Artificial intelligence (AI) has been revolutionizing much of early drug discovery, yet human clinical trials remain a bottleneck. Now some applications of the new technologies are bringing efficiencies to clinical research.
Sophisticated modeling and simulations suggests that AI can play a major role in improving the success rates of novel drugs entering human studies. The pharmaceutical industry has long struggled with efficiency rates. According to the the Congressional Budget Office (CBO), only 14% of drugs that enter clinical trials are ultimately approved by the FDA. The development process is complex and can take many years, with many drugs failing at various stages. Estimates are that the cost of developing one drug can be over $2 Billion.
The goal of modeling software in clinical research is to create programs that can simulate expensive clinical trials. Other technical industries use extensive simulations when building highly complex items like semiconductors and fighter jets. But in some ways, biology is actually harder to model due to the higher systemic complexity.
Simulations hold a multitude of promises for drug development. By pre-emptively modeling trials and getting accurate dosing, the probability of success can be improved and predictability increased. Ultimately, the moonshot is using trial simulations to be able to conduct completely safe patientless trials. Although completely virtual trials are still probably more than a decade away, machine learning and AI have advanced sufficiently to begin to manage human biology and data better.
QuantHealth, an AI-focused clinical trial design company based out of Tel Aviv, has announced the completion of more than 100 simulated clinical trials, reporting an 85% accuracy rate. The company has developed a proprietary AI-based Clinical-Simulator system that combines over 1 trillion data points across the clinical and pharmacological domains to optimize clinical development, according to CEO Orr Inbar.
The clinical trial simulation software enables scientists to holistically model a clinical trial with thousands of variations rapidly, allowing drug development experts to evaluate parameters on the basis of endpoint success, commercial viability, and protocol feasibility. The In-Silico platform generates evidence for how therapy will perform across all clinical phases, as soon as its mechanism is known, and preclinical evidence has been established. This synthetic evidence generation engine can be used to support trial planning, as well as indication selection, drug repurposing, and in-licensing asset evaluation.
In order to address the challenge of finding large swaths of data on which to train its models, AI companies have been working with health systems to gather data to train the model and QuantHealth has spent several years acquiring, processing and analyzing data from proprietary sources. QuantHealth has partnered with OMNY which represents data from 50 provider organizations nationally including hospital systems, nonprofits, community practices, pediatric hospitals and national cancer institutes that represent 78 million patients, or a third of Americans. They have licensed over 350 million patients' data from databases to get a complete picture of each patient and added five more databases in genetics, cell biology, pharmacology, and biological cascades. The company builds a foundation using clinical trial results and FDA data, with elements of real world data (RWD), and clinical knowledge graphs and then incorporates sponsor's own additional data.
QuantHealth's AI can predict phase 2 trial outcomes with 88% accuracy (compared to the actual success rate of 28.9%), and phase 3 trial outcomes with 83.2% accuracy (versus the industry average of 57.8%) according to the company's press releases and interviews. The technology is able to predict clinical trial outcomes with significantly higher accuracy than current success rates. The accuracy would allow users to answer mission-critical questions such as trial go/no-go, cohort optimization, drug repurposing, and more.
QuantHealth performs trial design for pharma companies and fine-tunes results by changing endpoints or population to determine different outcomes. It can be used for portfolio optimization, how well new drugs will perform against competitors when considering acquisitions, and health systems asking what patients will respond to different trials and their outcomes.
CEO Inbar states that QuantHealth works with five of the top ten big pharma companies, with most clients moving from pilots to multiple trials.
Another important area AI enhanced drug development simulations is accurately predicting appropriate dosing. Drugs are first tested in animals to confirm safety. But understanding how to move from animal doses to humans continues to be a difficult.
Certara is using AI to accelerate the drug development process by technology that can seamlessly incorporate simulations with other approaches to help model dosing based on prior non-human studies.
The US Food and Drug Administration (FDA) has had a long standing collaboration with a number of Certara software licenses for reviewing new drug and biologics applications. Additionally the FDA has awarded grants to expand its predictive models for assessing drug virtual bioequivalence (VBE) and to create a formulation toolbox for topically applied drugs. These capabilities will help enable safer, faster and more cost-effective product development of both complex generics and novel drugs.
AI helps Certara to mine millions of documents and unstructured data sources in a systematic and meaningful manner and also couple data that is in the public domain with a pharmaceutical company's proprietary data to build a unique database. The AI platform mines about six million public sources, including massive regulatory databases and associated filings, memos, and scientific meetings.
Writing in Pharma Focus Europe, Amin Rostami-Hodjegan and Piet van der Graaf describe applications utilized at Certara. AI helps to manage information overload, particularly when the facts are sparse and seemingly unconnected, by going through data and extracting elements that are useful. It also gives Certara confidence in models by gathering indirect evidence that verifies the model-informed decisions. Certara uses AI to build models that are a mathematical representation of drug physiology and build a biological map. The models help address questions such as, "How long will the molecule stay in the body? Will it turn into an undesirable metabolite?"
Simulations are also used to predict clinical endpoints in discovery for novel mechanisms and to identify new biomarkers. The models can capture fundamental biology and then simulate what happens to the biomarkers, cell types, and cytokines when they put a compound into that system. This allows researchers to predict clinical endpoints for novel mechanisms earlier in the development process.
AI is providing growing support for the drug development process. Its applications range from reviewing millions of data points and gleaning relevant information, to helping build biological maps and models of new mechanisms of action and predicting clinical endpoints.
AI-based clinical simulation systems can be a game-changer for the pharmaceutical industry. The ability to predict clinical trial outcomes with significantly higher accuracy than current success rates, it has the potential to save the industry billions of dollars and years of time.