Prof. Maurício Veloso Brant Pinheiro, Physics Department, UFMG
Prof. Sérgio Veloso Brant Pinheiro, Faculty of Medicine, UFMG
One sometimes finds what one is not looking for. When I woke up just after dawn on September 28, 1928, I certainly didn’t plan to revolutionize all medicine by discovering the world’s first antibiotic, or bacteria killer. But I suppose that was exactly what I did.
— Sir Alexander Fleming —
I. Introduction
In 2019, the pharmaceutical industry experienced a significant milestone as an AI model identified the experimental drug SU-3327 as a potential antibiotic. Originally crafted by a team of researchers at the Burnham Institute for Medical Research in the United States in 2009, this compound had initially been investigated for its potential in treating diabetes. However, due to unsatisfactory results in testing, the development for diabetes treatment was halted. Despite this initial setback, the latent potential of SU-3327 was not entirely overlooked, paving the way for its subsequent rediscovery in the realm of antibiotic research.
The in silico discovery that SU-3327 possesses potent antibiotic properties was accomplished through a deep learning approach by AI researchers, including bioengineer James Collins, at the MIT Jameel Clinic. This groundbreaking application of AI in drug discovery marked a paradigm shift in the field, highlighting the capability of machine learning algorithms to identify novel applications for existing drugs. This AI application bears resemblance to its counterpart in material science, where it is utilized for discovering and predicting new properties, and even forecasting new materials based on specified attributes.
After that, the experimental drug SU-3327 underwent a name change to “halicin,” inspired by the AI computer Hal900 from 2001: A Space Odyssey.
II. From Concept to Cure: The Grueling Journey of Drug Development
To fully comprehend the transformative impact of AI in the pharmaceutical industry, it is crucial to acknowledge the substantial challenges and expenses associated with developing a new drug. The creation of a novel pharmaceutical involves a complex, time-intensive, and financially demanding process with multiple stages, each presenting unique challenges and incurring significant costs. This process is often likened to the complexity of rocket science.
Spanning from the laboratory to the market, this intricate journey consists of numerous steps and potential points of failure. It commences with the identification of target candidates, and the process involves rigorous testing for efficacy and safety. At every stage, from early discovery through production and delivery, a thorough exploration, characterization, and understanding are imperative. A structured roadmap for drug development can be outlined as follows:
Drug Discovery:
- Target Identification: Researchers pinpoint specific molecules, proteins, or biological processes suitable for therapeutic intervention.
- Target Validation: Selected targets undergo validation to ensure their relevance and feasibility for drug development.
- Lead Discovery: Identification of potential compounds or molecules exhibiting desired therapeutic effects on the chosen target.
- Lead Optimization: Refinement of selected compounds to enhance their efficacy, safety, and pharmacological properties.
Pre-clinical Trials:
- In Vitro Studies: Potential drugs undergo testing in controlled laboratory environments using cells or tissues to assess safety and effectiveness.
- In Vivo Studies: Administration of the drug to animals for evaluating safety, efficacy, and potential side effects.
- Safety Pharmacology: Assessment of the drug’s impact on vital physiological functions to ensure it does not cause harmful effects.
Clinical Trials:
- Phase I – Safety: Involves a small group of healthy volunteers to evaluate safety, dosage range, and potential side effects.
- Phase II – Efficacy: Administered to a larger group to assess effectiveness and further evaluate safety.
- Phase III – Large-Scale Efficacy: Conducted on a much larger population to confirm efficacy, monitor side effects, and compare with existing treatments.
- Phase IV – Post-Marketing Surveillance: Continuous monitoring post-approval to identify and address unforeseen side effects or long-term impacts.
Regulatory Filings:
- New Drug Application (NDA) or Biologics License Application (BLA): Compilation of comprehensive data from pre-clinical and clinical trials, along with manufacturing information, to seek regulatory approval.
- Regulatory Review: Evaluation of submitted data by regulatory agencies (e.g., FDA, EMA) to determine if the drug is safe and effective for public use.
Post-Marketing Surveillance:
- Pharmacovigilance: Ongoing monitoring of the drug’s safety profile on the market, including identification, assessment, understanding, and prevention of adverse effects or other drug-related problems.
- Risk Management: Implementation of strategies to minimize identified risks and ensure the ongoing safety of the drug.
The journey of a new drug from the laboratory to the pharmacy shelf is thus a marathon, not a sprint. The timeline for drug development is extensive, often spanning more than a decade. In some instances, particularly in certain areas of biomedical research, this process can stretch up to over 30 years. This prolonged duration underscores the meticulous nature of the research involved and the comprehensive evaluation required at each stage. Each new medicine must successfully navigate through the entire research and development (R&D) process, a testament to the rigorous standards of safety and efficacy that must be met.
Parallel to the time investment, the financial commitment required to bring a new drug to market is substantial. The costs associated with drug development can range from hundreds of millions to a staggering billions of dollars. On average, the cost of developing a single medicine has reached an all-time high of $3 billion. This figure encompasses a wide array of expenses, including those incurred during the drug discovery phase, pre-clinical and clinical trials, regulatory filings, and post-marketing surveillance. Each of these stages represents a significant investment, both in terms of time and resources, further emphasizing the magnitude of the undertaking involved in drug development.
In conclusion, the development of a new drug is a complex, time-consuming, and costly process. Despite these challenges, the pharmaceutical industry continues to innovate, driven by the ultimate goal of improving patient health and quality of life. The significant time and financial investments underscore the importance of continued support and investment in pharmaceutical research and development. The rewards, in terms of improved patient outcomes and advancements in medical science, are well worth the investment.
The Grueling Journey of Drug Development
Years ago, I collaborated closely with colleagues from the ICB at Universidade Federal de Minas Gerais (UFMG) to develop innovative nanostructure drugs based on water-soluble fullerene. Our primary objective was to leverage the inherent properties of fullerene to create a therapeutic solution with significant potential against leishmaniosis.
This collaborative research yielded noteworthy outcomes, resulting in the successful production of two patents in 2011 and 2013.
Despite these achievements, the project encountered a significant hurdle as it failed to advance beyond the in vitro phase of the pre-clinical stage.
This setback was primarily attributed to the project’s struggle with inadequate financial support and a notable lack of enthusiasm from local pharmaceutical companies, hampering the crucial progression to in vivo experimentation.
III. Why Testing AI Drug Development with Antibiotics?
The global challenge of antibiotic resistance is becoming increasingly complicated and alarming as the crisis escalates due to factors such as the overuse of antibiotics, inadequate infection control measures, and the adaptability of microorganisms. This combination leads to the creation of antibiotic-resistant strains, posing a significant threat to public health. We are at a crucial point that demands new and smart strategies to combat antibiotic resistance and protect people worldwide.
Turning our attention to the pressing need for new antibiotics, the challenges posed by resistant bacterial strains and the dynamic landscape of infectious diseases are intensifying. Bacteria are continually improving their resistance to current drugs, presenting a serious problem for healthcare globally. The failure to rapidly develop new antibiotics to keep up with bacterial adaptations could result in severe consequences. Thus, it is of utmost importance to find innovative ways to produce antibiotics capable of addressing the growing problem of antibiotic resistance.
Additionally, a noticeable slowdown in the discovery and approval of new antibiotics has occurred in recent decades. This deceleration can be attributed to various factors, including economic challenges, regulatory hurdles, and pharmaceutical companies prioritizing other areas. The gravity of the situation underscores the need for renewed efforts in pharmaceutical innovation. A collective reassessment of existing approaches and policies is essential to reignite momentum in antibiotic discovery. The urgency emphasized earlier is now accentuated by the recognition of a critical need for renewed commitment and innovation in the field of antibiotic development.
These were the compelling facts that spurred the quest for a novel antibiotic, leveraging pre-existing drugs through the application of artificial intelligence (AI) methodologies. This innovative approach harnesses the power of AI to identify and optimize potential candidates from existing pharmaceutical compounds, paving the way for the development of more effective and targeted antibiotic treatments.
IV. The Rediscovery of Hailicin
The discovery of Halicin, a potent antibiotic, was indeed a significant milestone in drug development, achieved through the innovative application of artificial intelligence (AI). This event demonstrated the transformative potential of AI in streamlining and expediting multiple stages of the drug development marathon, significantly reducing both the time and financial investments traditionally associated with the process.
The team behind this groundbreaking discovery was led by James Collins, a bioengineer at the Massachusetts Institute of Technology (MIT). The team also included Regina Barzilay, an AI specialist, and Jonathan Stokes, a postdoc at MIT and the Broad Institute of MIT and Harvard.
They utilized a deep learning model, specifically a neural network, to predict molecules with antibacterial properties. This model was trained to learn the properties of molecules atom by atom. The algorithm was designed to predict molecular function without any assumptions about how drugs work and without labeling chemical groups, allowing the model to learn new patterns unknown to human experts.
The AI model was trained on approximately 2,500 molecules, including FDA-approved drugs, natural products, and other compounds. The model was trained on each of these data sets to predict which molecules would be effective against Escherichia coli. Once the model was trained, it was used to screen a library of about 6,000 compounds. The model was able to analyze these compounds in a matter of hours and identify Halicin as a potential antibiotic. Halicin was found to be structurally divergent from conventional antibiotics, suggesting that it kills bacteria using different mechanisms than those of existing drugs.
The AI’s identification of Halicin as an antibiotic was substantiated through a sequence of rigorous tests, commencing with in vitro cell testing against Staphylococcus aureus biofilms, and subsequently progressing to in vivo experiments carried out on mice. Halicin demonstrated noteworthy efficacy against drug-resistant strains of Clostridiodes difficile, Acinetobacter baumannii, Mycobacterium tuberculosis and other pathogens. This approach unveiled Halicin as one of the most potent antibiotics discovered to date, showcasing its effectiveness against a broad spectrum of bacteria, including strains resistant to all known antibiotics.
The discovery of Halicin’s antibiotic properties marks a significant milestone in the use of AI for drug discovery. It demonstrates the potential of AI to identify new uses for existing drugs and to accelerate the discovery of new drugs. As we move forward, the role of AI in drug discovery is expected to become increasingly prominent, opening up new possibilities for the development of novel therapeutic agents.
VI. Broadening the Scope: Beyond Antibiotics
Artificial Intelligence (AI) has emerged as a transformative force in drug discovery, revolutionizing traditional methodologies with its multifaceted and adaptable approach. Initially designed for antibiotic discovery, AI has showcased its versatility across various therapeutic areas, marking a paradigm shift.
Jacob Durrant, a Professor at the University of Pittsburgh, validates the extension of AI’s potential beyond antibiotics, emphasizing its adaptability in discovering drugs for diverse therapeutic purposes. Beyond infectious diseases, the AI-driven approach proves valuable in addressing complex conditions like cancer and neurodegenerative disorders, highlighting its broad applicability. In essence, AI is not merely a tool but a dynamic force reshaping pharmaceutical research, offering promising solutions for diverse medical challenges and advancing therapeutic options across fields.
Beyond Antibiotics
UFMG scientists, led by professors Renata Oliveira, Vinícius Maltarollo, José Eduardo Gonçalves, and Isabela César, have developed AI-assisted molecules effective against fungal infections like Cryptococcosis and Candidiasis.
Published in the Journal of Medicinal Chemistry, the study details the planning, synthesis, and evaluation of a new antifungal thiazolylhydrazone. Machine learning aided in substance selection, predicting biological activity, solubility, and lipophilicity.
The compounds demonstrated efficacy against various fungi, with modifications for enhanced water solubility.
The research, covering medicinal chemistry, bioavailability, microbiology, and cytotoxicity, is progressing to patenting.
UFMG now seeks collaborations for development and clinical trials, holding promise for immunocompromised patients, transplant recipients, and those with HIV. Co-authors include researchers from UFMG’s Faculty of Pharmacy and ICB, alongside collaborators from UFOP, FIOCRUZ, and the University of Tübingen.
The biotech sector is undergoing a remarkable transformation with the widespread integration of artificial intelligence (AI), particularly in the realm of drug discovery. The AI-driven pipeline, encompassing small-molecule drugs in discovery and clinical trials, is poised to experience an impressive annual growth rate of 37.3% between 2023 and 2030.
This surge in AI application is mirrored by a substantial increase in investments in AI-driven drug discovery over the past few years. Third-party investments have more than doubled annually for the last five years, surging from $2.4 billion in 2020 to surpassing $5.2 billion by the close of 2021. The momentum continued into 2022, reaching a staggering $24.6 billion. By the third quarter of 2023, investments had skyrocketed to an impressive $60.3 billion, highlighting the escalating interest and potential of AI as a pivotal tool in expediting drug discovery.
Validation of the transformative potential of AI-driven generative platforms is evident. While the exact number of AI-designed clinical stage assets as of 2024 remains elusive, the surge in AI utilization in drug development is unmistakable. In 2021, 132 entries were recorded, a stark contrast to the single entry in 2016, signifying a paradigm shift in the field of drug discovery.
AI’s influence on the economics of drug discovery is equally compelling. Expert surveys and analyses of scientific publications suggest that AI-enabled workflows could save up to 40% of the time needed to advance a new molecule for a challenging or poorly understood target to the preclinical candidate (PCC) stage. The potential cost savings for such a molecule until the PCC stage could be as high as 30%, collectively reducing the time and cost of drug discovery by approximately 25 to 50%.
The groundbreaking discovery of Halicin serves as a pivotal milestone, showcasing the transformative power of AI in reshaping pharmaceutical innovation and ushering in a new era of possibilities in medical science. Beyond its immediate impact on bacterial infections, the implications of AI in drug discovery extend to diverse therapeutic areas. The transformative influence of the AI approach is not only addressing antibiotic resistance but also reshaping pharmaceutical innovation, opening doors to novel possibilities in medical science.
VII. Alphafold
Surging ahead in the domain of machine-learning drug development is AlphaFold, a cutting-edge deep learning system crafted by DeepMind, a subsidiary of Alphabet Inc. (Google’s parent company). AlphaFold stands at the forefront, specializing in predicting protein folding—a pivotal element in unraveling the intricate three-dimensional structures of proteins.
The three-dimensional structure of a protein is essential for comprehending its function, and computationally predicting it has long been a challenge in biology. In the context of AlphaFold, “folding” refers to how a protein chain, composed of amino acids, adopts a specific three-dimensional structure, known as its folded state. The precise arrangement of atoms in this folded structure is crucial for the protein to perform its biological functions.
AlphaFold utilizes deep neural networks, a type of artificial intelligence, to predict the 3D structure of proteins. It gained considerable attention in the scientific community for its success in the Critical Assessment of Structure Prediction (CASP) competition. Notably, AlphaFold exhibited remarkable accuracy in predicting protein structures, surpassing many other methods and significantly advancing the field of structural biology.
The accurate prediction of protein structures through artificial intelligence (AI) holds profound implications for various biological applications. Several key areas stand to benefit significantly from this technological advancement:
- Drug Discovery: Understanding the intricate structure of proteins is pivotal in drug discovery, enabling the identification of potential targets for new drugs. AI plays a crucial role by swiftly and accurately predicting protein structures, thereby expediting the process and reducing the time and cost associated with traditional methods. Additionally, AI facilitates the design of drugs that can efficiently bind to these targets, enhancing the efficacy of treatments.
- Understanding Diseases: Proteins’ malfunction is often the root cause of many diseases. AI, by predicting the structure of these proteins, aids researchers in comprehending the molecular-level mechanisms underlying diseases. This knowledge can pave the way for the development of innovative therapeutic strategies.
- Personalized Medicine: AI’s ability to predict the structure of proteins, accounting for genetic variations among individuals, sets the stage for personalized medicine. Tailoring treatments to an individual’s unique genetic makeup becomes possible, offering more effective and targeted interventions.
- Enzyme Design: In industrial applications, AI contributes to the design of new enzymes. By predicting the structure of these enzymes, researchers can engineer them to catalyze specific chemical reactions, thereby optimizing industrial processes for greater efficiency.
- Vaccine Development: For infectious diseases, AI proves invaluable in predicting the structure of proteins located on the surface of pathogens. This information becomes instrumental in designing vaccines that can effectively trigger an immune response, contributing to more targeted and potent vaccine development.
In summary, the precise prediction of protein structures through AI stands poised to revolutionize various facets of biology and medicine. This technological advancement promises breakthroughs in drug discovery, enhanced understanding of diseases, personalized medicine, streamlined enzyme design for industrial applications, and more effective vaccine development.
VII. Conclusion
The future impact of AI-discovered drugs is profound and revolutionary. AI is not just expediting the discovery of new drugs; it is also predicting potential drug interactions and formulating personalized treatment plans, revolutionizing every stage of drug discovery.
This process commences with target identification, where AI is trained on extensive datasets, encompassing omics datasets (genomics, proteomics, metabolomics, transcriptomics, epigenomics, and microbiomics), phenotypic and expression data, disease associations, patents, publications, clinical trials, research grants, and more. This deep learning allows AI to comprehend the biological mechanisms of diseases and identify novel proteins and/or genes for targeting. Systems like AlphaFold further enhance this process by predicting 3D structures for 330,000 proteins, including all 20,000 proteins in the human genome, accelerating the design of appropriate drugs.
The subsequent stage involves molecular simulations. AI diminishes the necessity for physical testing of candidate drug compounds by enabling high-fidelity molecular simulations that can be entirely conducted on computers (i.e., in silico), overcoming the prohibitive costs of traditional chemistry methods. AI systems also predict crucial properties like toxicity, bioactivity, and physicochemical characteristics of molecules, bypassing simulated testing of drug candidates. This is a pivotal step in forecasting the properties of drugs.
In the de novo drug design stage, AI is revolutionizing the traditional paradigm of drug discovery, historically reliant on screening large libraries of candidate molecules. Some AI systems can generate promising and unprecedented drug molecules entirely from scratch. Once a set of promising “lead” drug compounds is identified, AI is employed to rank these molecules and prioritize them for further assessment, surpassing previous ranking techniques. This process is known as candidate drug prioritization.
Finally, in the synthesis pathway generation stage, AI extends beyond theoretical drug design, generating synthesis pathways for producing hypothetical drug compounds and, in some cases, suggesting modifications to compounds to streamline manufacturing. This underscores the practical applicability of AI in drug discovery.
In conclusion, the integration of AI in drug discovery is paving the way for a future where one person can accomplish what previously required the efforts of a hundred. This revolution in drug discovery and testing is anticipated to yield life-changing, game-changing drugs—on a scale and at a pace unprecedented in history.
#AI #AlphaFold #Antibiotics #Antioxidant #ArtificialIntelligence #C60 #Candidiasis #Cryptococcosis #Cure #DrugDevelopment #Drugs #FDA #Fullerene #Halicin #MachineLearning #Medicine #NeuralNetworks #Pharmacy #Thiazolylhydrazone
References:
Wouters, Olivier J., Martin McKee, and Jeroen Luyten. “Estimated research and development investment needed to bring a new medicine to market, 2009-2018.” Jama 323.9 (2020): 844-853.
Kissinger, Henry A., Eric Schmidt, and Daniel Huttenlocher. The age of AI: and our human future. Hachette UK, 2021.
Marchant, Jo. “Powerful antibiotics discovered using AI.” Nature (2020).
Stokes, Jonathan M., et al. “A deep learning approach to antibiotic discovery.” Cell 180.4 (2020): 688-702.
Margulis, Lynn, and Michael J. Chapman. Kingdoms and domains: an illustrated guide to the phyla of life on Earth. Academic Press, 2009.
https://ncats.nih.gov/research/our-impact/our-impact-drug-discovery-and-development
https://www.criver.com/eureka/why-does-drug-development-take-so-long
https://www.knowledgeportalia.org/costs-r-d
Copyright 2024 AI-Talks.org