FDA's New Measures Clarify the Pathway for AI-Driven Drug Development
Recently, at the 2018 American Society of Clinical Oncology (ASCO 2018) Annual Meeting, FDA Commissioner Dr. Scott Gottlieb stated that the FDA is attempting to implement new measures to promote more efficient drug development while increasing the market access opportunities for innovative drugs and therapies. Specifically, the FDA’s new policy encompasses various stages, including clinical trial recruitment and execution, as well as New Drug Application (NDA) submission and review. VCBeat (WeChat ID: vcbeat) has provided an interpretation of these developments.
Traditional eligibility criteria typically exclude elderly patients, those with poor physical performance, organ dysfunction, brain metastases, or other comorbidities; however, once a drug is marketed, these individuals are the most likely to receive such treatments. The outdated eligibility criteria undoubtedly restrict access to investigational drugs for patients in urgent need of treatment and have a negative impact on oncology clinical trials.
However, even with stringent exclusion criteria, the failure rate of drug trials remains considerably high. According to statistics from researchers at the Massachusetts Institute of Technology, the overall success rate for new anticancer drug trials in 2015 was only 8.3%, representing only a marginal improvement from 3.4% a decade earlier. This high failure rate drives up pharmaceutical development costs within the industry, which are ultimately passed on to the market, exerting upward pressure on drug prices.
Furthermore, outdated subject eligibility criteria have also constrained the diversity of competing pharmaceutical products. Typically, being the first to market confers a prolonged period of market exclusivity, thereby enabling monopoly pricing. If the market were flooded with a substantial number of alternative therapies, such exorbitant monopoly prices would be brought to an end.
One approach to untying this Gordian knot is to synchronize the delivery of care with the processing of the massive volumes of data generated from conducting real-world clinical trials on patients. This necessitates a systematic reconfiguration of key questions and answers within clinical trials, thereby optimizing medical care, trial design, and product upgrades.
Despite the risks associated with the expanded use of unclinically validated drugs in patients, timely and effective new medications can prevent further deterioration of a patient’s condition in real-world scenarios—for example, in diabetic patients with long-standing lymphoma and invasive disease who subsequently develop breast cancer.
Therefore, the FDA is seeking to provide additional support to clinical trial applicants to help them recruit more patients with diverse characteristics into their trials. To date, the FDA has issued guidance on the inclusion of adolescent patients in clinical trials of adult oncology drugs. According to this guidance, it is permissible to include adolescent patients in clinical trials of adult medications unless there is evidence that the investigational drug exhibits greater toxicity in younger patients than in adults.
Furthermore, even after expanding the coverage of clinical trial participants, certain human factors in drug review can still delay the market launch of new drugs. Specifically, discrepancies in the interpretation of review materials between applicants and regulatory authorities often result in documents and information that do not provide a basis for final decisions, thereby bogging down the review process. In response, the FDA is piloting an approach to gradually shift the focus of drug reviews toward data most relevant to safety and efficacy, requiring applicants to conduct substantive reviews before submitting their materials.
Specifically, when the applicant locks the database and decides to submit an application to the FDA, they will be required to begin sharing baseline data with the FDA, including key raw data and derived datasets, safety and efficacy tables and figures, study protocols and amendments, and draft package inserts.
Typically, the FDA can begin data verification and assess its adequacy and completeness while the applicant’s database is locked for 2–4 weeks. This enables reviewers and applicants to address data quality issues early through real-time interaction, thereby making the review process more efficient and thorough. Statistics indicate that this approach can free up 10% to 30% of reviewers’ time. Building on this foundation, the FDA will attempt to create a dynamic data submission system.
The Oncology Center of Excellence (OCE), established by the FDA one year ago, is currently implementing this pilot program. It employs the Real-Time Oncology Review (RTOR) program, which formats applications submitted by sponsors in accordance with the FDA’s substantive review standards for oncology drugs.
In practice, the table format specifically comprises two components: the applicant’s scope of work and the FDA’s assessment thereof. This structured table serves as a flexible data review platform, enabling the FDA to issue real-time approvals or disapprovals and incorporate new analytical conclusions. The early phase of the Real-Time Oncology Review (RTOR) pilot program focused on supplemental applications for new indications of already-approved anticancer drugs, with numerous applicants participating in the pilot. In later stages, the Office of Clinical Evaluation’s (OCE) RTOR approach may expand to include applications for innovative drugs and biologics, while also attempting parallel reviews of multiple supplemental applications.
The FDA’s new drug review policy has, in fact, imposed higher requirements on new drug development data in terms of both quality and quantity, thereby creating significant opportunities for artificial intelligence to play a pivotal role. Currently, numerous U.S.-based startups are already making attempts in this area.
A More Scientific Approach to Drug Discovery
Qrativ
Qrativ was established in 2017 as a joint venture between Mayo Clinic and nference, dedicated to integrating clinical expertise with artificial intelligence to advance new drug development. With direct access to Mayo Clinic’s specialized knowledge and clinical data, Qrativ is well-positioned to achieve significant breakthroughs in pharmaceutical R&D.
Currently, Qrativ has leveraged nference’s AI-powered knowledge synthesis platform, along with Mayo Clinic’s medical expertise and clinical data, to develop Darwin.ai, a drug therapy platform that facilitates systematic drug development, particularly for new treatments targeting rare diseases. Partners of Qrativ can utilize the Darwin.ai drug therapy platform to identify all potential applications of candidate drugs, including determining potential indications for rare diseases and pinpointing the patient subgroups most likely to respond favorably to a given candidate drug.
Engine Biosciences
Engine Biosciences is a biotechnology company that applies machine learning to genomics-driven drug discovery, with offices in San Francisco and Asia, pioneering the field of network biomedicine. Engine Biosciences’ proprietary patented technology is designed to decipher the complexity of biological networks, integrating high-throughput wet-lab experiments with AI algorithms for drug discovery and cellular reprogramming.
Engine Biosciences has garnered favor from numerous domestic and international investors, including Baidu and WuXi AppTec. Its technology has been applied in four areas: 1) drug repurposing, i.e., exploring new indications for known drugs; 2) novel target discovery, i.e., identifying biological factors that cause diseases; 3) precision medicine, which provides unique treatment plans based on patients’ specific genetic profiles; and 4) biological pathway analysis.
Cloud Pharmaceuticals
Cloud Pharmaceuticals, founded in 2014, is dedicated to improving health and well-being through computational design and the rapid discovery of novel therapies. Leveraging a proprietary design process that integrates AI technologies with cloud computing, Cloud Pharmaceuticals searches virtual molecular space and applies complex molecular modeling to design new drug candidates with development potential from the outset. This approach accelerates the drug development process at lower costs and with higher success rates, while enhancing targeting for difficult-to-treat indications.
Cloud Pharmaceuticals’ approach amplifies the power of cloud computing and artificial intelligence, fundamentally improving drug discovery and design. The company is establishing a broad-indication product pipeline that spans all stages of drug development, from initial discovery to clinical collaboration.
Numerate
Numerate, founded in 2007 and headquartered in San Bruno, California, applies novel machine learning algorithms at cloud scale to overcome major challenges in small-molecule drug discovery. In June 2014, Numerate raised $8 million in Series C funding from Atlas Venture and Lily Ventures.
Numerate’s drug design platform integrates advances in computer science and statistics with traditional medicinal chemistry approaches, while simultaneously addressing the factors that determine the success or failure of drug candidates. Numerate is leveraging this proprietary platform to develop a portfolio of therapeutic programs in cardiovascular, metabolic, and neurodegenerative diseases, with a focus on tackling challenges that are typically beyond the scope of conventional computer-aided drug discovery.
Atomwise
Atomwise, founded in June 2012, is a company that leverages supercomputers for drug research and development. By analyzing existing databases with supercomputers and employing AI and complex algorithms to simulate the drug development process, Atomwise assesses the risks of new drug candidates in the early stages of R&D. This approach reduces the cost of drug research to just a few thousand dollars and completes the assessment within days. Atomwise has achieved breakthroughs in identifying new drug targets, integrating results from binding affinity predictions and toxicity screening.
Atomwise has secured nearly $7 million in funding. Its unique advantage lies in extracting vast amounts of data spanning the entire human lifespan. By addressing the high costs and time-intensive nature of drug development, this capability tackles critical “life-or-death” challenges in the pharmaceutical industry. Atomwise claims to have achieved world-leading results in new drug discovery, binding affinity prediction, and toxicity screening.
Personalized Drug Customization
Insilico Medicine
Insilico Medicine, founded in 2014, specializes in leveraging AI technologies for novel drug discovery and lifespan extension, with a commitment to research on personal health management and aging-related issues. The company has developed a comprehensive, scalable drug knowledge management system that integrates annotated drugs, small molecules, biologics, and all other factors potentially influencing events at the molecular, cellular, and tissue levels. Leveraging its expertise in oncology and aging, Insilico Medicine utilizes individual patients’ gene expression data and signaling cloud modulation to guide drug discovery and enable targeted drug selection. PHARMA.AI, Young.AI, and NUTRIOMI are three flagship products of Insilico Medicine.
To date, Insilico Medicine has secured $8.26 million in financing. The company’s long-term goal is to collaborate with top-tier pharmaceutical companies to help analyze their drug databases and lead compounds, improve clinical trial enrollment rates, and enable accurate prediction of drug efficacy for patient groups and individual patients.
>>>>
Improving the Accuracy of Clinical Trials
Teckro
Teckro, founded in 2015, is dedicated to enhancing the speed and accuracy of clinical trials through information retrieval and machine learning technologies. Its solutions enable the instant capture and sharing of team knowledge, serving pharmaceutical companies and clinical researchers worldwide.
Teckro partners with pharmaceutical and biotechnology companies to streamline the clinical development process, which is the primary bottleneck in bringing effective and safe drugs to market. Teckro facilitates effective connections between trial personnel and the expertise they require, as well as among trial teams, sites, and patients. By enabling seamless collaboration among research teams, sites, and patients, Teckro significantly reduces the workload for clinical research associates (CRAs), medical monitors, data management professionals, and investigators.
Further Reading: