large populations' worth of data are needed to train the algorithms. Like licensed pharmaceuticals that go under inspection after going through clinical studies, the prediction algorithm should also be subject to audits following FDA approval. This method can explain the change in the predicted result because with time, deep learning algorithms' tools will take into consideration new factors. Continuous audits can help reduce the systemic biases. It is necessary to perform the algorithmic analysis on both synthetic and anonymous data. The regulatory boards that handle intellectual property should carry out post-marketing audits.^22,23

· AI in clinical trial design:

Clinical trials take six to seven years and a significant financial commitment to establish the safety and efficacy of a medicinal product in people for a specific illness condition. But only one in ten of the molecules that are put through these trials get cleared, which is a huge loss for the industry. Inadequate infrastructure, outdated technical requirements, and poor patient selection are all potential causes of these failures. Nevertheless, by using AI, these errors can be minimized thanks to the abundance of digital medical data that is currently accessible.^24,25

One third of the trial's duration is devoted to patient enrollment. Finding the right patients for a clinical study is essential to its success; without them, 86% of trials end in failure. Using patient-specific genome exposome profile analysis, AI can help select only a particular diseased population for recruitment in Phase II and III clinical trials. This can help in early prediction of the available therapeutic targets in the selected patients. Preclinical molecular discovery and early prediction of lead compounds that would pass clinical trials with consideration for the chosen patient population are made possible by other AI-based methods like predictive machine learning and other reasoning techniques.^24,25

Thirty percent of clinical trials fail due to patient dropout, which causes extra recruiting requirements to be met in order to complete the trial and wastes money and time. By closely monitoring the patients and assisting them in adhering to the intended protocol of the clinical research, this can be prevented. Ai Cure created mobile software to track schizophrenia patients' consistent medication intake during a Phase II trial. This resulted in a 25% increase in patient adherence and the successful completion of the clinical trial.^26,27

· Pharmaceutical market of AI:

Pharma companies are turning to AI in order to reduce the expensive burden and failure rates associated with VS. From US$200 million in 2015 to US$700 million in 2018, the AI market grew, and it is projected to reach $5 billion by 2024. AI is predicted to expand by 40% between 2017 and 2024, which means it will probably completely transform the pharmaceutical and medical industries. Numerous pharmaceutical corporations have invested in artificial intelligence (AI) and are still doing so. They have also worked with AI startups to provide vital healthcare technologies. One example of this is the partnership between the Royal Free London NHS Foundation Trust and Deep Mind Technologies, a Google company, for the treatment of acute kidney injury.^28,29,30

CONCLUSION:

In recent years, there has been a noticeable increase in interest in the application of AI technology for the analysis and interpretation of several significant pharmacy fields, such as development of medications, pharmaceutical dosage form design, polypharmacology, pharmacy for a hospital etc., These applications of AI technology are based on the idea that humans are capable of imagining knowledge, solving problems, and making decisions. It has been demonstrated that automated databases and workflows may be used to conduct efficient analysis using AI techniques. The application of AI techniques has made it possible to build new tactics, hypotheses, and assessments of numerous related elements with ease and at a lower cost, all while consuming less time.

Drug delivery systems are being revolutionized by AI, opening the door to individualized, tailored, and adaptable medicines. Pharmaceutical researchers and medical professionals can increase drug efficacy, reduce adverse effects, and improve patient outcomes by utilizing AI's strengths in data analysis, pattern identification, and optimization. The development of AI and its amazing tools continually tries to lessen the difficulties that pharmaceutical businesses confront, which has an impact on both the medication development process and the product's total lifetime. This could account for the rise in industry start-ups. The rising cost of medications and treatments is only one of the many complicated issues the healthcare industry is currently facing. As a result, society needs to make some very big adjustments in this area. AI-enabled pharmaceutical product production allows for the customization of drugs to meet the specific needs of each patient by controlling the dose, release parameters, and other necessary factors. Automation will become even more crucial as a result of the latest AI-based technologies, which will expedite the time it takes for products to reach the market while also improving product quality, safety throughout the production process, and cost-effective resource utilization. The primary concern surrounding the integration of these technologies is the potential loss of jobs and the stringent laws required to integrate artificial intelligence. But these tools aren't meant to take the place of people entirely they're just meant to make things easier. AI can help with hit compound identification quickly and easily, as well as with suggesting synthesis methods for these molecules, predicting the needed chemical structure, and comprehending drug-target interactions and their SAR.

AI can also significantly aid in the optimization and continued integration of the developed medication in the appropriate dosage form. Furthermore, AI can facilitate prompt decision-making, which can expedite the production of higher-quality products and ensure batch-to-batch consistency. Through thorough market analysis and prediction, AI can also help prove the product's safety and efficacy in clinical trials and ensure appropriate placement and costing in the market. AI is not yet on the market, and there are still some obstacles to overcome before this technology can be widely used, but it is very likely that in the not-too-distant future, the pharmaceutical sector will find AI to be a very useful tool.

REFERENCES:

1. Mak KK, Pichika MR. Artificial intelligence in drug development: present status and future prospects. Drug discovery today. 2019 Mar 1; 24(3): 773-80.

2. Sahu A, Mishra J, Kushwaha N. Artificial intelligence (AI) in drugs and pharmaceuticals. Combinatorial Chemistry & High Throughput Screening. 2022 Sep 1; 25(11): 1818-37.

3. Shakya DS. Analysis of artificial intelligence-based image classification techniques. Journal of Innovative Image Processing. 2020 Apr 28; 2(1): 44-54.

4. Chen M, Decary M. Artificial intelligence in healthcare: An essential guide for health leaders. InHealthcare management forum 2020 Jan (Vol. 33, No. 1, pp. 10-18). Sage CA: Los Angeles, CA: SAGE Publications.

5. Morrow E, Zidaru T, Ross F, Mason C, Patel KD, Ream M, Stockley R. Artificial intelligence technologies and compassion in healthcare: A systematic scoping review. Frontiers in Psychology. 2023 Jan 17; 13: 971044. Sunarti S, Rahman FF, Naufal M, Risky M, Febriyanto K, Masnina R. Artificial intelligence in healthcare: opportunities and risk for future. Gaceta Sanitaria. 2021 Jan 1; 35: S67-70.

6. Paul D, Sanap G, Shenoy S, Kalyane D, Kalia K, Tekade RK. Artificial intelligence in drug discovery and development. Drug discovery today. 2021 Jan; 26(1): 80.

7. Manikiran SS, Prasanthi NL. Artificial Intelligence: Milestones and Role in Pharma and Healthcare Sector. Pharma times. 2019; 51: 9-56.

8. Cherkasov A, Hilpert K, Jenssen H, Fjell CD, Waldbrook M, Mullaly SC, Volkmer R, Hancock RE. Use of artificial intelligence in the design of small peptide antibiotics effective against a broad spectrum of highly antibiotic-resistant superbugs. ACS chemical biology. 2009 Jan 16; 4(1): 65-74.

9. Das S, Dey R, Nayak AK. Artificial intelligence in pharmacy. Indian Journal of Pharmaceutical Education and Research. 2021 Apr 1; 55(2): 304-18.

10. https://www.hcltech.com/technology-qa/what-are-the-advantages-of-artificial-intelligence

11. Dara S, Dhamercherla S, Jadav SS, Babu CM, Ahsan MJ. Machine learning in drug discovery: a review. Artificial Intelligence Review. 2022 Mar; 55(3): 1947-99.

12. Kavasidis I, Lallas E, Gerogiannis VC, Charitou T, Karageorgos A. Predictive maintenance in pharmaceutical manufacturing lines using deep transformers. Procedia Computer Science. 2023 Jan 1; 220: 576-83.

13. Bagherian M, Sabeti E, Wang K, Sartor MA, Nikolovska-Coleska Z, Najarian K. Machine learning approaches and databases for prediction of drug–target interaction: a survey paper. Briefings in bioinformatics. 2021 Jan; 22(1): 247-69.

14. Kumar Y, Koul A, Singla R, Ijaz MF. Artificial intelligence in disease diagnosis: a systematic literature review, synthesizing framework and future research agenda. Journal of ambient intelligence and humanized computing. 2022 Jan 13: 1-28.

15. Chapman AB, Peterson KS, Alba PR, DuVall SL, Patterson OV. Detecting adverse drug events with rapidly trained classification models. Drug safety. 2019 Jan 21; 42: 147-56.

16. Elkin ME, Zhu X. Predictive modeling of clinical trial terminations using feature engineering and embedding learning. Scientific reports. 2021 Feb 10; 11(1): 3446.

17. Das S, Dey A, Pal A, Roy N. Applications of artificial intelligence in machine learning: review and prospect. International Journal of Computer Applications. 2015 Jan 1; 115(9).

18. Hassanzadeh P, Atyabi F, Dinarvand R. The significance of artificial intelligence in drug delivery system design. Advanced drug delivery reviews. 2019 Nov 1; 151: 169-90.

19. Russell S, Dewey D, Tegmark M. Research priorities for robust and beneficial artificial intelligence. AI magazine. 2015 Dec 31; 36(4): 105-14.

20. Duch W, Setiono R, Zurada JM. Computational intelligence methods for rule-based data understanding. Proceedings of the IEEE. 2004 Apr 19; 92(5): 771-805.

21. Dasta JF. Application of artificial intelligence to pharmacy and medicine. Hospital pharmacy. 1992 Apr 1; 27(4): 312-5.

22. Jiang F, Jiang Y, Zhi H, Dong Y, Li H, Ma S, Wang Y, Dong Q, Shen H, Wang Y. Artificial intelligence in healthcare: past, present and future. Stroke and vascular neurology. 2017 Dec 1; 2(4).

23. Gobburu JV, Chen EP. Artificial neural networks as a novel approach to integrated pharmacokinetic—pharmacodynamic analysis. Journal of pharmaceutical sciences. 1996 May; 85(5): 505-10.

24. Sakiyama Y. The use of machine learning and nonlinear statistical tools for ADME prediction. Expert opinion on drug metabolism & toxicology. 2009 Feb 1; 5(2): 149-69.

25. Agatonovic-Kustrin S, Beresford R. Basic concepts of artificial neural network (ANN) modeling and its application in pharmaceutical research. Journal of pharmaceutical and biomedical analysis. 2000 Jun 1; 22(5): 717-27.

26. Zhang ZH, Wang Y, Wu WF, Zhao X, Sun XC, Wang HQ. Development of glipizide push-pull osmotic pump-controlled release tablets by using expert system and artificial neural network. Yao xue xue bao= Acta Pharmaceutica Sinica. 2012 Dec 1; 47(12): 1687-95.

27. Ma J, Sheridan RP, Liaw A, Dahl GE, Svetnik V. Deep neural nets as a method for quantitative structure–activity relationships. Journal of chemical information and modeling. 2015 Feb 23; 55(2): 263-74.

28. Mayr A, Klambauer G, Unterthiner T, Hochreiter S. DeepTox: toxicity prediction using deep learning. Frontiers in Environmental Science. 2016 Feb 2; 3: 80.

29. Bishop CM. Model-based machine learning. Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences. 2013 Feb 13; 371(1984): 20120222.

30. Merk D, Friedrich L, Grisoni F, Schneider G. De novo design of bioactive small molecules by artificial intelligence. Molecular informatics. 2018 Jan; 37(1-2): 1700153.

Received on 21.08.2024 Revised on 09.09.2024

Accepted on 24.09.2024 Published on 17.12.2024

Available online on December 23, 2024

Asian Journal of Pharmaceutical Research. 2024; 14(4):403-410.

DOI: 10.52711/2231-5691.2024.00064

This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. Creative Commons License.