Some of the largest blockbuster therapies – including Keytruda in oncology, Myozyme in rare diseases, and Ozempic – were overlooked for years or even decades during early research.

At Explority, we trained the first Large Language Models (LLMs) capable of predicting the probability of success in clinical trials for early-stage therapies – identifying therapeutic programs with the highest success potential:

• For commercial drug development by R&D teams

• For deal sourcing by VC funds and business development teams

The core problem: We make billion-dollar decisions with limited predictive power

Various AI models are now widely used to help design and discover new drugs. However, until now, there have been no AI systems capable of evaluating and prioritizing early-stage therapeutic programs in the context of initial research results and decades of historical clinical successes and failures.

Yet the most critical decisions for pharmaceutical companies, biotech startups, and investors are:

• Which new R&D opportunities to pursue

• Which biological targets to prioritize

• Which startups and therapeutic concepts to invest in

• Which assets to license or acquire

A failed drug discovery program can cost hundreds of millions of dollars and years of scientific effort.

Why Explority AI changes the equation

LLMs are extremely powerful at pattern recognition, with out-of-the-box models already surpassing human experts in predicting the most probable research outcomes (Luo, X. et al.). However, recognizing the patterns associated with long-term clinical successes and failures required specialized training on a uniquely structured dataset.

To create this dataset for Explority AI, we processed and organized 1M+ drug discovery research papers across 5,846 rare diseases by linking preceding publications with downstream clinical and regulatory outcomes.

On top of this structured foundation, we train LLMs as classifiers to recognize patterns in the early results of therapeutic programs that later succeed in clinical trials versus those that do not, with top-ranked therapies demonstrating a 4× higher trial success rate than the average preclinical drug (How Explority AI is trained and validated).

Benefits of AI-driven probability of success assessments

Assessing probability of success is especially valuable in early drug development, where weak signals are scattered across thousands of scientific papers, datasets, and biological hypotheses.

Rather than replacing scientific expertise, Explority AI augments it by transforming fragmented biomedical knowledge into structured, actionable predictions.

From the first scientific publication to orphan drug designation by the U.S. Food and Drug Administration or European Medicines Agency, Explority AI tracks research papers across 5,846 rare diseases, highlighting new therapies with the highest potential for clinical success and helping ensure that promising treatments are recognized and prioritized earlier and with a better ROI.