Designing new drugs can be a tedious and laborious process. The result has been that relatively few effective new drugs have been clinically implemented in the past several decades. Utilizing the power of artificial intelligence could speed up the process of discovering novel molecules for hard-to-treat diseases. Bypassing the typical method of trial-and-error may benefit patients in a more timely and cost-effective manner. In this article, we provide a guide to AI-designed drugs and the importance of AI in drug discovery.
Definition and Importance of AI-Designed Drugs
Drug design and development has traditionally faced challenges of prohibitive cost and protracted time (mean duration of 12 years). In the past few decades, artificial intelligence (AI) techniques, which include machine learning (ML) and deep learning (DL), have been developed for drug design and development. Today, AI is used throughout clinical research and practice, from early drug discovery to real-world data mining. With advancements in technology, computer-aided drug design (CADD) integrated with AI algorithms addresses traditional drug design and development obstacles to efficiently bring AI-designed drugs to patients in need.
AI-designed drugs are a class of drugs developed using AI technology at any point during the drug discovery process. The discovery of these drugs leverages the power of ML algorithms to identify potential targets and design molecules that can interact with those targets.
The Intersection of AI and Pharmaceuticals
AI is applied extensively in drug design and development, including preclinical and clinical development, de novo drug design, activity scoring, virtual screening (VS), and in silico evaluation of the properties of a drug molecule. In addition, AI is used in the classification of active and inactive molecular motifs, monitoring drug release, biomarker development, pharmaceutical manufacturing, bioactivity identification and physiochemical properties, prediction of toxicity, and identification of mode of action. AI is used to effectively mine massive amounts of pharmacological data and identify target proteins, thereby decreasing health hazards associated with preclinical trials, reducing cost substantially, and improving the drug success rate.
The Role of AI in Drug Discovery
Overview of traditional drug discovery methods
Traditional drug discovery methods face significant challenges, which result in high-cost, time-consuming, complicated, and laborious steps to advance a new drug. Difficulties arise throughout the entire drug discovery process, from design to testing, and these include low efficacy and off-target effects and the need for conventional wet laboratory testing. Advancements in AI technologies have shifted the focus of drug discovery and development away from traditional methods toward bioinformatics tools. Combined with widely used data resources, AI is changing the drug discovery landscape.
Advantages of AI in drug design
The application of AI-based methods enables scientists to overcome the challenges of traditional drug discovery and development and improves the ability of sponsors to efficiently discover effective drugs. CADD, integrated with AI, ML, and DL technologies, can handle vast amounts of biological data, leading to reduced time and cost of drug development.
In addition, AI and ML facilitate effective and efficient solutions to complex problems, including:
- developing effective drug design, chemical synthesis, drug screening, polypharmacology, and drug repurposing
- recognizing hit and lead compounds
- expediting drug target validation
- optimizing drug structure design
- VS of compounds from large chemical libraries (more than 106 million compounds)
Key technologies powering AI-designed drugs
AI is used throughout clinical research, and DL methods like artificial neural networks (ANN) have played a role in modernizing drug discovery. As clinical trials shift to decentralized models, the use of wearable medical technology has grown. Besides automatically collecting and processing data inputs, AI can automate decision-making regarding device notifications and generate recommendations based on patterns in their health data.
Key technologies and resources powering AI-designed drugs include:
- data resources such as ChEMBL and DrugBank
- algorithms to develop AI-based models for drug discovery
- AI techniques for predicting drug toxicity, drug bioactivity, and drug physicochemical property
- AI-based models for de novo drug design, drug-target interaction, and binding affinity prediction
- advanced AI in drug synergism/antagonism prediction and nanomedicine design
- unsupervised clustering of drugs or patients aimed at identifying potential drug compounds or suitable patient populations
- supervised ML approaches to improve therapeutic drug monitoring
- natural language processing to mine electronic health records (EHR) to obtain real-world data
- modern ML technologies that complement traditional CADD and accelerate candidate optimization
- data pre-processing and applications of big data and AI methods for the accurate analysis of biomedical data and developing predictive models in drug design.
How AI Designs Drugs
Advancements in AI have impacted drug design and discovery and changed nearly every stage of the process, including target identification, molecular simulations, prediction of drug properties, de novo drug design, candidate drug prioritization, and synthesis pathway generation. Seo & Lee (2024) describe applications of big data and AI-driven technologies in CADD, including predicting absorption, distribution, metabolism, excretion, and toxicity properties, as well as finding binding sites in target proteins and conducting structure-based virtual screening (SBVS).
According to Gupta et al. (2021), ML and DL algorithms have been used in peptide synthesis, SBVS, ligand-based VS, toxicity prediction, drug monitoring and release, pharmacophore modeling, quantitative structure-activity relationships (QSARs), drug repositioning, polypharmacology, and physiochemical activity. In addition, novel data mining, curation, and management techniques support modeling algorithms.
Data collection and analysis
Data pre-processing methods and applications of big data and AI techniques enable the accurate and thorough analysis of large volumes of raw data and the design of prediction models. With decentralized trials, data collection and processing are challenging as patients are off-site and have to regularly submit their participation data. Contract research organizations (CROs) and medical research institutions can leverage AI to solve these issues by creating algorithms to analyze patient data. Additionally, AI can optimize and generate notifications that prompt patients to complete electronic clinical outcome assessments (eCOA) for a more reliable data pool.
Machine learning algorithms in drug discovery
Applying ML methods to biomedical data has the potential for developing personalized therapies, drug repurposing, and drug discovery. AI and ML techniques have contributed to drug discovery over the past 15 – 20 years and have had a transformative effect on clinical research, from peptide synthesis to molecule design, VS to molecular docking, QSARs to drug repositioning, protein misfolding to protein-protein interactions, and molecular pathway identification to polypharmacology. In oncology, Petinrin et al. (2023) study the application of ML and DL techniques to analyze metastatic cancer research data, most of which are collected as PET/CT and MRI imaging data. Kolluri et al. (2022) describe the positive disruption brought about by AI/ML in clinical trial design, conduct, and analysis.
Predictive modeling and simulations
Predictive ML models can be used for drug response simulation. Researchers utilize biosimulation to run human testing of drugs virtually to gauge the human response to a medical compound. Biosimulation uses math-based computer simulations to replicate a body’s biological processes and systems. Through AI and ML, the simulation creates predictions based on models and data, which researchers can use to gather information about the drug’s effects.
In a study by Khatami et al. (2021), highly predictive ML models were leveraged for patient drug response simulation. The authors demonstrated that their approach can also deconvolute a drug’s mechanism of action and propose combination therapies.
Target identification and validation
Drug discovery aims to identify proper targets and drug candidates in response to unmet health needs. Targets can be identified and validated using AI, enhancing the designed drugs’ success rate. Targets, e.g., genes involved in disease pathophysiology, are identified through gene expression, genome-wide association studies (GWAS), identification of risk genes, and data mining of published literature. ML-based biology network analysis is applied to large, complex data to identify reliable potential novel targets. Identifying elusive targets paves the way for diversifying discovery pipelines while ML technologies accelerate candidate optimization.
Examples of AI-Designed Drugs
Notable success stories in pharmaceuticals
In recent years, AI-designed drugs have made the news with historic milestones. In 2020, the first-ever AI-designed drug molecule entered human clinical trials. A year later, an AI system predicted the protein structures for 330,000 proteins and has since expanded to over 200 million. In 2022, the first-ever AI-discovered molecule based on an AI-discovered novel target started Phase I clinical trials. In early 2023, the Food and Drug Administration (FDA) granted its first Orphan Drug Designation to a drug discovered and designed using AI.
Breakthroughs in disease treatment and prevention
One of the lasting effects of COVID-19 is the impact on management problems of health services, notably the prevention, diagnosis, and treatment of breast cancer. Silva-Aravena et al. (2023) present a decision support strategy for health teams based on ML tools and explainability algorithms (XAI), making it possible to predict the disease and interpret the impact on patient health. They demonstrate that using ML allows health teams to offer early and personalized alerts for each patient.
Hospitals and medical research organizations have begun to design disease-detection algorithms that examine and analyze symptoms, medical histories, and procedures preceding a diagnosis. The algorithm then identifies whether a patient is likely to develop a disease or is in the early stages of a medical condition. This allows for more proactive care and treatments, delaying disease progression or limiting symptoms.
Ongoing research and development
AI/ML-based drug discovery companies are developing data-driven platforms to support the end-to-end drug discovery process. In addition, Sarkar et al. (2023) anticipate that technological advances such as “message-passing paradigms,” “spatial-symmetry-preserving networks,” and “hybrid de novo design” will contribute more significantly to the development of AI-designed drugs in the future.
Challenges and Limitations
Inconsistent data management is a challenge as CROs face difficulties integrating AI technologies with their data capture and processing, especially on a global scale. A lack of standardization complicates data consolidation an AI program can analyze. A recent study by Petinrin et al. (2023) on ML and DL in metastatic cancer research highlights several limitations, notably the black-box nature, high computational cost, and risk of overestimation for generality due to the non-diverse population in clinical trial datasets. Medical research and genetic databases are skewed toward Caucasian and European patients, resulting in an inherent bias in research databases. Due to accessibility and affordability issues, AI tech may only be available to a limited group of organizations.
Ethical considerations
Ethical concerns include the potential for:
- AI to be used to make decisions affecting the health and well-being of the general population, e.g., which drugs to develop and which clinical trials to conduct
- bias in AI algorithms, which could result in unequal access to care and the unfair treatment of particular groups of people.
- job losses in the pharmaceutical industry due to automation
To address these limitations, Gonzalez et al. (2023) suggest the following measures: ensuring AI systems are trained on diverse and representative data and regularly reviewing and auditing AI systems for bias.
Data privacy and security concerns
AI systems need large amounts of data to process and analyze, which creates a risk of accessing or misusing personal information. Gonzalez et al. (2023) call for strong data privacy and security protocols and compliance with regulations pertaining to the collection and use of sensitive medical data.
Regulatory and legal hurdles
The FDA requires a significant amount of data from clinical trials before it will approve a drug. While AI algorithms can help identify potential drug targets and design molecules that interact with those targets, they cannot replace the need for clinical trials. Additionally, the FDA requires a thorough understanding of the mechanism of action before approving a drug, but depending on the complexity of the AI algorithm, this may not be transparent.
The Future of AI-Designed Drugs
Potential impact on healthcare and medicine
AI has transformed the drug discovery process and continues to improve efficiency and accuracy, accelerate drug development, and provide the capacity to develop more effective and personalized treatments.
Emerging trends and innovations
Advanced applications of AI in drug design include:
- AI in drug synergism and antagonism prediction – exploring possible drug combinations at a lower cost and with more efficiency
- AI in nanomedicine design, e.g., an ANN for nanomedicine composition optimization
- AI in the design of oligonucleotide therapeutics, a novel class of drugs composed of short strands of DNA or RNA
Collaboration between scientists and AI systems
Pharmaceutical companies are establishing collaborations with AI companies to expedite drug discovery and development. Collaborations between pharmaceutical scientists and AI researchers fuel the development of innovative therapeutics. The benefits of such collaborations include:
- creation of powerful algorithms and ML models based on the combined expertise, knowledge, and experience of the two parties
- improves accuracy and efficiency of clinical trials as AI algorithms can analyze data and identify trends and potential adverse effects
- improves accessibility and affordability of healthcare
- analyzing data from large populations to identify patterns that help predict the effectiveness of drug candidates for specific populations
- help to identify new targets and improve the effectiveness of existing treatments.
Conclusion
Recap of the significance of AI in drug design
AI is used throughout clinical research and practice, from early drug discovery to real-world data mining. With advancements in technology, computer-aided drug design (CADD) integrated with AI algorithms addresses traditional drug design and development obstacles to efficiently bring AI-designed drugs to patients in need.
The evolving landscape of pharmaceuticals
Advancements in AI technologies have shifted the focus of drug discovery and development away from conventional methods toward bioinformatics tools. The successful application of AI depends on the availability of high-quality data, addressing ethical concerns, and recognizing the limitations of AI-based approaches.
Encouragement for further exploration and research in AI-designed drugs
A review of AI-based drug discovery research from 1991 to 2022 proposed additional exploration and research on:
- AI algorithms with the best accuracy for specific tasks in drug discovery
- the potential of unsupervised learning algorithms for drug discovery
- whether the explainability or accuracy of AI be prioritized in drug discovery
- prediction accuracy to classify an algorithm as reliable for drug discovery
- increasing the explainability of specific algorithms
- acceptance by pharmacists and customers, e.g., patient trust in AI-designed drugs
- data management and security, e.g., protecting AI algorithms from malicious manipulation, protecting data from theft
- law and regulation, e.g., ensuring privacy regulations such as HIPAA, compliance with government principles for the pharmaceutical industry, and ensuring quality assurance processes for a ML algorithm.
Vial x Battery Bio: Reimagining AI-Designed Drugs
From decentralization to biosimulation, contract research organizations (CROs) and medical research institutions leverage AI to transform the execution of clinical research and trials. CROs like Vial leverage these technologies to make the drug discovery and development process quicker and more efficient, e.g., AI can assist in patient enrollment by parsing through population data to detect demographics and subgroups that will benefit the most from participating in the trial.
Technology-Driven CRO
Vial is a next-generation, technology-first CRO reimagining clinical trials to deliver faster, more efficient trial results at dramatically lower costs for biotech sponsors. Vial CRO’s modern, intuitive technology platform integrates trial onboarding, patient enrollment, site communication, and data collection into one connected system. By deploying technology at every step, we are driving efficiencies in speed and cost savings for innovative biotech companies of all sizes. Learn more.
Reimagining drug discovery
Battery Bio is Vial CRO’s internal drug discovery arm. The Battery Bio hyper-scalable platform designs therapeutics, and the automated lab turns AI designs into reality in days. The lab is scalable and fully automated through next-generation robotics. Battery Bio believes that a dramatic reduction in the cost of trials combined with a high-velocity clinical strategy will transform medicine and enable scientists to bring curative therapies to patients dramatically faster.
