Leveraging Artificial Intelligence (AI) tools for Substance Use Disorders (SUD) drug discovery and development (R43/R44 - Clinical Trial Not Allowed)

Background. SUDs are a set of complex and chronic diseases with an interplay between genetic and environmental factors. In the constellation of SUDs, the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM V) provides a framework that includes ten separate classes of substances and covers 11 different criteria for their diagnosis. Currently, there are a limited number of pharmaco-therapeutics for either tobacco use disorder (TUD) or opioid use disorder (OUD) approved by the Food and Drug Administration (FDA). Additionally, there are still no approved therapies for other SUDs, such as stimulant or cannabis use disorders, despite the current understanding of neurological mechanisms and treatments being clinically evaluated. The costs and time to develop a new molecular entity (NME) from the discovery phase to a successful registration have been increasing steadily over the years. Multiple stakeholders, including government, academia, and the pharmaceutical industry are searching for ways to increase efficiency in research and development (R&D) to save time and money in drug development. Non-traditional drug development approaches and novel predictive analyses may provide additional venues to create innovative insights leading to discovering novel targets, treatments, and mechanisms of action to treat SUD. Artificial Intelligence (AI) has become central to a growing number of businesses and incorporated in their business models and has led to the acceleration and advanced development of the applications, such as AI-related powered chatbots, eCommerce, Communication, manufacturing, cybersecurity, surveillance, and smart technologies. Such growth results from several technological developments, namely, increased computational capacity, improved and advanced neural network algorithms, and more available diverse, complex sets of data (“Big Data”). AI technologies can extract concepts and relationships from Big Data, learn independently from the data patterns, and significantly augment human capabilities. AI technologies include computer vision, robotics, machine learning (ML), including deep and reinforcement learning, and natural language processing (NLP). In health care, AI is being investigated to analyze patient Big Data, such as electronic health records (EHR) of historical treatments and current care of patients to create more effective and better patient outcomes along with identifying new diagnostic tools and novel analyses. Recent partnerships between advanced computational approach-based companies and pharmaceutical companies are utilizing AI technologies to improve therapeutic research and development from early discovery to clinical trials. In 2020, the notice NOT-OD-21-011 demonstrated the continued National Institutes of Health (NIH) and National Science Foundation (NSF) interest in the growing area of AI technology use to improve science and public health. Adopting AI and related technology development may allow researchers to take advantage of such untapped advancements and identify complex patterns in the consumer and environmental data relevant to SUD drug development. Applications of AI-related tools to drug development are being increasingly investigated in all stages of the therapeutic pipeline, ranging from medicinal chemistry to clinical trials, with the ultimate goal of improving the R&D efficiency. AI technologies have also assisted medicinal chemistry areas through computer-aided novel drug design for the NME characteristics such as the “Lipinski” rules. In silo screens have been performed when the three-dimensional structural data of a protein target can be utilized in drug-docking studies. With the observation of novel & multigenic targets for common disorders, traditional one-disease-one-target dominance is being reconsidered for polypharmacology-based therapeutics. The polypharmacology strategy can also include AI-related predictions of adverse drug reactions and computational toxicology. AI technology use can accelerate drug repurposing of clinical-stage therapeutics, which have the advantage of moving directly into clinical trials, thereby saving more time and money. For “First in class” NMEs by either target-based or phenotypic drug discovery, researchers found that phenotypic screening yielded significantly more of these novel NMEs. While the phenotypic approaches have the advantage of unbiased target investigation, this approach does require the extra step of target deconvolution analyses where AI technologies can assist in image analyses and feature analyses. In the investigation of complex or unmet disorders, the “omics”-based analyses can yield new drug targets and new biomarkers. Additional areas where AI technologies are in use include reviewing biomedical research literature for many important areas (e.g., reproducibility, human evidence, target identification, data repositories for genetics, biomarkers identification, clinical trial design, patent searches). The incorporation of AI-related technologies has had recent successes and is predicted to accelerate traditional and innovative areas of SUD drug development. In 2020, the first reported AI-developed therapy is being used in clinical trials to treat obsessive-compulsive disorder (OCD) patients. The development time for this AI-based drug development was 12 months, which is significantly shorter than the routine exploratory research phase of drug development that can take up to five years. This “proof of concept” exemplifies that AI-based technologies can drastically save many years spent on drug profiling and hold the potential to cut down the expense of the drug development process as well. AI can be leveraged to develop a set of novel biomarkers and secondary outcomes for clinical SUD studies. In addition, AI technologies may analyze the basic science of SUD models and patient data separately, then facilitate the linking of this data with clinical outcomes to help optimize personalized treatments of SUD. These approaches hold the additional promise to decrease drug development cost and time by providing quantitative values to assess candidate molecules' efficacy, even before performing laboratory experiments or pre-clinical tests. AI platforms may help normalize and annotate Big Data resources, including hundreds of millions of human biological, pharmacological, and clinical data points. The development of AI network-based algorithms and the analyses of these data may identify precision medicine approaches and repurpose existing NMEs for SUD indications. AI-related predictions of personalized medicine (e.g., identifying patients who would potentially benefit more from the drug or experience a more significant side-effect profile) may be particularly useful in this context. Research Objectives. This Funding Opportunity Announcement (FOA) aims to develop AI technologies specifically for the SUD drug development, addressing knowledge discovery, drug discovery, drug testing, drug repurposing, and clinical trial designs that can be subject to automation. Applications focusing on alcohol use disorder or pain as the primary indication will not be supported under this National Institute on Drug Abuse (NIDA) funding opportunity. Areas of interest include, but are not limited to, the development and utilization of AI-based tools in the following areas for SUD drug discovery and development: 1. Drug target identification and validation: (a) AI-driven tools and platforms that automate the transformation of diverse streams of biomedical and healthcare data, such as longitudinal EHR, next-generation sequencing, and other “-omic” data into mechanistic computer models to be representative of individual patients. (b) AI models using biomedical and clinical data that can draw novel insights about drug candidates and model the whole biological systems to help identify novel pathways, targets, and biomarkers. 2. Target-based and phenotypic drug discovery: (a) AI-related systems that can design new molecules employing phenotypic, high-content screening data along with assessing their potency, selectivity, and binding affinity. (b) AI-based modeling at each stage of the process, from the hit identification and expansion through lead design/optimization to absorption, distribution, metabolism, and excretion (ADME)/toxicity predictions. (c) AI technology platforms that work with data points obtained from different studies of high-content, low-throughput phenotypic assays, high-throughput screening, structure-based design, and traditional computational methods. 3. Polypharmacology discovery: (a) AI algorithms that include data from studies of the genome, proteome, metabolome, and the lipidome of the biological samples to evaluate the complex biological networks playing roles in SUD and help identify medications for specific patient populations and sift through the drug candidates that are likely to fail. (b) AI-systems that probe vast chemical spaces and identify the most promising drug candidates. 4. Drug repurposing: (a) AI tools that analyze and visualize Big Data (e.g., biomedical, clinical, research) and identify new patterns and insights to understand how drugs work and which of them are useful in satisfying unmet medical needs in the SUD area. (b) AI based platforms utilizing Big Data resource includes hundreds of millions of human subjects, biological, pharmacological, and clinical data points, normalized and annotated to build network-based algorithms to repurpose existing drugs or identify drug candidates or identify drug targets related to SUD indications. 5. Biomarkers development: (a) AI tools that identify molecular signatures and potential biomarkers for assessing the drug response, advancing deep-learning screens of biomarkers from patient data, and “multi-omic” modeling approaches utilizing real-world biomedical data, including gene expression, protein interaction networks, and clinical records. (b) AI based models that identify potential responders to a molecular targeted therapy before the drug is tested in humans. 6. Predictions of Adverse Drug Reactions: (a) Machine learning workflow that can enhance predictions of Adverse Drug Reactions (ADRs) through drug-gene interactions. 7. Clinical Trial Design: (a) AI-based systems that correlate large and diverse datasets such as EHRs, the medical literature, and trial databases for the improved patient–trial matching (cohort composition) and recruitment before a trial starts, as well as for monitoring patients automatically and continuously during the trial, thereby allowing improved adherence control and yielding more reliable and efficient endpoint assessment. Special Considerations National Advisory Council on Drug Abuse Recommended Guidelines for the Administration of Drugs to Human Subjects: The National Advisory Council on Drug Abuse (NACDA) recognizes the importance of research involving the administration of drugs with abuse potential, and dependence or addiction liability, to human subjects. Potential applicants are encouraged to obtain and review these recommendations of Council before submitting an application that will administer compounds to human subjects. The guidelines are available on NIDA's Web site at http://www.drugabuse.gov/funding/clinical-research/nacda-guidelines-administration-drugs-to-human-subjects. Points to Consider Regarding Tobacco Industry Funding of NIDA Applicants: The National Advisory Council on Drug Abuse (NACDA) encourages NIDA and its grantees to consider the points it has set forth with regard to existing or prospective sponsored research agreements with tobacco companies or their related entities and the impact of acceptance of tobacco industry funding on NIDA's credibility and reputation within the scientific community. Please see http://www.drugabuse.gov/about-nida/advisory-boards-groups/national-advisory-council-drug-abuse-nacda/council-statements/points-to-consider-regarding- for details. Data Harmonization for Substance Abuse and Addiction via the PhenX Toolkit: NIDA strongly encourages investigators involved in human-subjects studies to employ a common set of tools and resources that will promote the collection of comparable data across studies and to do so by incorporating the measures from the Core and Specialty collections, which are available in the Substance Abuse and Addiction Collection of the PhenX Toolkit (www.phenxtoolkit.org). Please see NOT-DA-12-008 (http://grants.nih.gov/grants/guide/notice-files/NOT-DA-12-008.html) for further details.