Abstract
Artificial intelligence (AI) is a technology that combines machine learning (ML) and deep learning. It has numerous usages in the domains of medicine and other sciences. Artificial intelligence can forecast the behavior of a drug's target protein and predict its desired physicochemical qualities. AI's potential to enhance healthcare services offerings formerly unheard-of opportunities for cost reserves, enhanced overall clinical and patient outcomes. The recent development of research in the biomedical field, encompassing fields such as genomics, computational medicine, AI, and algorithms for learning, has led to the demand for novel technology, a fresh workforce, and new standards of practice set the stage for the revolution in healthcare. By connecting these health statistics with cutting-edge AI technologies, precise insights into patient treatment can be obtained. Moreover, AI can aid in the search for new drugs by foretelling the target protein's two-dimensional structure. In the current review, an overview of the latest AI-based technologies and how they may be employed to reduce product development time to market and snowballing product quality, cost-effectiveness, as well as security throughout the manufacturing process is detailed.
Graphical Abstract
Keywords: Artificial intelligence, Computational medicines, Drug development, Healthcare industry, Machine learning
Introduction
Artificial intelligence (AI) describes the replication of human intelligence in computers designed to think and act like people. It is a technique that allows machines to simulate various complex human abilities that were initially proposed in 1956 at a meeting organized by Marvin Minsky and John McCarthy [1]. Artificial intelligence (AI) plays a crucial role in various aspects of pharmaceuticals and healthcare, including drug discovery, patient data analysis, and production processes. Drug discovery is categorized into (i) Virtual Screening: AI algorithms can analyze large chemical databases to predict the potential of molecules as drug candidates, saving time and resources in the early stages of drug discovery, (ii) Drug Design: AI can assist in designing novel drug compounds with specific properties, optimizing them for efficacy and safety, (iii) Target Identification: by analyzing biological data, AI can identify novel drug targets and pathways for therapeutic intervention, accelerating the process of target selection, (iv) Optimization of Clinical Trials: AI can analyze patient data to optimize clinical trial design, identify suitable patient populations, and predict trial outcomes, leading to more efficient and cost-effective trials. Moreover, for Patient Data Analysis, AI assists in (i) Personalized Medicine: AI can analyze patient data, including genomic, clinical, and lifestyle information, to tailor treatments to individual patients, improving efficacy and minimizing side effects, (ii) Disease Diagnosis and Prognosis: AI algorithms can analyze medical images, electronic health records (EHRs), and other patient data to assist in disease diagnosis, prognosis, and treatment planning, (iii) Predictive Analytics: AI can predict patient outcomes, such as the likelihood of hospital readmission or disease progression, enabling proactive interventions and personalized care plans [2].
Furthermore, for the production process, AI assists in (i) Process Optimization: AI can optimize manufacturing processes in pharmaceutical production, improving efficiency, reducing costs, and ensuring product quality and consistency, (ii) Supply Chain Management: AI algorithms can optimize supply chain logistics, predicting demand, managing inventory, and identifying potential disruptions to ensure a steady supply of medications, (iii) Quality Control: AI can analyze data from production processes in real-time to detect deviations from quality standards and prevent defects or contamination [3]. In all the above-mentioned areas, AI empowers pharmaceutical companies and healthcare providers to make data-driven decisions, accelerate innovation, and improve patient outcomes. However, it's essential to address challenges such as data privacy, regulatory compliance, and the need for interdisciplinary collaboration to fully realize the potential of AI in healthcare and pharmaceuticals. In the pharmaceutical industry, the digitization of data has increased rapidly in recent years. However, with digitization, the information is problematic to learn, research, and use to resolve challenging clinical issues [4]. The fact that AI can automate substantial amounts of data makes it more conducive to its regular use [5].
Additionally, AI approaches are the most effective in diagnosing various disorders. Artificial intelligence (AI) offers unheard-of chances to save costs, improve patient and clinical group outcomes, and improve healthcare services. The models utilized extend beyond computerization and include "family" models for patients [6, 7], and medical care professionals for data generation and recommendations, as well as data sharing for collaborative assessment structure. AI can also pinpoint specific populations or geographic regions with higher-than-average rates of disease or high-risk behaviors. Investigators have successfully computed correlations between the constructed atmosphere and the prevalence of obesity by using deep learning classifiers in diagnostic procedures [8, 9]. Artificial Intelligence is a technology that uses a variety of sophisticated tools and networks to mimic human intelligence. But it doesn't threaten to take the place of a person's physical presence [10, 11]. To make judgments on their own and accomplish preset objectives, hardware and software that can grasp, analyze, and learn from input data are used in artificial intelligence applications. The intricacy involved in the formulation and manufacturing of pharmaceuticals has led to a growing use of artificial neural networks, or ANNs, for behaviour simulation [12]. Neural network models and genetic algorithms have proven to be incredibly effective when the formulator has to come up with a formulation that meets tight, frequently conflicting requirements. The optimization process's objectives can be easily and explicitly defined with the help of fuzzy logic. Fuzzy logic can be used to make hard pills that dissolve quickly, for instance [13].
More recently, novel techniques such as neuro-fuzzy logic have been created, combining the power of fuzzy logic to represent complex ideas simply with neural networks' capability to "learn" from data. These methodologies enable the "mining" of data for information and its presentation as applicable rules that can direct future research. The emergence of biomedical science—including genomics, digital medicine, artificial intelligence, and its subset, machine learning (ML)—and the resulting need for new skills, a new labour force, and new standards of practice provide the foundation for the transformation of healthcare. Precision medicine models, therapies, care delivery, regenerative treatment, and diagnostics can all be enhanced and revolutionized by genomics, biometrics, tissue engineering, and the vaccine industry [14]. By giving them access to patient data, digital health solutions give medical professionals a more comprehensive picture of their patients' health. Additionally, they enable doctors to give their patients more health-related facts. These present significant chances to enhance treatment results and effectiveness, however there are concerns that these modalities may have greater psychological problems, especially given how frequently patients, the public, and professionals utilize social media (SM) and instant messenger applications (IMAs) [15]. Furthermore, big data is created by leveraging the learning process from data sources like research articles, user experiences, and the exploration of large datasets. This process accelerates the application of machine learning and artificial intelligence in medical systems.
The collected data comes from a variety of sources, including wearables, MIDs, telemedicine, mHealth, health information systems (HISs), and other AI-powered medical technology [16, 17]. Moreover, patient healthcare data is included in electronic health records or EHRs. By connecting these health statistics with cutting-edge AI technologies, precise insights into patient treatment can be obtained. AI has been a popular option for large data applications in the medical field [18]. Furthermore, by employing analytical algorithms to enhance EHRs, big data analytics helps healthcare practitioners to provide better clinical services [19]. To improve data analysis, these analytics also take advantage of AI advancements to filter large data on several criteria [20]. The current perspective ends with a detailed review of the usage and drawbacks of AI in the pharmaceutical sector and its versatile application.
Limitations of artificial intelligence
Despite significant progress, there are still several gaps and limitations in AI's current application within the pharmaceutical field and pharmaceutical manufacturing viz. Data Quality and Availability: AI models heavily rely on high-quality, well-curated data. However, in pharmaceuticals, obtaining large and diverse datasets can be challenging due to data privacy concerns, regulatory restrictions, and the proprietary nature of pharmaceutical data. Improving data quality and accessibility remains a significant challenge; (ii) Interpretability and Explainability: Many AI models used in pharmaceuticals, such as deep learning models, are often considered "black boxes," making it challenging to interpret their decisions and explain their reasoning. Explainable AI (XAI) techniques are needed to enhance transparency and trust in AI-driven insights, especially in critical applications like drug discovery and clinical decision-making; (iii) Robustness and Generalization: AI models trained on specific datasets may lack robustness and generalization when applied to new data or real-world scenarios. Ensuring the reliability and generalizability of AI algorithms across diverse populations and healthcare settings is crucial for their widespread adoption in pharmaceuticals and healthcare; (iv) Regulatory and Ethical Challenges: The regulatory landscape for AI in healthcare and pharmaceuticals is still evolving. Ensuring compliance with regulatory requirements, such as FDA guidelines for AI-based medical devices and drug development, poses challenges for industry stakeholders. Additionally, ethical considerations, including bias, fairness, and patient privacy, need to be addressed in AI-driven applications in pharmaceuticals; (v) Integration with Traditional Methods: While AI shows promise in augmenting traditional drug discovery and development processes, there is a need to integrate AI-driven approaches with existing methodologies seamlessly. Collaborations between AI experts, pharmaceutical scientists, and clinicians are essential to leverage the strengths of both AI and traditional approaches effectively; (vi) Adoption and Acceptance: Despite the potential benefits, there may be resistance to adopting AI technologies in pharmaceuticals due to factors such as organizational inertia, lack of awareness or understanding, and concerns about job displacement. Overcoming these barriers and fostering a culture of innovation and collaboration is crucial for the widespread adoption of AI in the pharmaceutical industry; (vii) Unexplored Areas in pharmaceutical manufacturing: While AI has shown promise in optimizing certain aspects of pharmaceutical manufacturing, such as process optimization and quality control, there are still many unexplored areas. For example, leveraging AI for real-time monitoring and adaptive control of manufacturing processes, optimizing formulations and drug delivery systems, and enhancing predictive maintenance and fault detection in manufacturing equipment are areas that hold potential for further exploration and innovation [21].
Addressing these gaps and limitations will require concerted efforts from industry stakeholders, policymakers, regulators, and the research community to unlock the full potential of AI in pharmaceuticals and pharmaceutical manufacturing. Inefficiencies in drug testing and barriers in personalized medicine represent significant challenges that AI faces in the pharmaceutical domain:
Drug testing inefficiencies
Data Quality and Quantity: AI models require access to large, diverse, and high-quality datasets for effective training. However, in drug testing, obtaining sufficient and relevant data, especially for rare diseases or novel drug targets, can be challenging. Limited data availability can hinder the development and validation of AI models for drug testing [22, 23]. Prediction Accuracy: While AI algorithms can predict the efficacy and safety of drug candidates based on molecular structure and biological data, their accuracy may vary. Predicting complex biological interactions and outcomes accurately remains a challenge, leading to potential inaccuracies in drug testing results [24, 25]. Validation and Regulatory Compliance: Ensuring the reliability and regulatory compliance of AI-based drug testing methods is essential for their adoption in the pharmaceutical industry. Validating AI models against established standards and regulatory requirements, such as those set by the FDA, can be resource-intensive and time-consuming.
Barriers to personalized medicine
Data Integration and Privacy: Personalized medicine relies on the integration of diverse data sources, including genomic, clinical, and lifestyle data, to tailor treatments to individual patients. However, integrating and analysing such data while ensuring patient privacy and data security can be challenging. Regulatory requirements, such as HIPAA in the United States, add complexity to data sharing and analysis efforts [26]. Interpretability and Transparency: AI-driven personalized medicine models often lack interpretability and transparency, making it challenging for clinicians and patients to understand the rationale behind treatment recommendations. Enhancing the explainability of AI models in personalized medicine is crucial for building trust and acceptance among healthcare providers and patients [27]. Clinical Implementation and Adoption: Integrating AI-driven personalized medicine approaches into clinical practice poses challenges related to workflow integration, clinician training, and reimbursement. Healthcare systems may face barriers in adopting and scaling personalized medicine initiatives due to factors such as cost, infrastructure requirements, and regulatory constraints. Addressing these inefficiencies and barriers requires multidisciplinary collaboration among researchers, clinicians, regulators, and industry stakeholders. Efforts to improve data access and quality, enhance algorithmic transparency and interpretability, streamline regulatory processes, and promote education and awareness about AI in healthcare are essential for realizing the full potential of AI in drug testing and personalized medicine. Additionally, fostering a culture of innovation and collaboration within the pharmaceutical industry can facilitate the development and adoption of AI-driven approaches to improve patient care and outcomes [28].
Certainly, there are several gaps and limitations in AI's current application within the field of pharmaceutical manufacturing:
Process optimization complexity
Pharmaceutical manufacturing processes are highly complex and involve multiple stages, from drug formulation to packaging. While AI has shown promise in optimizing certain aspects of these processes, such as production scheduling or quality control, fully optimizing the entire manufacturing workflow remains a significant challenge due to the intricate interplay of variables and constraints [29].
Data quality and integration
AI models require access to vast amounts of high-quality data to make accurate predictions and recommendations. However, in pharmaceutical manufacturing, data sources are often heterogeneous, fragmented, and of varying quality. Integrating data from disparate sources, such as production sensors, laboratory instruments, and manufacturing execution systems (MES), while ensuring data integrity and security, presents a significant challenge [30].
Real-time monitoring and control
Real-time monitoring and control of pharmaceutical manufacturing processes are crucial for ensuring product quality, consistency, and compliance with regulatory standards. While AI-based systems can analyze large volumes of data in real-time and make predictive insights, implementing real-time decision-making capabilities in manufacturing environments poses technical and operational challenges.
Interpretability and trustworthiness
AI models used in pharmaceutical manufacturing, such as machine learning algorithms for process optimization or anomaly detection, often lack interpretability and transparency. Understanding the underlying factors driving AI-driven recommendations or decisions is essential for building trust among manufacturing stakeholders and regulatory authorities.
Regulatory compliance and validation
Pharmaceutical manufacturing is subject to stringent regulatory requirements, including Good Manufacturing Practice (GMP) regulations and guidelines from regulatory agencies such as the FDA and EMA. Validating AI-based systems for compliance with regulatory standards and ensuring their reliability and robustness in manufacturing environments require comprehensive testing and validation procedures.
AI-based technologies in medical product development offer transformative potential, but they also present significant regulatory and ethical challenges. These implications are as follows:
Regulatory Implications
Regulatory approval processes
Standards and Guidelines: Regulatory agencies like the FDA (U.S. Food and Drug Administration), EMA (European Medicines Agency), and MHRA (UK Medicines and Healthcare products Regulatory Agency) require rigorous evaluation of AI-based technologies. These evaluations include assessing the safety, efficacy, and reliability of AI models, often through clinical trials or real-world evidence.
Validation and verification
AI models must undergo thorough validation to ensure they work as intended across different populations and conditions. Verification ensures the model's outputs are accurate and reproducible.
Post-market surveillance
Once an AI-based product is approved, ongoing monitoring is essential. Regulators require systems for tracking performance and detecting issues after the product is on the market. This includes adverse event reporting and real-world data analysis.
Regulatory Frameworks
Adaptive Regulations: AI technologies evolve rapidly, requiring adaptive regulatory frameworks that can keep pace with technological advances. This includes updating guidelines to accommodate new types of AI models and applications.
Explainability and Transparency: Regulators emphasize the need for AI models to be interpretable, particularly in critical applications like medical product development. Transparency about how decisions are made by AI systems is crucial for gaining regulatory approval.
International Harmonization:
Global Collaboration: There is a need for international harmonization of regulations to facilitate the global use of AI in medical product development. Differing standards between countries can complicate the approval process and limit the widespread adoption of AI technologies.
Regulatory Sandboxes: Some countries are experimenting with "regulatory sandboxes" that allow companies to test AI technologies in a controlled environment before they are fully regulated. This can help in understanding how AI systems will perform in real-world settings.
Data Privacy and Security:
Compliance with Data Protection Laws: AI systems often require large datasets, which may include sensitive patient information. Compliance with data protection laws such as the GDPR (General Data Protection Regulation) in Europe and HIPAA (Health Insurance Portability and Accountability Act) in the U.S. is critical.
Security Standards: Ensuring that AI systems are secure from cyberattacks and data breaches is essential. Regulators may require evidence of robust cybersecurity measures before approving AI-based medical products.
Ethical Implications
Bias and Fairness
Data Bias: AI models are only as good as the data they are trained on. If the training data is biased or unrepresentative, the AI may produce biased or unfair outcomes. This can lead to disparities in healthcare, such as certain populations receiving less accurate diagnoses or treatment recommendations.
Fairness in Decision-Making: Ethical concerns arise when AI models disproportionately impact certain groups, particularly if the technology is used in critical areas like drug approval, patient selection for clinical trials, or personalized medicine.
Transparency and explainability
Black Box Models: Many AI models, especially deep learning models, operate as "black boxes," where the decision-making process is not easily interpretable. This lack of transparency can lead to mistrust among healthcare professionals and patients.
Informed Consent: Patients need to understand how AI is being used in their care, particularly when it comes to AI-driven medical products. Ensuring that patients provide informed consent is an ethical necessity.
Accountability and liability
Responsibility for Outcomes: Determining who is accountable when an AI system makes a mistake is complex. Is it the developer of the AI, the healthcare provider using it, or the regulatory body that approved it? Clear guidelines on liability are necessary.
Ethical AI Development: Developers have an ethical responsibility to ensure that AI systems are designed and tested rigorously to avoid harm. This includes ongoing monitoring and updating of AI models to prevent outdated or incorrect predictions.
Patient autonomy
Role of AI in Decision-Making: AI systems should support, not replace, human decision-making in healthcare. Maintaining patient autonomy means ensuring that AI serves as a tool for healthcare providers, rather than making decisions without human oversight.
Right to Opt-Out: Patients should have the right to opt-out of AI-driven care if they are uncomfortable with its use, provided that alternative options are available.
Social implications
Access and Equity: AI has the potential to improve healthcare access, but there is a risk that it could exacerbate existing inequalities if certain populations have less access to AI-driven medical products.
Impact on Employment: The increasing use of AI in healthcare and medical product development could lead to job displacement in certain areas, raising ethical concerns about the societal impact.
Ethical AI governance
Ethics Committees: Establishing ethics committees to oversee the development and deployment of AI in medical product development is crucial. These committees can ensure that ethical considerations are integrated into every stage of the AI lifecycle.
Continuous Ethical Review: As AI systems evolve, continuous ethical review is necessary to address new challenges and ensure that the technology remains aligned with societal values.
Adoption and integration challenges
Integrating AI-driven solutions into existing manufacturing workflows and systems can be challenging due to factors such as compatibility with legacy systems, organizational inertia, and cultural resistance to change. Overcoming these adoption barriers and fostering a culture of innovation and collaboration within pharmaceutical companies are essential for realizing the full potential of AI in manufacturing.
Supply chain optimization
While AI has been applied to optimize certain aspects of pharmaceutical supply chains, such as demand forecasting or inventory management, there are still unexplored opportunities for leveraging AI to enhance supply chain resilience, agility, and sustainability in the face of global disruptions, such as pandemics or natural disasters.
Addressing these gaps and limitations will require concerted efforts from pharmaceutical companies, technology providers, regulatory agencies, and other stakeholders to develop and deploy AI-driven solutions that effectively address the unique challenges of pharmaceutical manufacturing while ensuring product quality, safety, and compliance with regulatory standards [31].
Evaluating the effectiveness of AI in speeding up drug trials or enhancing drug safety
Trial design optimization
AI can help optimize clinical trial design by identifying suitable patient populations, selecting relevant endpoints, and refining inclusion/exclusion criteria. By leveraging patient data and predictive analytics, AI can improve trial efficiency and accelerate patient recruitment, leading to faster completion of trials.
Predictive modelling
AI algorithms can analyze diverse datasets, including preclinical data, electronic health records (EHRs), and genetic information, to predict drug safety profiles and identify potential adverse events. By detecting safety signals early in the drug development process, AI can help pharmaceutical companies mitigate risks and prioritize resources more effectively [32].
Drug repurposing and candidate selection
AI-driven approaches, such as virtual screening and drug repurposing algorithms, can expedite the identification of promising drug candidates for specific indications. By repurposing existing drugs or identifying novel uses for approved compounds, AI can shorten the drug discovery timeline and accelerate the development of new treatments [33].
Real-time safety monitoring
AI-based systems can monitor real-world data, such as adverse event reports, social media discussions, and patient forums, to detect safety concerns associated with marketed drugs. By analyzing large volumes of unstructured data in real-time, AI can provide early warning signs of potential safety issues, enabling timely intervention and risk management.
Drug-drug interactions and pharmacovigilance
AI algorithms can analyze drug interaction databases and electronic health records to identify potential drug-drug interactions and adverse drug reactions. By predicting drug interactions and assessing their clinical significance, AI can improve drug safety by guiding prescribing decisions and reducing the risk of adverse events.
Regulatory compliance and pharmacovigilance
AI-driven systems can automate pharmacovigilance processes, such as adverse event detection, reporting, and signal detection, to ensure compliance with regulatory requirements. By streamlining pharmacovigilance workflows and enhancing data analysis capabilities, AI can help regulatory agencies and pharmaceutical companies monitor drug safety more efficiently [34].
Decision support systems
AI-powered decision support systems can assist healthcare professionals in evaluating the safety and efficacy of drugs, incorporating patient-specific factors and real-time evidence into clinical decision-making. By providing personalized recommendations and treatment guidelines, AI can improve patient outcomes and enhance drug safety in clinical practice [35].
To evaluate the effectiveness of AI in speeding up drug trials or enhancing drug safety, stakeholders can assess metrics such as trial duration, patient recruitment rates, safety signal detection sensitivity, regulatory compliance, and clinical outcomes. Longitudinal studies and real-world evidence can provide insights into the impact of AI-driven interventions on patient outcomes, healthcare utilization, and overall drug safety and effectiveness. Additionally, collaboration between industry stakeholders, regulatory agencies, academic researchers, and technology providers is essential for validating AI-driven approaches and ensuring their widespread adoption in drug development and pharmacovigilance.
Methodology
The databases searched included PubMed, Scopus, and ScienceDirect using keywords 'AI in pharmaceuticals, AI in drug discovery, patient data analysis and production process' and 'machine learning in drug development,' from 2010 to 2024.
Diagnostic advancement through AI
The most noteworthy fact about AI in healthcare is how it can be used to program surgeon robots and collect and interpret significant data. This section covers the many methods and applications of AI in healthcare applications, along with disease indications, indicative challenges, and a methodology for simulating illness diagnosis with AI and additional learning models [36].
Basis for AI in modelling illness detection
A machine's capacity to examine how a human learns is known as artificial intelligence (AI), for example, by identifying patterns in a given scenario and identifying images. AI in healthcare transforms the way data is gathered, evaluated, and created for patient care [37]. The system's basic abstract design is called system planning. It contains the framework's opinions, its plan of action, and how it functions within precise guidelines. The customer can become aware of the constraints and limitations of the framework with the support of a firm understanding of its design. A visual representation of the illness recognition model applying practical machines and deep learning cataloging techniques is presented in Fig. 1. Real-world data must be maintained and prepared in advance of being handled by the computation [38]. Real-world data frequently contains errors about the metrics used, but these errors cannot be practiced due to the valid explanation. Information pre-processing then cycles the raw data, removes errors, and saves it for later examination. Information osmosis is the process of combining data from several sources. The data is then updated to reflect any blend errors, which are fixed through following techniques: -
Fig. 1.
The disease detection system's framework
Information alteration: Standardized data are used in this progression, depending on the calculation that is provided. Standardization of information can be implemented in several ways [39]. In the majority of information mining computations, this evolution is required since the data strives for maximum accuracy. After that, information is shared and developed. Information lessening: The strategy's progression focuses on reducing the information to more beneficial levels. Informational assortment and assessment: Informational index preparation and testing are two distinct phases of the information-gathering process. The preparation data is used to evaluate the real-world data examples [40]. A similar informational index, which is equivalent to the data needed for preparation and assessment, is usually used to copy experimental data. Testing the framework's accuracy is the next jiffy step after pre-handling the model.
Systematic model
To determine the possibility of a particular manifestation function given commitment parameters, analytical showing procedures are used, and they are highly effective in the expectation of disease. It often imagines what the person is going through based on their information indicators and previous judgments [41, 42].
Medical imaging used to diagnose ailments
"Clinical imaging" describes the arrangement of methods that provide internal body images. To obtain photographs of the body of an individual for therapeutic intents, such as finding, evaluating, or diagnosing a wound, fracture, or disease, certain procedures and cycles are used [43]. Computed tomography (CT) scan results are excellent illustrations of reminiscent imaging that can be used to accurately measure, mediate, and conclude the harms and dysfunctions that real advisers frequently deal with [44]. Additional reflection indicates that imaging modalities, like X-rays and MRI, are overutilized for difficult and intricate jobs (Table 1).
Table 1.
Types of medical imaging and their characterizations
| Medical Imaging Types | Characterizations |
|---|---|
| Radiographic imaging | Imaging utilizes ionising electromagnetic radiation, like X-beams, to see objects |
| Fluoroscopy | By continually providing X-beams at a lowered portion rate, it creates continuous images of the body's internal structures, resulting in moving projection radiographs of lower quality |
| Angiography | Angiography is used to detect aneurysms, releases, blockages, the development of new arteries, and the implantation of catheters and stents |
| DEXA | It is sometimes referred to as dual X-beam absorptiometry or bone densitometry when used in osteoporosis examinations |
| Computed tomography (CT) | With computed tomography imaging, a large amount of ionizing radiation associated with a PC is used to create images of both hard and soft tissues |
| Magnetic resonance imaging | A clinical test called magnetic resonance imaging (MRI) filtering uses radiofrequency radiation and a strong magnet to produce an image of the patient's body |
| Ultrasound imaging | It uses high repetition wideband sound waves in the megahertz range, which are reflected to variable degrees by tissue, to create three-dimensional images |
| Bone scan | It's an imaging process that uses a radioactive substance to identify the areas of the bone that are healing |
| Electron microscopy | Magnification of minute features can be achieved with electron microscopy, an instrument with a high settling power |
| Nuclear medicine | Using atomic characteristics, nuclear medicine encompasses the detection and management of illnesses |
| Magnetic resonance angiography scans | An appealing reverberation angiogram that provides incredibly detailed images of the body's veins is magnetic resonance angiography |
Disease symptoms and issues with assessment
The illness could be mild, severe, harsh, or ongoing. Lethal refers to a condition that has the potential to be fatal, while chronic and severe describe the stages of a disease. Furthermore, there may be warning signs for more serious and restorative illnesses or situations that are not related to one another. Here are a few illnesses along with their symptoms and warning signs for potential activities:
Symptoms of a heart attack include greater chest pain, dyspnea, heaving, cold sweats, discomfort, anxiety, crushing, or a broadening sensation in the middle of the chest that lasts for a prolonged period [45].
Facial listing, arm weakness, trouble speaking, quick smiles or equalization, sudden loss of feeling or weakness, blindness, disorientation, or acute pain are all indicators of a stroke [46].
Skin mole alterations, frequent flushing and redness of the face and neck, jaundice, chronic skin sores, newly acquired moles on the skin, and thick, red skin with brilliant patches are some of the indications of skin issues [47].
Reproductive health indicators include unusual vaginal discharge, intense pelvic/stomach pain, feeling full in the lower abdomen, tingling, mimicking, or discomfort around the genital area, blood loss or spotting between periods, pain or uneasiness during sexual activity, and real or sore feminine dying [48].
The adverse effects of lung problems include blood in the cough, dyspnea, breathing problems, persistent hacking, recurrent episodes of pneumonia or bronchitis, and puffing [49].
The indications of headache issues (apart from common strain headaches) include sudden attacks of migraine, "the most noticeably awful migraine of your life," and headaches associated with intense activity, queasiness, heaving, and immobility [50]. We have already covered the range of disease indicators and their symptoms. On the other hand, medication errors related to illness detection occur very frequently, can result in severe consequences, and are currently the basis for exceptional patient safety outcomes. Here, we have important concerns for different diagnostic approaches for identifying specific illnesses [51].
Analysis that is inadvertently delayed or, conversely, overlooked as a result of a conclusive delight of more astounding evidence.
Breakdowns in occasionally admitting to mind; elicitation or comprehension of side effects, images, and research facility results; detailing and weighing of difference investigation; and optimal development and strength arrangement or appraisal are among the flaws or malfunctions in the analytical course that are crucial to a missed finding or a conceded conclusion.
Different AI technologies used in drug discovery
The development of many pharmacological compounds is encouraged by the huge chemical space, which consists of over 1060 molecules [52]. Due to limitations in the medicine development process caused by a lack of advanced technology, AI helps address this issue by making the procedure more expensive and time-consuming [53]. AI can expedite therapeutic target validation, identify hit and lead compounds, and simplify drug structure design [54]. Figure 2 illustrates many uses of AI in drug discovery. Even with its benefits, AI still has to deal with a lot of data, including data size, growth, diversity, and uncertainty. Millions of compounds may be included in the data sets that pharmaceutical companies have available for drug development. Traditional machine-learning tools may not be able to handle this volume of data to tackle these problems. However, by utilizing recently developed artificial intelligence (AI) techniques, such as data mining (DL) and pertinent modelling studies to track the benefits of deep learning (DL) in the pharmaceutical industry's drug development process, big data modelling and analysis can be used to evaluate the safety and effectiveness of therapeutic molecules. In 2012, Merck sponsored a QSAR ML competition. DL models showed much greater predictivity than traditional ML approaches for fifteen drug candidate ADMET (absorption, distribution, metabolism, excretion, and toxicity) data sets [55, 56]. The vast virtual chemical space that depicts the locations and characteristics of molecules seems to be mapping out molecules. Chemical space visualisation aims to gather positional information about molecules inside the space to facilitate the search for bioactive substances. Virtual screening (VS) helps identify pertinent compounds for further analysis. Several chemical spaces, like ChemBank, PubChem, DrugBank, and ChemDB, are open to the public. In addition to structure- and ligand-based approaches, a range of in silico techniques for virtually screening compounds from virtual chemical spaces provide enhanced profile analysis, faster removal of nonlead compounds, and more economical therapeutic molecule selection [52]. Drug design tools, such as coulomb matrices and molecular fingerprint identification, consider the physical, chemical, and toxicological characteristics when selecting a lead ingredient [57].
Fig. 2.
Role of Artificial Intelligence (AI) in medicine development
Algorithms in drug discovery and development
Algorithms
Deep Learning (DL) Models
Convolutional Neural Networks (CNNs): Used for analysing molecular structures and predicting drug interactions.
Generative Adversarial Networks (GANs): Generate potential drug candidates by simulating molecular structures.
Reinforcement Learning (RL): Optimizes the drug discovery process by simulating various scenarios and learning from them.
Support Vector Machines (SVMs): Classify compounds based on their potential effectiveness or toxicity.
Quantitative Structure–Activity Relationship (QSAR) Models: Predict the chemical activity of compounds based on their structure.
Applications
Target Identification: AI models predict which proteins or genes a drug should target.
Lead Compound Identification: Algorithms screen large chemical libraries to find promising compounds.
Predicting Drug Toxicity: AI helps predict adverse effects of new drugs, reducing the likelihood of failure in clinical trials.
Limitations
Data Availability: The accuracy of predictions depends heavily on the quality and quantity of data.
Model Interpretability: Complex models like DL can be black boxes, making it difficult to understand how they arrived at a prediction.
Generalization: Models trained on specific datasets may not generalize well to new compounds or diseases.
Manufacturing and quality control
Algorithms
Machine Learning (ML) Models
Predictive Maintenance Models: Use sensor data to predict equipment failures before they happen, ensuring consistent production quality.
Anomaly Detection Algorithms: Identify defects in real-time by analyzing data from production lines.
Statistical Process Control (SPC): Applies ML to monitor and control the quality of manufacturing processes.
Computer Vision (CV):
CNNs: Used for visual inspection to detect defects in medical products like pills, syringes, or implants.
Optical Character Recognition (OCR): Ensures correct labeling and packaging.
Applications
Real-Time Monitoring: AI monitors production processes to detect deviations from the standard quality.
Automation of Quality Control: Reduces human error by automating visual inspections and other QC tasks.
Process Optimization: AI optimizes manufacturing parameters to enhance product consistency and reduce waste.
Limitations
Complexity of Processes: Variability in manufacturing processes can make it difficult for AI models to generalize across different production lines.
Data Integration: Combining data from different sources (e.g., sensors, cameras, QC tests) can be challenging and may affect model accuracy.
Scalability: Implementing AI at scale in manufacturing requires significant infrastructure and coordination.
Clinical trials
Algorithms
Natural Language Processing (NLP)
Text Mining: Extracts relevant information from medical literature to aid in trial design.
Sentiment Analysis: Analyzes patient feedback to assess the impact of the treatment.
Predictive Modeling: Survival Analysis Models: Predict patient outcomes based on trial data.
Patient Matching Algorithms: Ensure the selection of appropriate patients for trials by analyzing genetic, demographic, and health data.
Applications
Patient Recruitment: AI identifies suitable candidates for clinical trials, speeding up the recruitment process.
Trial Design: Helps in designing trials by analyzing previous trial data, reducing the chances of failure.
Outcome Prediction: AI models predict the success of clinical trials, helping to prioritize promising treatments.
Limitations
Bias in Data: If the training data is biased, the AI models may produce skewed or unfair results.
Regulatory Challenges: The use of AI in clinical trials is still a relatively new area, and regulatory guidelines are evolving.
Ethical Concerns: The use of AI in patient selection and outcome prediction raises ethical questions, particularly regarding fairness and transparency.
Post-Market Surveillance
Algorithms
NLP and Text Mining:
Adverse Event Detection: NLP algorithms scan reports and social media for mentions of adverse events.
Sentiment Analysis: Gauges public opinion about medical products, identifying potential issues early.
Anomaly Detection: ML models analyze usage patterns to detect any deviations that could indicate a quality issue.
Applications
Monitoring Product Performance: AI continuously monitors real-world product performance to detect any emerging issues.
Automated Reporting: Streamlines the reporting process for adverse events, ensuring timely updates.
Risk Management: AI helps in assessing and managing risks associated with medical products by analyzing vast amounts of post-market data.
Limitations
Data Privacy: Collecting and analyzing post-market data raises concerns about patient privacy.
Interpretation of Results: Identifying the root cause of an issue detected by AI can be challenging, as the model might flag non-issues.
Regulatory Compliance: Ensuring AI systems comply with regulatory standards in different regions can be complex.
General Limitations in AI for Medical Products
Regulatory Hurdles
AI-based systems must meet stringent regulatory standards before they can be used in medical product development and manufacturing.
Integration with Existing Systems
Integrating AI technologies into existing healthcare and manufacturing systems can be difficult and expensive.
Ethical and Legal Concerns:
The use of AI in healthcare raises questions about liability, informed consent, and the potential for unintended consequences.
Artificial intelligence for drug screening
It can require over a decade to find and develop a therapeutically optimal drug, with an average development cost of US$2.8 billion. But then again, nine out of ten drugs fail Phase II clinical trials and eventually receive final regulatory body approval [58, 59]. In addition to being employed for VS, these algorithms—which additionally predict in vivo activity and toxicity based on synthesis feasibility—include RF, SVMs, DNNs, RF-Nearest-Neighbour classifiers, and extreme learning machines [58, 60]. Alliance amid IT companies and biopharmaceutical companies, such as Bayer, Roche, and Pfizer, has resulted in providing the platform that streamlines the search for medicines in fields including immuno-oncology and heart issues [52].
Prediction of the physicochemical characteristics
The drug's physicochemical characteristics, which indirectly influence the drug's pharmacokinetics and target receptor family, must be considered when developing a new dosage form. These characteristics include the drug's solubility, surface area, dissolution, partition coefficient (logP), degree of ionization, and intrinsic permeability [61]. Numerous and significant AI-based techniques can aid in predicting physicochemical properties. For example, dosage form optimization generates large data sets for which machine learning (ML) software is well-trained [62]. Drug design methods use molecular descriptors such as electron density around the molecule, SMILES strings, potential energy measurements, and three-dimensional atomic coordinates to generate viable compounds through DNN and predict their properties. [63]. Since AI can forecast a dosage form's desirable bioactivity as well as its expected physicochemical qualities, it plays a crucial role in drug development.
Prediction of bioactivity
The affinity of pharmaceutical substances for the target protein or receptor determines how effective they are. Drug molecules cannot exhibit a pharmacological response or have a therapeutic impact if they do not interact with or have affinity for the targeted protein. Additionally, synthesized drug molecules may occasionally form detrimental connections with undesirable proteins or receptors. Consequently, drug-target binding affinity (DTBA) determines the predictive capacity of drug-target interactions. By considering the traits or similarities between a medicine and its target, AI-based methods can determine a drug's binding affinity. The drug and target's chemical moieties are found using feature-based interactions, and these are subsequently utilized to build the feature vectors. In contrast, a similarity-based interaction considers how similar the medicine and the target are. [64]. DTBA has been ascertained using a variety of ML and DL techniques, including KronRLS, SimBoost, DeepDTA, and PADME.
Prediction of toxicity
Anticipating the toxicity of any medicinal chemical is crucial to avert negative outcomes. The expense of drug discovery is increased since animal trials are often carried out following exploratory research employing cell-based in vitro assays to identify a compound's toxicity. There are numerous web-based programs available to assist with cost reduction, including Toxtree, pkCSM, admetSAR, and LimTox [63]. Cutting-edge AI-based methods use input feature-based toxicity projections or search for commonalities between substances. The Tox21 Data Challenge was established by the US Food and Drug Administration (FDA), the Environmental Protection Agency (EPA), and the National Institutes of Health in an attempt to evaluate various computer programs for estimating the toxicity of 12,707 medications and environmental chemicals [62]. A machine learning algorithm called DeepTox outperformed all other methods and was capable of correctly predicting the toxicity of a molecule based on 2,500 predefined toxicophoric properties, identifying both static and dynamic properties of molecules' chemical descriptors, such as molecular weight and Van der Waals number [65]. Table 2 summarizes the various AI technologies used in drug development.
Table 2.
Some AI technologies for the drug discovery process
| Tools | Details |
|---|---|
| DeepChem | MLP model that uses an AI system based on Python to identify a promising candidate for drug discovery |
| DeepTox | Computer programs that predict the toxicity of twelve thousand drugs |
| DeepNeuralNetQSAR | A Python-based system that helps detect a compound's molecular activity and is powered by computational tools |
| DeepBind | A computational instrument for examining protein-DNA/RNA binding |
| DeepPurpose | Drug-target interaction, drug-drug interaction, and protein function prediction using a deep learning library |
| ORGANIC | A molecular formation tool that facilitates the generation of molecules with desired properties |
| PotentialNet | Forecasts the binding affinity of ligands using NNs |
| PathDSP | A deep learning approach that uses cancer cell lines to predict medication sensitivity |
| Hit Dexter | Using machine learning to forecast which molecules may react to biochemical tests |
| DeltaVina | Using a scoring algorithm to determine the drug-ligand binding affinity |
| Neural graph fingerprint | Aids in predicting the characteristics of new compounds |
| AlphaFold | Predicts 3D structures of proteins |
| AtomNet | Predicting bioactivity with a deep convolutional neural network |
| Chemputer | Aids in the standard format reporting of chemical synthesis procedures |
| Chemical VAE | A framework based on auto-encoders to produce novel compounds |
| TorchDrug | A versatile framework for drug discovery models based on PyTorch |
| Graph level representation | Recognizing drug discovery graph depiction |
| DeepGraphMol | A computational technique that combines reinforcement learning and graph neural networks to create molecules with the desired features |
AI in drug designing
Prediction of the target protein structure
In order to effectively treat any disease state, the right target must be chosen while developing a therapeutic moiety. Some of the overexpressed proteins have a complex role in the development of the disease. For the purpose of creating a medicinal compound that specifically targets a disease, it is imperative to predict the structure of the target protein. Artificial intelligence (AI) can help with structure-based drug development by predicting the 3D protein structure. This is because the design is compatible with the target protein location's chemical environment, which helps anticipate a compound's impact on the target and helps with safety issues before synthesis or manufacturing [66]. In order to anticipate the 3D target protein structure, the distance between neighbouring amino acids and the corresponding angles of the peptide bonds were examined using the AI tool AlphaFold, which is based on DNNs. With 25 out of 43 structures properly predicted, AlphaFold demonstrated extraordinary performance. Reputable researcher Al Qurashi used recurrent geometric networks (RNNs), which take into account the three steps of computation, geometry, and assessment, to predict the structure of proteins. Subsequently, the basic sequence of the protein was encoded, and a partially completed backbone was retrieved from the geometric unit upstream, along with the torsional angles of a specific residue, to create a new backbone. The 3D structure was the end product of the previous element. The root mean square deviation (dRMSD) approach was utilized to evaluate the structural disparity between the expected and experimental structures. The RGN values were adjusted to maintain a minimal dRMSD between the anticipated and experimental structures [67]. AlphaFold was not going to be able to anticipate the protein structure as quickly as Al Qurashi's and his colleagues' AI technique. When it comes to predicting protein structures, AlphaFold is probably more accurate when the sequences are comparable to the reference structures [68]. The 2D structure of a protein was predicted by a study based on feed-forward supervised learning and backpropagation error approach, using MATLAB and a nonlinear three-layered neural network toolbox. Neural networks functioned as algorithm learners and performance assessors, and the input and output data sets were trained in MATLAB. 62.72% of the predictions of the 2D structure were realized [69].
Forecasting interactions between drugs and proteins
The study of drug-protein interactions is essential to determine the therapy's effectiveness. Once a drug's protein interaction is determined its efficacy and effectiveness can be predicted effortlessly. Moreover, poly-pharmacology is avoided, and medication repurposing is permitted [66]. Better therapeutic efficacy has been ensured via ligand–protein interactions that can be accurately predicted using a wide range of AI techniques [66, 70]. To uncover nine unique compounds and their interactions with four important targets, one researcher, Wang et al., created a model that operates on fifteen thousand protein–ligand interactions using core protein sequences and structural features of small molecules [71]. AI's capacity to foresee how medications will interact with targets has also been applied to avoid polypharmacology and facilitate the repurposing of approved medications. Phase II clinical trials may be initiated with a prescription drug that has been approved [52]. Relaunching an old medicine costs US$8.4 million, whereas launching a new medical entity costs US$41.3 million, therefore this also saves money [72]. The 'guilt by association' technique can be used to anticipate the novel relationship between a medicine and an illness, which can be either a network driven by computation or knowledge [73]. Among the techniques frequently used in the machine learning (ML) approach in a computationally driven network are SVM, NN, logistic regression, and deep learning. Logistic regression platforms such as “PREDICT”, “SPACE” and other machine learning algorithms take into consideration several aspects when repurposing medicines, such as the similarity between drugs and diseases, target molecule similarity, chemical composition and patterns of gene expression [74]. One well-known example is “Ligand Express”, a cloud-based proteome screening AI tool created by Cyclica. Receptors that can generate both on- and off-target interactions when interacting with a certain small molecule (whose chemical description is included in the SMILE string) are identified using this method. Understanding the medication and its potential negative effects is helpful [75].
AI in de novo drug design
The de novo drug design approach has been extensively utilized in recent years to create new medication compounds. Evolving DL approaches are replacing the conventional de novo medicinal design technique, which has drawbacks such as convoluted synthesis pathways and challenging bioactivity prediction [63]. Additionally, computer-aided synthesis planning, which can predict many synthesis paths for each molecule, can offer millions of synthetic structural options [76]. Popova et al. [77] developed a method called “Enforcement Learning for Structural Evolution” for de novo drug development that produces unique compounds using generative and predictive DNNs. Here, the predictive models anticipate the attributes of the created chemical, while the generative model, based on the stack memory, generates molecules more specific to the SMILE strings. Merk et al. [78] have synthesized retinoid X and PPAR agonist compounds with the intended therapeutic effects without the need for intricate regulations by using a generative AI model. The researchers' success in creating five compounds—four of which demonstrated notable modulatory effects in cell tests—highlights the value of generative AI in the creation of novel chemicals. The pharmaceutical business may benefit from using AI to create novel molecular designs because it can provide online learning, simultaneously optimize previously learned data, and suggest new synthetic routes for compounds, expediate designing and development [79, 80].
Role of AI in optimization of drug formulation
Connectivity of software and product optimization
Up until now, the main force behind AI acceleration and optimization has been hardware breakthroughs. However, recent developments from AI start-ups and academic organizations have resulted in new solutions that operate at the software level of the machine learning technology stack, cutting down on the amount of time that AI and ML models need to train, infer, use energy, and use memory [81]. These technologies are upending the AI sector, therefore knowing how to measure, apply, and leverage them is critical to business success.
Reduce AI hardware ownership costs by improving the performance of ML models on CPUs.
Optimizing AI accelerators that are already in use (GPUs, FPGAs, DSPs)
Supporting the complete implementation of additional data science and machine learning initiatives
Increasing the competitiveness of edge devices and IoT goods
At the Software for AI Optimization Summit, the newest software optimization technologies for AI/ML processes will be shown. This implies that innovators and early adopters of data science and machine learning may be able to train and infer AI and ML models more quickly, cheaply, and with high levels of performance and accuracy [82].
Software related to product optimization for AI devices
Product optimization is a hurdle that many firms must overcome. Any activity, procedure, or strategy that seeks to create something as good, useful, or efficient as possible—a design, system, or decision—is what we refer to as optimization. The process of making decisions that maximize quality, minimize energy use, and minimize expenses is referred to as optimization. Currently, the industry is placing a lot of attention on analytics and digitization [82]. Massive volumes of data generated everyday by thousands of sensors, even inside a single production site, enable this laser-like focus. Before, the utilization of data from sensors and other systems was restricted due to a lack of knowledge, the absence of suitable technologies, and a lack of data pipelines for collecting data for additional analysis. With the AI platform, creating an app from the ground up is simple. You can use the built-in formula for this. Because of the drag-and-drop interface, it's simple to use. Chatbots are computer programs designed to simulate human behaviour, and their use is growing. Deep learning software can also be advantageous for speech and image recognition [83]. A computer that has access to data can learn from it by using machine learning software. In other words, artificial intelligence refers to the science and technology that enables a system, program, or computer to perform the creative and intellectual duties of a human, independently resolving issues and reaching judgments [83, 84]. By enabling users to discover new information, artificial intelligence software systems are intended to assist people in gradually increasing their performance and productivity. Artificial intelligence systems that offer an analytical report to enhance planning, reasoning, thinking, problem-solving, and learning include machine learning and deep learning [85].
Applications with AI-based advancements
Drug-delivery nanorobots powered by AI
Integrated circuits, sensors, power sources, and safe data backups are crucial parts of nanorobots, and computational technologies such as artificial intelligence maintain them current [86, 87]. When creating implantable nanorobots for the regulated delivery of medications and genes, factors including dose modification, sustained release, and control release need to be taken into account. Furthermore, the medications' release needs to be automated with the help of AI tools like integrators, NNs, and fuzzy logic [88]. Implants with microchips can be utilized for both programmed release and body position detection (Fig. 3).
Fig. 3.
AI-based nanorobots for drug delivery
AI in conjunction with medication administration and prediction of synergism and antagonist
Many drug combinations are authorized and sold to treat complicated illnesses, including cancer and tuberculosis since they can work in concert to promote a quick recovery [89, 90]. For example, cancer rehabilitation requires a combination of six or seven therapies. Selecting specific and promising medications for combination therapy is a laborious procedure. It necessitates high throughput drug screening of a wide range of drugs. Logistic regression with artificial neural networks and network-based modelling can be used to test drug combinations, which can improve the overall dosing schedule [89, 91]. Evidence about the antagonistic or synergistic effects of drugs taken together can provide credence to the notion that combination drug delivery may yield greater efficacy. The core regulatory inference algorithm used "material regulatory genes and quot; predicts 56% synergistic accuracy. Moreover, network-based Laplacian regularized least squares and radio frequency (RF) synergistic medicine combination can be applied for the same objective [91]. Li et al. [92] used RF to create a model for medication combinations that work well together in order to predict synergistic anticancer treatment combinations. Using a range of networks and gene expression patterns, the model's developers were able to precisely forecast 28 combinations of synergistic anticancer drugs. Only three of the combinations were recognized by the researchers, though other combinations may prove to be significant. Likewise, Mason et al. [93] use a machine learning technique called Combination Synergy Estimation on a dataset of 1540 antimalarial drugs to forecast possible synergistic antimalarial combos.
The advent of AI in nanomedicine
In order to identify, alleviate, and keep track of complicated illnesses like HIV, cancer, malaria, asthma, and other inflammatory diseases, nanomedicine integrates medicine and nanotechnology. Recent breakthroughs in medicinal and diagnostic domains have increased the use of drug administration using nanoparticles because of their heightened therapeutic qualities and efficacy [90, 94]. The combination of artificial intelligence and nanotechnology can provide solutions to several formulation development glitches [95]. The multipurpose mesoporous silica nanoparticles known as silicasomes—which are laden with irinotecan and a peptide known to penetrate tumours called iRGD—were made possible with the assistance of AI. iRGD tripled or quadrupled the absorption of silicasomes because it improved treatment results and overall survival by promoting silicasome transcytosis [94] (Fig. 4).
Fig. 4.
AI in nanotechnology
Applications of AI in healthcare
Artificial intelligence for medicine discovery
There are a few medicines that best biopharmaceutical companies have found. Pfizer is utilizing IBM Watson, a framework that's based on machine learning, to help within the revelation of immuno-oncology solutions [96]. Whereas Roche backup Genentech is depending on an fake insights system from GNS Healthcare in Cambridge, Massachusetts to assist with its look for cancer drugs, Sanofi has chosen to utilize Exscientia's counterfeit insights (AI) stage to seek for pharmaceuticals for metabolic illnesses. Comparable organizations or inner activities are display in nearly all critical biopharmaceutical companies [96].
Application of artificial intelligence in clinical research
The best way to validate the suitability and safety of new medicines is through regular "straight" clinical research. Created fundamentally for the assessment of mass-market pharmaceuticals, the long-standing, well-tried convention comprising discrete and well-defined stages of randomized controlled trials (RCTs) has for the most part not changed within the final few decades. Fake insights can make strides in efficiency and clinical improvement results, as well as diminish the length of clinical trial cycles. The impact of AI on the biopharma esteem chain is secured in three reports, the third of which is this one [97].
Medical consideration
Counterfeit insights have an effect on quiet results within the restorative field. Therapeutic AI companies create a framework that underpins understanding on all fronts. Clinical insights too provides insights to patients to help them progress their quality of life by analyzing their restorative information (Table 3).
Table 3.
Some major firms around the world via. artificial intelligence in therapeutic disciplines
| S.no | Company | Purpose |
|---|---|---|
| 1 | AiCure (New York City) Patient-oriented | Makes use of behavioral data, audio, and video to enhance comprehension of the relationship between patients, illness, and therapy |
| 2 | Aidence (Amsterdam, The Netherlands) Clinician-oriented | AI for radiologists: enhancing lung cancer diagnosis and therapy |
| 3 | Aiva Health (Los Angeles) Administrative and Operational-oriented | The first care assistant powered by voice: matches patients with the appropriate doctor for communication |
| 4 | Babylon Health (London) Administrative and Operational-oriented | Uses AI and NLP to build a universally accessible, reasonably priced health system |
| 5 | Bot MD (Singapore) Clinician-oriented | Bot assistant: responds to clinical inquiries, automatically arranges files and photos, and transcribes spoken case notes |
| 6 | Suki (San Francisco) Clinician-oriented | Voice-activated medical digital assistant |
| 7 | Insitro (San Francisco) Patient-oriented | Uses computational genomics and powerful machine learning to cut down on the time and expense patients must spend finding new drugs |
Maternal healthcare
The utilising artificial intelligence (AI) and electronic health data to predict if expecting moms are much more likely to experience delivery complications. Utilising digital technology to provide more pregnant patients with access to high-acuity, or more frequent and complex care, as well as regular care. High-risk obstetric ladies who allow birth at low-acuity clinics run a more noteworthy chance of encountering extreme maternal dreariness as compared to giving birth in higher-acuity clinics with more grounded assets and proficient encounter.
Manifestation of AI in Indian healthcare
A wide range of stakeholders, such as Google and Microsoft, have also joined forces to work on several projects aimed at developing AI infrastructure throughout India [98]. The policy landscape in India is driving the application of AI in healthcare delivery. The year, 2018 saw the approval of "NITI Aayog," the policy think tank of the Indian government, to create a national plan for artificial intelligence and other emerging technologies (14). Healthcare is one of the five industries that NITI Aayog concentrated on because it is thought to gain from AI the most in meeting societal requirements (14). In this endeavour, NITI Aayog has embraced the motto "AI for all" (#AI for All). Along with ensuring proper data privacy and security, NITI Aayog also aims to strike a balance between the need for innovation and ethical issues [99].
AI applications in Indian medical therapy
Based on mapping body warm inserted with an AI innovation, Niramai Wellbeing Analytics (Bangalore) made a non-invasive, low-cost strategy to test for early breast cancer in 2016 [100]. The term "Niramai" stands for Non-invasive risk assessment utilising artificial intelligence and machine learning. It uses "Thermalytix" technology, which enables tumour identification five years ahead of clinical exams or mammography. With several technologies, such as Daksh, a paperless labour monitoring device, Janitri technologies focuses on providing healthcare during pregnancy. Keytar is a non-invasive cardiotocography (CTG) device that tracks a pregnant woman's uterine contractions and can monitor a baby's heart rate while it is still inside her. In isolated, rural locations, this gadget is lightweight, non-intrusive, and simple to operate [101]. In the Indian context, utilizing chatbots to gather family history can be quite beneficial because patients typically travel long distances to cities or towns just to sit and answer questions. It can save a tonne of time to do this. In genetics, family history is very significant. Voice bots in India have the potential to overcome several literacy issues that chatbots encounter. One of the challenges in developing chatbots is figuring out how to convert the majority of them into regional languages that the average person can understand. Additionally, chatbots are limited in areas where literacy rates are low [102].
Limitations and challenges within the application of AI framework in therapeutic science
Artificial intelligence systems have constraints and obstacles similar to those faced by any other sector, including healthcare. The moment has arrived for us to shift from a reactive to a proactive approach when it comes to the drawbacks of new technologies. Here, we talk about those difficulties with a specific emphasis on those that are related to healthcare.
Accessibility of statistics
To create an artificially intelligent system, data collection must occur first, then problem identification and solution strategy development. In order to create models that function well, a lot of high-quality data must be accessible. Because of concerns about patient privacy and recent instances of data breaches at large organizations, data collection is a sensitive topic. The capacity to analyze and dissect tremendous sums of information as well as store it has expanded due to mechanical breakthroughs. Recognizable proof of a specific person from a swarm is made conceivable by innovations such as facial acknowledgment and hereditary investigation. Patients and the common open have the proper to privacy and the caution to choose how much of their individual data they need to share. These days, protections firms can obtain patient data through information breaches, which might result within the refusal of a patient's restorative protections claim since the guarantors considers the patient is more expensive due to their hereditary composition. Due to quiet protection, there's less information accessible, which confines show preparing and keeps a show from performing at its best.
Creating biased models
Prejudiced data
Subsequently, a certain amount of the composed data, known as the training data set, is utilized to build AI systems; the remaining data is placed aside for testing, and is also known as the testing data set. Therefore, the final model will likewise be biassed if the data collection is biassed, i.e., if it specifically targets a particular race, gender, or age group. As such, the information gathered needs to accurately reflect the intended audience.
Data preprocessing
Even after collecting objective data, a biased model can still be created. Prior to the acquired data being utilized to train the algorithm, pre-processing is required. Human information entry errors and other various reasons are often the cause of errors in the raw data that was gathered. Mathematical reasoning is sometimes used to remove or modify these entries. Data preparation must be done carefully to ensure that the resulting data collection is not biassed.
Prototypical assortment
Out of all the models and algorithms accessible, the approach that best suits the task at hand must be selected. As such, the process of selecting the model is essential. Bias models are overly straightforward models that fall short of accurately capturing the trends in the dataset.
Bestowing prejudiced models
The client of a misleadingly shrewd framework ought to get these models on a very basic level. This gives the user more control over how to use and comprehend the model's output. For example, it is possible to assess a model's performance using a variety of metrics, including F1 score, AUC score, accuracy, precision, and recall [103]. But the statistic needed for each challenge varies. When showing performance metrics of a model to an artificially intelligent system user, they need to ensure that the metrics are relevant to the current problem and not only the ones with the highest scores.
Patchy data
Another barrier to the widespread application of AI is the incapacity of models- such as regression, classification, clustering, natural language processing, etc. that an organization invests time and money in creating and implementing for a specific task to be quickly and readily transferred to another organization without requiring recalibration. Data communication across healthcare organizations is sometimes prohibited or unavailable due to privacy concerns. This results in fragmented data, which lowers the reliability of a model.
Black boxes
Artificial intelligence is now and then alluded to as "dark boxes" due to the complexity of the numerical methods included. Upgrading the models’ interpretability and availability is fundamental. There's much work to be tried in this region, despite a few later endeavours [104].
Imaginable impression of artificial intelligence on remedial care in the future
Though the initial attempts in this field have been challenging, we believe that AI will soon be able to provide accurate diagnoses and recommendations for treatments. Fast progress in AI image analysis indicates that most radiology and cancer images will eventually be analysed by computers. In medicine, speech and text recognition have been used for tasks including note-taking and patient communication. The main issue with artificial intelligence (AI) in several healthcare domains is not so much if the technologies will function or not, but rather, how to apply them to routine professional practice. For wider application, artificial intelligence (AI) systems need approval from regulatory bodies, integration using electronic health record (EHR) systems, enough standardization so that similar products work similarly, education of clinicians, financing from either private or public payer organizations, and ongoing updates in the field. There are solutions to these challenges, but they will take much more time to implement than it did to create the technology. As a result, we predict that AI will have some impact on clinical practice within the next 5 years, and more widely within the next 10 years.
Conclusion
AI and its cutting-edge capabilities are constantly evolving to overcome the obstacles pharmaceutical companies face. The obstacles have an impact on both the creation of novel drugs and the product's total shelf life. This could explain the increase in start-ups in the business. A major obstacle confronting the healthcare sector at the moment is the rising expense of medications and therapies. As a result, society must drastically change in this regard. AI-enabled pharmaceutical product manufacturing enables the adjustment of dosage, release parameters, and other essential elements to match the unique requirements of every patient. By leveraging the most recent AI-based technology, It will be feasible to reduce the amount of time it takes for products to go on sale. while simultaneously improving product quality, cost-effectiveness, and safety throughout the production progression. These advantages show how important automation is. Artificial Intelligence (AI) holds great promise for optimizing and subsequently integrating the created drug in the proper dose form. Although artificial intelligence (AI) is not yet commercially available and there are still certain challenges to be solved before this technology can be extensively employed, it is quite probable that the pharmaceutical industry will find AI to be a very helpful tool very soon.
Funding
No funding was received for conducting this study.
Data Availability
Not applicable.
Declarations
Financial and non-financial interests
The authors have no relevant financial or non-financial interests to disclose.
Competing interest
There is no completing interest to declare.
Footnotes
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Contributor Information
Nidhi Srivastava, Email: nidhimpharm@gmail.com.
Ajar Nath Yadav, Email: ajarbiotech@gmail.com, Email: ajar@eternaluniversity.edu.in.
References
- 1.Varela FJ, Thompson E, Rosch E. The embodied mind: Cognitive science and human experience. MIT Press, Cambridge; 2014. [Google Scholar]
- 2.Shabbir J, Anwer T. Artificial intelligence and its role in near future. J Latex Class Files. 2015;14:1–11. [Google Scholar]
- 3.Lavecchia A, Giovanni CD. Virtual screening strategies in drug discovery: a critical review. Curr Med Chem. 2013;20:2839–60. [DOI] [PubMed] [Google Scholar]
- 4.Ramesh A, et al. Artificial intelligence in medicine. Ann R Coll Surg Engl. 2004;86:334–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Miles J, Walker A. The potential application of artificial intelligence in transport. IEE Proc Intell Transport Syst. 2006;153:183–98. [Google Scholar]
- 6.Musleh M, Alajrami E, Khalil A, Nasser B, Barhoom A, Naser S. Predicting liver patients using artificial neural network. J Acad Inf Syst Res. 2019;3:1–11. [Google Scholar]
- 7.Dabowsa N, Amaitik N, Maatuk A, Shadi A. A hybrid intelligent system for skin disease diagnosis. In: Conference on engineering and technology. 2017; pp 1–6. 10.1109/ICEngTechnol.2017.8308157.
- 8.Bhatt V, Pal V. An intelligent system for diagnosing thyroid disease in pregnant ladies through artificial neural network. In: Conference on advances in engineering science management and technology. 2019; pp 1–10. 10.2139/ssrn.3382654
- 9.Plawiak P, Ozal Y, Tan R, Acharya U. Arrhythmia detection using deep convolution neural network with long duration ECG signals. Comput Biol Med. 2018;102:411–20. 10.1016/j.combiomed.2018.09.009. [DOI] [PubMed] [Google Scholar]
- 10.Yang Y, Siau K. A Qualitative Research on Marketing and Sales in the Artificial Intelligence Age. MWAIS 2018 Proceedings. Presented at the Midwest Association for Information Systems Conference, Saint Louis, Missouri. 2018.
- 11.Wirtz BW, Weyerer JC, Geyer C. Artificial intelligence and the public sector—applications and challenges. Int J Public Adm. 2019;42(7):596–615. [Google Scholar]
- 12.Paul D, Sanap G, Shenoy S, Kalyane D, Kalia K, Tekade RK. Artificial intelligence in drug discovery and development. Drug Discov Today. 2021;26(1):80–93. 10.1016/j.drudis.2020.10.010. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Miles JC, Walker AJ. The potential application of artificial intelligence in transport. IEE Proc Intell Trans Syst. 2006;153:183–98. [Google Scholar]
- 14.Ng R, Tan KB. Implementing an individual-centric discharge process across Singapore Public Hospitals. Int J Environ Res Public Health. 2021;18(16):8700. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.World Health Organization. Global Strategy on Digital Health 2020–2025. 2021; pp 7–13.
- 16.Briganti G, Le Moine O. Artificial Intelligence in Medicine: Today and Tomorrow. Front Med. 2020;7:27. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Hu J, Perer A, Wang F. Data-driven analytics for personalized healthcare. In: Weaver CA, Ball MJ, Kim GR, Kiel JM, editors. Healthcare Information Management Systems. Cham: Springer; 2016. p. 529–54. [Google Scholar]
- 18.Dash S, Shakyawar SK, Sharma M, Kaushik S. Big data in healthcare: Management, analysis and future prospects. J Big Data. 2019;6:54. [Google Scholar]
- 19.Zhang C, Ma R, Sun S, Li Y, Wang Y, Yan Z. Optimizing the electronic health records through big data analytics: a knowledge-based view. IEEE Access. 2019;7:136223–31. [Google Scholar]
- 20.Rawat S. How Is Big Data Analytics Using AI? 2021. Available online: https://www.analyticssteps.com/blogs/how-big-dataanalytics-using-ai. Accessed on 11 Jan 2023.
- 21.Jiménez-Luna J, Grisoni F, Weskamp N, Schneider G. Artificial intelligence in drug discovery: recent advances and future perspectives. Expert Opin on Drug Discov. 2021;16(9):949–59. [DOI] [PubMed] [Google Scholar]
- 22.Wassermann AM, Lounkine E, Hoepfner D, et al. Dark chemical matter as a promising starting point for drug lead discovery. Nat Chem Biol. 2015;11:958–66. [DOI] [PubMed] [Google Scholar]
- 23.Engels MFM, Gibbs AC, Jaeger EP. A cluster-based strategy for assessing the overlap between large chemical libraries and its application to a recent acquisition. J Chem Inf Model. 2006;46(6):2651–60. [DOI] [PubMed] [Google Scholar]
- 24.Bradshaw J, Kusner MJ, Paige B, Marwin HS, Segler S, Hernandez-Lobato JM. A generative model for electron paths. In: Proc. 7th International Conference on Learning Representations, ICLR 2019, New Orleans, LA, USA. April 26–30, 2019;6–9. https://openreview.net/forum?idr¯1x4BnCqKX.
- 25.Do K, Tran T, Venkatesh S. Graph transformation policy network for chemical reaction prediction. In Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining Association for Computing Machinery, Anchorage, AK, USA. 2019;750–760.
- 26.Wallace SE, Gaye A, Shoush O, Burton PR. Protecting personal data in epidemiological research: DataShield and UK law. Public Health Genom. 2014;17:149–57. [DOI] [PubMed] [Google Scholar]
- 27.Laurie G, Sethi N. Towards principles-based approaches to governance of health-related research using personal data. Eur J Risk Regul. 2013;4:43–57. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Agaku IT, Adisa AO, Ayo-Yusuf OA, Connolly GN. Concern about security and privacy, and perceived control over collection and use of health information are related to withholding of health information from healthcare providers. J Am Med Inform Assoc. 2014;21:374–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Bruno A, Costantino G, Sartori L, et al. The in silico drug discovery toolbox: applications in lead discovery and optimization. Curr Med Chem. 2019;26:3838–73. [DOI] [PubMed] [Google Scholar]
- 30.Subramanian I, Verma S, Kumar S, Jere A, Anamika K. Multi-omics Data Integration, Interpretation, and Its Application. Bioinform Biol Insights. 2020;14:1177932219899051. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Selvaraj C, Chandra I, Singh SK. Artificial intelligence and machine learning approaches for drug design: challenges and opportunities for the pharmaceutical industries. Mol Divers. 2022;26:1893–913. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Gawehn E, Hiss JA, Schneider G. Deep learning in drug discovery. Mol Inform. 2016;35(1):3–14. [DOI] [PubMed] [Google Scholar]
- 33.Bender A, Cortés-Ciriano I. Artificial intelligence in drug discovery: what is realistic, what are illusions? Part 1: ways to make an impact, and why we are not there yet. Drug Discov Today. 2020;26(2):511–24. [DOI] [PubMed] [Google Scholar]
- 34.Chu Y, Kaushik AM, Wang X, et al. DTI-CDF: a cascade deep forest model towards the prediction of drug-target interactions based on hybrid features. Brief Bioinform. 2021;22(1):451–62. [DOI] [PubMed] [Google Scholar]
- 35.Jiménez-Luna J, Grisoni F, Schneider G. Drug discovery with explainable artificial intelligence. Nat Mach Intell. 2020;2(10):573–84. [Google Scholar]
- 36.Kumar Y, Singla R. Federated learning systems for healthcare: perspective and recent progress. In: Rehman MH, Gaber MM (eds). Studies in computational intelligence, vol965. Springer, Cham, 2021. 10.1007/978-3-030-70604-3_6
- 37.Ali M, Tengnah J, Sooklall R. A predictive model for hypertension diagnosis using machine learning techniques. In: Jude HD, Balas VE, editors. Telemedicine Technologies. Elsevier; 2019. p. 139–52. 10.1016/B978-0-12-816948-3.00009-X. [Google Scholar]
- 38.Jo T, Nho K, Saykin AJ. Deep learning in Alzheimer’s disease: diagnostic classification and prognostic prediction using neuroimaging data. Front Aging Neurosci. 2019;11. 10.3389/fnagi.2019.00220. [DOI] [PMC free article] [PubMed]
- 39.Naser S, Naseer I. Lung cancer detection using artificial neural network. J Eng Inf Syst. 2019;3:17–23. [Google Scholar]
- 40.Sarao V, Veritti D, Paolo L. Automated diabetic retinopathy detection with two different retinal imaging devices using artificial intelligence. Graefe’s Arch Clin Exp Ophthalmol. 2020;258:2647–54. 10.1007/s00417-020-04853-y. [DOI] [PubMed] [Google Scholar]
- 41.Keenan T, Clemons T, Domalpally A, Elman M, Havilio M, Agron E, Chew E, Benyamini G. Intelligence detection versus artificial intelligence detection of retinal fluid from OCT: age-related eye disease study 2: 10 year follow on study. Ophthalmology. 2020;128:100–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Rajalakshmi R, Subashini R, Anjana R, Mohan V. Automated diabetic retinopathy detection in smartphone-based fundus photography using artificial intelligence. Eye. 2018;32:1138–44. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Bibault J, Xing L. Screening for chronic obstructive pulmonary disease with artificial intelligence. Lancet Digit Health. 2020;2:e216–7. [DOI] [PubMed] [Google Scholar]
- 44.Chen Y, Sha M, Zhao X, Ma J, Ni H, Gao W, Ming D. Automated detection of pathologic white matter alterations in Alzheimer’s disease using combined diffusivity and kurtosis method. Psychiatry Res Neuroimaging. 2017;264:35–45. [DOI] [PubMed] [Google Scholar]
- 45.Aggarwal Y, Das J, Mazumder PM, Kumar R, Sinha RK. Heart rate variability features from nonlinear cardiac dynamics in the identification of diabetes using artificial neural network and support vector machine. Biocybern Biomed Eng. 2020;40(3):1002–9. [Google Scholar]
- 46.Lukwanto R, Irwansyah E. The early detection of diabetes mellitus using a fuzzy hierarchical model. Proc Comput Sci. 2015;59:312–9. [Google Scholar]
- 47.Rodrigues DA, Ivo RF, Satapathy SC, Wang S, Hemanth J, Filho PPR. A new approach for classification skin lesion based on transfer learning, deep learning, and IoT system. Pattern Recognit Lett. 2020;136:8–15. [Google Scholar]
- 48.Kather J, Pearson A, Halama N, Krause J, Boor P. Deep learning microsatellite instability directly from histology in gastrointestinal cancer. Nat Med. 2019;25:1054–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49.Ma F, Sun T, Liu L, Jing H. Detection and diagnosis of chronic kidney disease using deep learning-based heterogeneous modified artificial neural network. Future Gener Comput Syst. 2020;111:17–26. [Google Scholar]
- 50.Mueller FB. AI (Artificial Intelligence) and hypertension research. Curr Hypertens Rep. 2020;22:70. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 51.Chuang C. Case-based reasoning support for liver disease diagnosis. Artif Intell. 2011;53:15–23. [DOI] [PubMed] [Google Scholar]
- 52.Mak KK, Pichika MR. Artificial intelligence in drug development: present status and prospects. Drug Discov Today. 2019;24:773–80. [DOI] [PubMed] [Google Scholar]
- 53.Vyas M, Thakur S, Riyaz B. Artificial intelligence: the beginning of a new era in the pharmacy profession. Asian J Pharm. 2018;12:72–6. [Google Scholar]
- 54.Sellwood MA, Ahmed M, Segler MSH, Brown N. Artificial intelligence in drug discovery. Fut Sci. 2018;10:2025–8. [DOI] [PubMed] [Google Scholar]
- 55.Zhu H. Big data and artificial intelligence modeling for drug discovery. Annu Rev Pharmacol Toxicol. 2020;60:573–89. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 56.Ciallella HL, Zhu H. Advancing computational toxicology in the big data era by artificial intelligence: data-driven and mechanism-driven modeling for chemical toxicity. Chem Res Toxicol. 2019;32:536–47. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 57.Chan HS, Shan H, Dahoun T, Vogel H, Yuan S. Advancing drug discovery via artificial intelligence. Trends Pharmacol Sci. 2019;40(8):592–604. [DOI] [PubMed] [Google Scholar]
- 58.Alvarez-Machancoses Ó, Fernández-Martínez JL. Using artificial intelligence methods to speed up drug discovery. Expert Opin Drug Discov. 2019;14:769–77. [DOI] [PubMed] [Google Scholar]
- 59.Fleming N. How artificial intelligence is changing drug discovery. Nature. 2018;557:S55–S55. [DOI] [PubMed] [Google Scholar]
- 60.Dana D, Gadhiya SV, St. Surin LG, Li D, Naaz F, Ali Q, Paka L, Yamin MA, Narayan M, Goldberg ID. Deep learning in drug discovery and medicine; scratching the surface. Molecules. 2018;23:2384. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 61.Zang Q, Mansouri K, Williams AJ, Judson RS, Allen DG, Casey WM, Kleinstreuer NC. In silico prediction of physicochemical properties of environmental chemicals using molecular fingerprints and machine learning. Chem Inf Model. 2017;57:36–49. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 62.Yang X, Wang Y, Byrne R, Schneider G, Yang S. Concepts of artificial intelligence for computer-assisted drug discovery. Chem Rev. 2019;119:10520–94. [DOI] [PubMed] [Google Scholar]
- 63.Hessler G, Baringhaus K-H. Artificial intelligence in drug design. Molecules. 2018;23:2520. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 64.Ozturk H, Özgür A, Ozkirimli E. DeepDTA: deep drug–target binding affinity prediction. Bioinformatics. 2018;34:i821–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 65.Mayr A, Klambauer G, Unterthiner T, Hochreiter S. DeepTox: toxicity prediction using deep learning. Front Environ Sci. 2016;3:80. [Google Scholar]
- 66.Wan F, Zeng J. Deep learning with feature embedding for compound– protein interaction prediction. BioRxiv. 2016;2016:086033. [Google Scholar]
- 67.AlQuraishi M. End-to-end differentiable learning of protein structure. Cell Syst. 2019;8:292–301. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 68.Hutson M. AI protein-folding algorithms solve structures faster than ever. Nature News. 2019. 10.1038/d41586-019-01357-6. [DOI] [PubMed]
- 69.Avdagic Z, Purisevic E, Omanovic S, Coralic Z. Artificial intelligence in prediction of secondary protein structure using CB513 database. Summit Transl Bioinf. 2009;2009:1–5. [PMC free article] [PubMed] [Google Scholar]
- 70.Tian K, Shao M, Wang Y, Guan J, Zhou S. Boosting compound-protein interaction prediction by deep learning. Methods. 2016;110:64–72. [DOI] [PubMed] [Google Scholar]
- 71.Wang F, Liu D, Wang H, Luo C, Zheng M, Liu H, Zhu W, Luo X, Zhang J, Jiang H. Computational screening for active compounds targeting protein sequences: methodology and experimental validation. J Chem Inf Model. 2011;51:2821–8. [DOI] [PubMed] [Google Scholar]
- 72.Persidis A. The benefits of drug repositioning. Drug Discov World. 2011;12:9–12. [Google Scholar]
- 73.Koromina M, Pandi MT, Patrinos GP. Rethinking drug repositioning and development with artificial intelligence, machine learning, and omics. OMICS. 2019;23:539–48. [DOI] [PubMed] [Google Scholar]
- 74.Park K. A review of computational drug repurposing. Transl Clin Pharmacol. 2019;27:59–63. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 75.Li Z, Li X, Liu X, Fu Z, Xiong Z, Wu X, Zheng M et al. KinomeX: a web application for predicting kinome-wide polypharmacology efect of small molecules. Bioinformatics. 2019;35:5354–5356 [DOI] [PubMed]
- 76.Corey E, Wipke WT. Computer-assisted design of complex organic syntheses. Science. 1969;166:178–92. [DOI] [PubMed] [Google Scholar]
- 77.Popova M, Isayev O, Tropsha A. Deep reinforcement learning for de novo drug design. Sci Adv. 2018;4:eaap7885. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 78.Merk D, Friedrich L, Grisoni F, Schneider G. De novo design of bioactive small molecules by artificial intelligence. Mol Inf. 2018;37:1700153. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 79.Segler MHS, Kogej T, Tyrchan C, Waller MP. Generating focused molecule libraries for drug discovery with recurrent neural networks. ACS Cent Sci. 2018;4:120–31. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 80.Schneider G, Clark DE. Automated de novo drug design: are we nearly there yet? Angew Chem. 2019;131:10906–17. [DOI] [PubMed] [Google Scholar]
- 81.Castle BC, Forbes RA. Impact of quality by design in process development on the analytical control strategy for a small-molecule drug substance. J Pharm Innov. 2013;8(4):247–64. [Google Scholar]
- 82.Barmpalexis P, Kachrimanis K, Georgarakis E. Solid dispersions in the development of a nimodipine floating tablet formulation and optimization by artificial neural networks and genetic programming. Euro J Pharm Biopharm. 2011;77(1):122–31. [DOI] [PubMed] [Google Scholar]
- 83.Papavasileiou V, Koulouris A, Siletti C, Petrides D. Optimize manufacturing of pharmaceutical products with process simulation and production scheduling tools. Chem Eng Res Des. 2007;85:1086–97. [Google Scholar]
- 84.Achilleos EC, Calandranis JC, Petrides DP. Quantifying the impact of uncertain parameters in the batch manufacturing of active pharmaceutical ingredients. Pharm Eng. 2006;26:34–40. [Google Scholar]
- 85.Vaghani SS, Patel SG, Jivani RR, Jivani NP, Patel MM, Borda R. Design and optimization of a stomach-specific drug delivery system of repaglinide: application of simplex lattice design. Pharm Dev Technol. 2012;17:55–65. [DOI] [PubMed] [Google Scholar]
- 86.Hassanzadeh P, Atyabi F, Dinarvand R. The significance of artificial intelligence in drug delivery system design. Adv Drug Deliv Rev. 2019;151–152:169–90. [DOI] [PubMed] [Google Scholar]
- 87.Luo M, Feng Y, Wang T, Guan J. Micro-/nanorobots at work in active drug delivery. Adv Funct Mater. 2018;28:1706100. [Google Scholar]
- 88.Fu J, Yan H. Controlled drug release by a nanorobot. Nat Biotechnol. 2012;30:407–8. [DOI] [PubMed] [Google Scholar]
- 89.Calzolari D, Bruschi S, Coquin L, Schofield J, Feala JD, Reed JC, McCulloch AD, Paternostro G. Search algorithms as a framework for the optimization of drug combinations. PLoS Comput Biol. 2008;4:e1000249. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 90.Wilson B, Geetha KM. Artificial intelligence and related technologies enabled nanomedicine for advanced cancer treatment. Future Med. 2020;15:433–5. [DOI] [PubMed] [Google Scholar]
- 91.Tsigelny IF. Artificial intelligence in drug combination therapy. Brief Bioinform. 2019;20:1434–48. [DOI] [PubMed] [Google Scholar]
- 92.Li X, Xu Y, Cui H, Huang T, Wang D, Lian B, Li W, Qin G, Chen L, Xie L. Prediction of synergistic anti-cancer drug combinations based on drug target network and drug induced gene expression profiles. Artif Intell Med. 2017;83:35–43. [DOI] [PubMed] [Google Scholar]
- 93.Mason DJ, Eastman RT, Lewis RPI, Stott IP, Guha R, Bender A. Using machine learning to predict synergistic antimalarial compound combinations with novel structures. Front Pharmacol. 2018;9:1096. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 94.Ho D, Wang P, Kee T. Artificial intelligence in nanomedicine. Nanoscale Horiz. 2019;4:365–77. [DOI] [PubMed] [Google Scholar]
- 95.Sacha GM, Varona P. Artificial intelligence in nanotechnology. Nanotechnology. 2013;24:452002. [DOI] [PubMed] [Google Scholar]
- 96.Agrawal P. Artificial intelligence in drug discovery and development. J Pharmacovigil. 2018;6:1000e1173. [Google Scholar]
- 97.Lee E. How do we build trust in machine learning models? Available at SSRN 3822437. 2021.
- 98.Dhanabalan T, Sathish A. Transforming Indian industries through artificial intelligence and robotics in industry 4.0. J Mech Eng. 2018;9:835–45. [Google Scholar]
- 99.Mahajan A, Vaidya T, Gupta A, et al. Artificial intelligence in healthcare in developing nations: The beginning of a transformative journey. Cancer Res Stat Treat. 2019;2:182–9. [Google Scholar]
- 100.Bhattacharya S, Pradhan KB, Bashar MA, Tripathi S, Semwal J, Marzo RR, Bhattacharya S, Singh A. Artificial intelligence enabled healthcare: a hype, hope or harm. J Family Med Prim Care. 2019;8:3461–4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 101.Rinkoo A, Panjiyar A, Songara D, Sharma A, Singh P. From guidelines to practice: assessing the implementation fidelity of partograph use for monitoring labour in primary healthcare settings in india using a tablet-based mobile application. J Health Med Informat. 2019;10:334. [Google Scholar]
- 102.Singh J, Shah D. Employment in the healthcare industry in India. The Arthniti. 2020;1:86–94. [Google Scholar]
- 103.Grus J. Data Science from Scratch First Principles with Python. 1st ed. Sebastopol: O’Reilly; 2015. [Google Scholar]
- 104.Lucas GM, Gratch J, King A, Morency LP. It’s only a computer: Virtual humans increase their willingness to disclose. Comput Human Behav. 2017;37:94–100. [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
Not applicable.





