The failure rate of drugs in clinical development remains one of the most persistent challenges in drug development. Traditional models often fail during this process due to either inadequate data or inadequate tools for preparing and analyzing the data that many industry companies already possess. Organizing and structuring this critical data into a well-structured, AI-ready format is needed to uncover game-changing insights, but it can be complicated to collect or cost-prohibitive for many small or microcap biotechnology companies.  

BullFrog AI is working to overcome these traditional limitations by focusing on causal artificial intelligence that is purpose-built for life sciences rather than adapted from general analytics tools. Unlike conventional correlation-based machine learning that merely shows relationships and patterns, BullFrog’s causal AI provides the magnitude and direction of those relationships. Especially in complex biological systems, this approach provides an understanding of disease drivers and associated pathways, which is crucial for making discoveries that lead to successful drugs.  

Drug Development & Delivery recently interviewed Vin Singh, Founder and CEO of BullFrog AI, to discuss the company’s technology, integrating AI into the drug development process, and more.  

Can you tell us a little bit about BullFrog AI and your solutions for drug development and discovery?  

BullFrog AI uses AI and machine learning to overcome a costly problem in the drug development ecosystem: stubbornly low success rates in drug development. From the start, our vision has been to help life sciences companies make better decisions earlier, reduce wasted capital, and ultimately bring more effective therapies to patients faster. There are many companies talking about AI in drug development, yet none have our core platform, which we licensed from Johns Hopkins Applied Physics Lab, and very few, if any, have our causal inference capabilities. Our bfLEAP® platform is designed to uncover critical biomarkers, disease drivers and pathways, not just patterns. That distinction matters because it leads to insights that are actionable and explainable. 

What made you and your team decide to tackle this specific problem with your AI solutions?  

BullFrog AI was born out of a seemingly obvious and costly problem in the drug development ecosystem. Despite massive advances in biology and data generation, drug development success rates remain stubbornly low. Too often, promising therapies fail late in development because earlier decisions were made with incomplete or misleading insights.cThe idea was to use AI and Machine Learning to change that equation. Rather than relying solely on correlations or traditional statistical approaches, we wanted to build a system that could analyze high-dimensional, multi-modal biological data and uncover the underlying drivers of disease and corresponding treatment response. That thinking ultimately led us to bfLEAP®.  

We then invested heavily in innovation and have built a very powerful proprietary capability in causal AI, which moves us beyond correlations and enables us to identify the drivers of disease. From the start, the vision was to help life sciences companies make better decisions earlier, reduce wasted capital, and bring more effective therapies to patients. Through considerable hard work and dedication, we’ve now completed our end-to-end AI workflow from AI data preparation, bfPREP™, to clear and defensible recommendations and decision making, bfARENAS™.   

What makes BullFrog’s solutions different from other companies focused on bringing AI to the life science industry?  

Our end-to-end solutions in this space are bfPREP™, bfLEAP®, and bfARENAS™, each of which offers capabilities at key stages in the drug development process. bfPREP™ prepares messy large-scale datasets for AI consumption, a product as a service that a lot of these companies need. This then feeds directly into our core platform, bfLEAP®, which is a system of proprietary technologies and methodologies including causal AI, designed to analyze complex biomedical datasets to provide insights into drivers of disease: uncovering patient subgroups with shared molecular signatures, flagging genes and pathways driving disease biology, and much more. What makes it different is its ability to work with incomplete, shallow, wide, and multi-modal data, which is the reality of drug development data. The platform has almost infinite scalability. bfLEAP® then feeds into bfARENAS™, which builds upon the AI analytics engine to competitively rank potential drug targets and disease drivers using criteria that are established by the client. These tools provide actionable and explainable insights that clinical teams can trust to make development decisions. This end-to-end approach, from data preparation all the way to delivering key insights that can enable better-informed decision-making at key moments in the drug development process, is ultimately what separates us from others in the market. 

Can you describe an example of how BullFrog’s solutions provided value to a clinical program using AI and machine learning?  

Absolutely. We recently completed a post-hoc analysis for an oncology-focused company, Eleison Pharmaceuticals, and their Phase 3 pancreatic cancer study. We used bfLEAP® to analyze an incredibly large, complex, and messy dataset, allowing us to identify biomarkers and a patient subgroup that showed an almost three-fold improvement in overall survival relative to patients on the best supportive care. These results were presented at the American Society of Clinical Oncology’s Gastrointestinal Cancers Symposium (ASCO-GI) meeting in January of this year.  

Eleison and their partners were very happy, and they will decide what is best to do with the findings from that study, but their conclusions for the presentation were that our platform is capable of successfully identifying patient subgroups within existing clinical trial data. We hope those insights can be leveraged to make a difference for patients suffering from such a horrible disease. 

What does optimal AI-enhanced clinical development look like in your mind?  

We recently announced our new capability, bfARENAS™, which slots into our end-to-end solution that includes bfPREP™ and bfLEAP®, resulting in a scenario-based design engine that can help define clinical trial strategies and build diversified, risk-balanced R&D portfolios. We’re very excited to see how some of our existing and new clients will be able to apply this new solution to enable robust recommendations and highlight clear pathways for strategies with a higher probability of success. In our estimation, there are not solutions similar on the market, being able to provide explainable and actionable recommendations and insights instead of weighted spreadsheets and scorecards that can give you multiple top choices.  

More broadly, AI integration is only going to become more accepted in this field, but it’s important that we’re doing so with patients in mind. We can’t just be applying AI for AI’s sake. At BullFrog AI, we really want to help educate the industry about how to properly prepare the right data to enable clinical trial teams to gain critical insights and improve drug development success rates. The goal must remain improving and saving lives with better outcomes for patients. 

What are the biggest obstacles to broader AI adoption in the drug development industry?  

One of the largest barriers to using AI in life sciences is data readiness. Many organizations struggle with scattered, inconsistent, or error-prone data. This is why we created bfPREP™. bfPREP™ is designed to clean, harmonize, and structure biological and clinical data so it can be effectively used by AI systems. 

Another hurdle in this space is adoption. AI has enormous potential in drug development, but many organizations are still early in their AI journey. They may not have the infrastructure, data practices, or internal alignment needed to fully leverage these tools. That’s why we place so much emphasis on explainability and data preparation. Trust is essential, especially in regulated industries like healthcare. 

Lastly, the broader biotech market environment can be a significant challenge. Capital has become more selective, and companies are under increasing pressure to demonstrate efficiency and capital discipline. At the same time, this environment actually highlights the value of our platform. When success rates are low and resources are constrained, tools that help teams make better-informed, faster decisions become even more critical.