Artificial intelligence as a tool in drug discovery and development

Abstract

The rapidly advancing field of artificial intelligence (AI) has garnered substantial attention for its potential application in drug discovery and development. This opinion review critically examined the feasibility and prospects of integrating AI as a transformative tool in the pharmaceutical industry. AI, encompassing machine learning algorithms, deep learning, and data analytics, offers unprecedented opportunities to streamline and enhance various stages of drug development. This opinion review delved into the current landscape of AI-driven approaches, discussing their utilization in target identification, lead optimization, and predictive modeling of pharmacokinetics and toxicity. We aimed to scrutinize the integration of large-scale omics data, electronic health records, and chemical informatics, highlighting the power of AI in uncovering novel therapeutic targets and accelerating drug repurposing strategies. Despite the considerable potential of AI, the review also addressed inherent challenges, including data privacy concerns, interpretability of AI models, and the need for robust validation in real-world clinical settings. Additionally, we explored ethical considerations surrounding AI-driven decision-making in drug development. This opinion review provided a nuanced perspective on the transformative role of AI in drug discovery by discussing the existing literature and emerging trends, presenting critical insights and addressing potential hurdles. In conclusion, this study aimed to stimulate discourse within the scientific community and guide future endeavors to harness the full potential of AI in drug development.

Key Words: Artificial intelligence; Drug discovery; Drug development; Decision-making; AI-driven medicine; Healthcare; Public health

Core Tip: Embracing artificial intelligence (AI) expedites drug discovery and development by streamlining computational chemistry, molecular modeling, and data mining. Leveraging AI-driven algorithms enhances the accuracy and efficiency of identifying potential drug candidates and predicting their pharmacological properties. Integrating machine learning and deep learning frameworks into pharmaceutical research optimizes decision-making, accelerates drug design cycles, and ultimately advances novel therapies for various diseases.

INTRODUCTION

Drug discovery and development is a long, expensive, and complex process that can often take more than 10 years from molecule identification to medical drug approval and placement on the market. Each stage in the process carries a risk of failure, and most drug applicants never reach the market. This makes the process of drug innovation and development both expensive and inefficient[1,2].

In recent years, the use of artificial intelligence (AI) in this industry has increased significantly. Drug discovery requires the analysis of large databases of chemical compounds. This can be achieved rapidly using machine learning techniques[3]. These techniques have their limitations because even a little change in the molecular structure of the drug can drastically alter its effect. Drug discovery involves the analysis and comparison of the properties of different molecular structures and components. In this context, AI tools can automatically scan large datasets quickly, using a composition safety check to pick out the most effective model for a certain goal[4,5].

Several public libraries store chemical and biological data, including ChEMBL[6] and PubChem[7]. They contain information on millions of molecules for various disease targets. These libraries are machine-readable and are used for drug discovery models, including for drug candidate compounds targeting severe acute respiratory syndrome coronavirus 2[8]. AI (mostly machine learning techniques) has also been implemented to evaluate toxicity. For example, the DeepTox platform is used as a model to evaluate the toxicity of certain compounds[9]. Another platform, MoleculeNet, can be used to translate molecular structures and predict toxicity[10].

The assessment of drug-target interactions is another important stage of drug design. The binding affinity between the drug and its target is important for the final product. Molecular docking, one of the most common approaches to predict affinity, is used to study the binding and complex formation between two molecules, such as receptor-ligand interactions[11,12].

Different pharmaceutical companies have used AI to improve drug discovery. Verge Genomics uses AI to predict the effects of some new drugs on patients with Alzheimer’s disease and Parkinson’s disease[13]. Despite their use of automated data analysis, certain drug studies have failed. Most neurological diseases are polygenic, but the company drug data targets one gene. In 2018, Verge Genomics developed an algorithm to identify the pathogenic genes and select drugs to target them all. Thus, the company successfully utilized the vast potential of AI and machine learning algorithms by identifying drugs for neurodegenerative diseases. In 2018, Bayer and Merck received Food and Drug Administration approval to use AI algorithms to support clinical decision making for chronic thromboembolic pulmonary hypertension[14]. This form of chronic thromboembolic pulmonary hypertension is very rare and affects approximately 5/1000000 people annually worldwide. The symptoms resemble chronic obstructive pulmonary disease or asthma, complicating its diagnosis even after numerous medical tests.

Novartis currently uses AI algorithms to classify digital images of different cells[15,16]. Each cell is treated with different experimental molecules. The algorithms group and test molecules with similar effects. Finding biologically active molecules requires complicated analysis. Therefore, to speed up this screening process, Novartis research teams use machine learning algorithms to predict which unknown molecules might be worth exploring.

In 2018, the biotech company Cyclica collaborated with Bayer[17], using AI machine learning to determine the polypharmacological profiles of small molecules and develop more affordable drugs. The company created Ligand Express, an integrated network of cloud technologies expanded with AI that enhances drug design, screening, and personalization.

In line with those mentioned above, AI-driven drug design, development, and delivery are highly trendy topics to discuss. We hypothesize that integrating AI into drug discovery and development processes will significantly enhance efficiency, accuracy, and innovation, leading to the discovery of novel therapeutic agents and the optimization of existing drugs. Our goals for this review were: (1) To assess the current state of AI applications in drug discovery, including machine learning algorithms, predictive modeling, and virtual screening techniques; (2) To explore recent advancements in AI-driven platforms and technologies that have revolutionized drug discovery, such as deep learning algorithms, generative models, and molecular design tools; (3) To investigate how AI has impacted various stages of drug development, from target identification and lead optimization to clinical trials and post-marketing surveillance; (4) To showcase case studies and success stories where AI-driven approaches have led to the discovery of promising drug candidates or repurposing of existing drugs for new indications; (5) To identify key challenges and limitations associated with AI in drug discovery, such as data quality issues, ethical considerations, and regulatory hurdles; and (6) To provide insights into potential future directions and emerging trends in AI-enabled drug discovery, including the integration of multi-omics data, collaborative AI platforms, and AI-driven personalized medicine approaches. The current review is crucial as it addresses the intersection of two rapidly evolving fields: AI and drug discovery. Our review aimed to provide a comprehensive summary for AI experts, drug developers, and healthcare professionals by synthesizing the latest research, methodologies, and best practices. The insights gained from this review will inform academia and industry and contribute to the ongoing efforts to accelerate drug discovery, improve patient outcomes, and address unmet medical needs globally.

SEARCH STRATEGY

We conducted a comprehensive search across multiple databases, including PubMed, Scopus, Web of Science, IEEE Xplore, and Google Scholar. We focused on various types of studies, such as review articles, research articles, case studies, clinical trials, and meta-analyses. Using Boolean operators, our search terms included (“Artificial intelligence” OR “AI”) AND (“Drug discovery” OR “Drug development”) AND (“Machine learning” OR “Deep learning” OR “Neural networks”) AND (“Pharmaceuticals” OR “Medications” OR “Compounds”), (“Computational chemistry” OR “Chemoinformatics”) AND (“Drug design” OR “Molecular modeling”), and (“Data mining” OR “Big data analytics”) AND (“Pharmacology” OR “Therapeutics”). This strategy allowed us to gather relevant literature on the application of AI in various stages of drug discovery and development, including computational chemistry, molecular modeling, data mining, and pharmacological analysis. The paper selection is presented in Figure 1.

Figure 1 Identification, screening, and selection of papers to include.

CURRENT LANDSCAPE OF DRUG DISCOVERY

Traditional methods and limitations

Historically, drug discovery has relied heavily on trial-and-error methods, where researchers would test compounds to see if they had the desired therapeutic effect[18]. However, traditional methods are associated with high failure rates and lengthy timelines. It often takes years and significant resources to develop a single drug, with many candidates failing in clinical trials[19]. Moreover, traditional methods often lack the depth of understanding required to fully grasp complex biological systems and disease mechanisms[20].

Evolution of technology in pharmaceutical research

High-throughput screening techniques allow researchers to quickly screen large libraries of compounds for potential drug candidates, speeding up the initial stages of drug discovery[21]. Advances in genomics, proteomics, and other omics technologies have provided researchers with a deeper understanding of biological systems and disease pathways[22]. Computational approaches have become increasingly important, allowing researchers to simulate molecular interactions and predict the properties of potential drug candidates[23].

Emergence of AI in drug discovery

AI techniques, particularly machine learning and deep learning, have revolutionized drug discovery by analyzing large datasets, predicting molecular properties, and identifying potential drug candidates[24]. AI algorithms can perform virtual screening of compound libraries to identify molecules with the highest likelihood of binding to specific targets, reducing the time and cost associated with experimental screening[25]. AI models can predict the pharmacokinetic and pharmacodynamic properties of compounds, helping researchers prioritize the most promising candidates for further development[26].

AI algorithms can analyze patient data to identify biomarkers, predict treatment responses, and tailor therapies to individual patients, leading to the development of more effective and targeted treatments[27]. The integration of AI into drug discovery and development has the potential to significantly accelerate the process, reduce costs, and improve the success rate of bringing new drugs to market. However, challenges, such as data quality, regulatory considerations, and ethical concerns, must be addressed to fully realize the benefits of AI in pharmaceutical research.

APPLICATIONS OF AI IN DRUG DISCOVERY

Target identification and validation

The process of drug discovery starts with molecule searching. Certain small molecule databases can match specific health problems. Ranges of docking software are developed to explore molecular bioactivity[28]. Researchers also use analogues of already known molecules[29].

High-throughput screening and data analysis

High-throughput screening is a process for identifying potential molecules to be included in further analysis. The molecules identified in the previous step are screened to identify the most applicable to the health problem of interest. The process also includes molecule stability and interactions[30]. Wide varieties of methods are employed, including neural network, multiple linear regression, decision tree, and the analysis of variance[31,32].

Predictive modeling for drug design

Drug design is an important process. It includes the form (for example, tablet or solution) as well as the excipients (the other ingredients). AI is also applied at this stage. The tested parameters include the blend bulk and tapped density, flowability, angle of repose, appearance, friability, resistance to crushing, and time of disintegration of the tablet[33].

Biomarker discovery and validation

In the era of molecular medicine, biomarker discovery enhances the drug discovery process. Biomarker discovery requires a large number of samples to be collected and thoroughly analyzed in a uniform manner. Validation ensures that the marker is reproducible and reliable, and that its sensitivity and specificity are acceptable. AI could be employed in this step. Within the drug development context, biomarkers are used as an outcome measure in clinical trials, helping the identification and validation of drug targets. Thus, the right treatment for each patient based on the biomarkers tested would be found[34].

AI-DRIVEN DRUG DEVELOPMENT

In recent years, AI technology has revolutionized pharmaceutical research, ushering in a new era of drug development and revolutionizing medicine discovery, testing, and patient delivery. This revolution can transform various stages of the drug development pipeline, from accelerating preclinical research to optimizing clinical trial design and enabling personalized medicine approaches.

Accelerating preclinical research

Preclinical research involves extensive and time-consuming safety and efficacy assessments of potential drug candidates. This process is expensive, challenging, and often unsuccessful. Here, AI can streamline data analysis, predict drug interactions, and identify promising compounds. Machine learning algorithms have been developed to predict the toxicity of potential drug candidates[35]. Using AI-driven platforms, researchers can rapidly screen thousands of compounds and prioritize the most promising candidates, significantly reducing the time and cost of preclinical testing[36].

Optimization of clinical trial design

Conventional trial designs are often flawed and inefficient, leading to high costs, lengthy timelines, and sometimes inconclusive results. AI can improve patient recruitment, trial outcomes, and novel therapy development by tailoring inclusion criteria and treatment protocols based on predictive analytics[37]. In addition, AI-powered algorithms can analyze patient data, identify relevant biomarkers, and stratify patient populations to optimize trial design[38]. AI-driven simulations allow researchers to explore virtual trial scenarios, refine study protocols, and mitigate risks before initiating costly and time-consuming clinical trials.

Personalized medicine approaches

The advancement of personalized medicine is one of the most promising applications of AI in drug development. The paradigm of disease treatment and drug development is shifting towards personalized therapies to achieve better results for individual patients. AI can accelerate this trend by improving diagnostics, collecting personalized information, and assisting clinical decisions[39]. AI algorithms can store and analyze patient data, such as genetic profiles, clinical histories, and lifestyle factors[40]. This can significantly alleviate the burden of extensive data collection and analysis on researchers, facilitating their speedy and efficient work by allowing them to focus on the clinical scenario. In addition, AI technology can identify biomarkers associated with drug responses or disease progression, further improving targeted therapies with maximal efficacy and minimal adverse effects[41]. In conclusion, AI-driven drug development can transform the pharmaceutical industry by optimizing preclinical research, clinical trial design, and personalized treatment. Researchers can benefit from AI to expedite the discovery and delivery of innovative therapies that address unmet clinical needs and improve patient outcomes.

We present an overview of AI utilization in drug development in Figure 2.

Figure 2 An entire workflow of artificial intelligence applications in drug discovery and development, highlighting the different stages and the corresponding artificial intelligence techniques used. AI: Artificial intelligence; QSAR: Quantitative structure-activity relationship. The figure was generated using brainstorming from OpenAI. (2024). ChatGPT [Large language model]. https://chatgpt.com/c/d95bc2d2-a53a-492d-a78d-6946bc43cef2.

CHALLENGES AND ETHICAL CONSIDERATIONS

Data quality and bias in AI models

Ensuring the quality and reliability of data used to train AI models is crucial for accurate predictions and decision-making. Biases, inaccuracies, and incompleteness in the data can lead to flawed results[42]. AI models can inherit biases present in the training data, leading to biased predictions or decisions. This is particularly concerning in healthcare, where biases related to race, sex, or socioeconomic status can impact patient outcomes[43]. Developing techniques to identify and mitigate biases in AI models, as well as ensuring diverse and representative training datasets, are essential for ethical AI applications in drug discovery[44].

Interpretability and transparency

Many AI models, particularly deep learning models, are often considered ”black boxes” because their internal workings are not easily interpretable by humans. This lack of transparency raises concerns about how decisions are made and undermines trust in AI systems[45]. There is a growing need for interpretable AI models, where the reasoning behind predictions or recommendations can be understood by domain experts and regulatory authorities[46]. Establishing mechanisms to ensure accountability for AI-driven decisions, including transparency about model training and validation, is crucial for maintaining ethical standards in drug discovery[47].

Regulatory and ethical implications

AI applications in drug discovery must adhere to regulatory requirements set forth by agencies, such as the Food and Drug Administration and European Medicines Agency. Ensuring compliance with regulations designed for traditional drug development processes presents challenges due to the unique nature of AI technologies[48]. Demonstrating the safety and efficacy of AI-generated drug candidates or treatment recommendations is essential for regulatory approval. Robust validation and testing procedures are necessary to mitigate risks to patients[49]. Protecting patient privacy and securing sensitive healthcare data used in AI applications is paramount. Adhering to data protection regulations, such as the General Data Protection Regulation and the Health Insurance Portability and Accountability Act, is essential to maintain trust and ethical standards[50].

Addressing these challenges and ethical considerations is crucial for the responsible and ethical use of AI in drug discovery and development. Collaborative efforts between researchers, regulatory bodies, and ethicists are essential to develop guidelines and frameworks that promote the ethical use of AI while maximizing its potential benefits in healthcare.

FUTURE DIRECTIONS AND INNOVATIONS

Advancements in AI technologies

Deep learning is advanced machine learning that could be applied in the field of drug discovery. It is a neural network that can extract information from public databases and create scientific conclusions based on them. Deep learning is applicable to reduce the costs of the clinical trials by predicting their outcome before they start[51]. Another promising application of AI in the field of drug discovery is drug repurposing. Finding new applications for already existing drugs reduces the time and cost of their development[52]. Another new trend in the field of drug discovery and development is AI application in nanotechnologies, especially nanocarriers. AI is also crucial in smart drug release systems that deliver the medicine when it is needed[53].

Collaborative approaches and industry trends

Application of AI in the field of drug discovery requires a multidisciplinary approach by default. Collaboration between researchers, clinical experts, engineers, and data managers is crucial. Thus, multidisciplinary education is required to meet the new demands of pharmaceutical trends[53].

Potential impact on drug development pipelines

AI can speed up the drug discovery and development process and reduce costs. The resources not allocated to drug discovery could be invested into drug searching for different diseases. This could have a large positive impact on public health.

INTEGRATION OF AI INTO MAINSTREAM DRUG DISCOVERY

The use of AI in mainstream drug research signifies a watershed moment set to transform the pharmaceutical business. Adoption strategies and industry readiness are critical components of this transformation, which need strong frameworks for AI adoption and organizational readiness[54].

Training and skill development programs are critical for providing professionals with the requisite skills in AI-driven approaches, guaranteeing smooth integration and realizing the potential advantages. Overcoming industrial opposition and skepticism requires proactive actions to address concerns about AI technology dependability, ethical issues, and data security. Collaboration among stakeholders, regulatory authorities, and AI developers is critical for building confidence and accelerating wider adoption[54]. As AI evolves and demonstrates its usefulness in drug development, proactive involvement, ongoing education, and open communication will be critical in managing difficulties and maximizing the promise of AI-driven advances in the pharmaceutical sector.

In Table 1, we present the AI techniques used for drug development, their application areas, key benefits, and challenges.

Table 1 Applications of artificial intelligence in drug discovery and development.

Application area	AI techniques used	Key benefits	Challenges
Target identification	Machine learning, deep learning	Identifying novel drug targets, high accuracy	Data quality, complexity of biological systems
Drug screening	Virtual screening, predictive models	Faster screening of compounds, cost-effective	False positives/negatives, model validation
Lead optimization	QSAR models, reinforcement learning	Improved candidate selection, reduced development time	Integration with traditional methods, data scarcity
Preclinical development	Image analysis, natural language processing	Enhanced understanding of drug toxicity and efficacy	Interpretation of complex data, standardization
Clinical trials	Predictive analytics, patient recruitment algorithms	Optimized trial design, better patient stratification	Ethical concerns, data privacy
Personalized medicine	Genomic data analysis, personalized algorithms	Tailored treatments, improved patient outcomes	Data integration, regulatory issues

AI: Artificial intelligence; QSAR: Quantitative structure-activity relationship.

CONSIDERATIONS FOR SMALL AND LARGE PHARMACEUTICAL COMPANIES

Customization of AI tools

Small pharmaceutical companies may benefit from customizable AI tools that are tailored to their specific research needs and capabilities. Customization allows them to focus on targeted areas of drug discovery where AI can provide the most value. Larger pharmaceutical companies may have the resources to develop or acquire sophisticated AI platforms that can be customized for various stages of drug development, from target identification to clinical trial optimization. Customization enables them to integrate AI seamlessly into their existing workflow and infrastructure[55].

Resource allocation and return on investment

Small pharmaceutical companies must carefully allocate limited resources when investing in AI technologies. They need to assess the potential return on investment of implementing AI tools and prioritize projects with the highest likelihood of success. Large pharmaceutical companies have greater financial resources and may allocate significant budgets to AI initiatives[55]. However, they also face pressure to demonstrate tangible return on investment and ensure that AI investments align with strategic business objectives.

GLOBAL COLLABORATIONS AND DATA SHARING

Small pharmaceutical companies may lack access to large datasets necessary to train AI models effectively. Collaborating with academic institutions, research organizations, or larger pharmaceutical companies can provide access to diverse datasets and expertise, facilitating more robust AI applications. In contrast, large pharmaceutical companies often have extensive internal datasets, but collaboration with external partners can still be beneficial. Engaging in global collaborations and data sharing initiatives allows them to access additional resources, validate AI models across diverse populations, and accelerate drug discovery efforts[53].

Both small and large pharmaceutical companies can leverage AI in drug discovery and development, but considerations, such as customization of AI tools, resource allocation, and global collaborations differ based on their size, resources, and organizational capabilities.

CONCLUSION

This review showed that integrating AI can revolutionize drug discovery and development. Utilizing machine learning algorithms, deep learning techniques, and data analytics, AI can expedite target identification, optimize lead compounds, and predict pharmacokinetics and toxicity. By acknowledging the inherent challenges, including insufficient resources and model interpretability on a larger scale, this review demonstrated the need for robust validation and ethical considerations in AI-driven drug development. Nevertheless, the future demonstrates a pressing need for industry stakeholder and research community collaboration to overcome these hurdles and harness the full potential of AI to drive innovation and improve patient outcomes in medical research.