Skip to Main Content

A new algorithm, developed by machine learning company Applied XL, identifies important signals in clinical trials data. Now, life sciences professionals can access this model through STAT Trials Pulse.

What if you could be alerted when several trials in a particular class of drugs stopped early due to efficacy issues? Or see when there’s an unexpected uptick in an emerging therapeutic area? Or even be notified when key trials deviate from their regular trajectories? 

This event-based lens of assessing clinical activity is central to the new STAT Trials Pulse platform, which mines publicly available data and leverages a methodology that replicates the research process of journalists. Rather than taking a snapshot view of the clinical landscape, the system continuously tracks, dynamically categorizes, and contextualizes thousands of daily changes across multiple data sources to find the most important updates to a particular user. 

Not all of these signals are relevant, but those that are provide early indications of broader trends. Through the platform, life sciences professionals can then further refine topics, events, and trials, analyzing these data faster and more efficiently than ever before. And these feedback loops from experts and users of the platform alike is what strengthens the performance of the algorithm and improves the experience over time.

“New data footprints are being constantly created and, if properly triangulated, can provide new ways to monitor the evolution of a clinical trial — but all of that data still requires human interpretation. The machine learning algorithms in STAT Trials Pulse decide if a signal is noteworthy or not based on weights and parameters defined by journalists and industry experts who have deep domain knowledge of the space.” said Francesco Marconi, Applied XL co-founder and CEO who was previously the R&D chief at The Wall Street Journal.  

Leveraging the knowledge of experts to continuously improve accuracy

The platform is built on the belief that the most reliable information is generated by combining the editorial perspective of humans and the scale of artificial intelligence. As part of the process of algorithmic accuracy training, the technology developed by Applied XL uses a “human-in-the-loop AI system” that guarantees continuous improvement of event detection.

STAT journalists and expert sources participate in the workflow by providing feedback on machine-generated classifications. They help surface new labels or identify relationships that did not previously exist, ensuring a dynamic and editorially sound classification model. For example, defining the difference between a trial stopping early because of efficacy vs. futility, or linking the name of a drug brand name with its initial investigational drug designation.

“This isn’t about machine learning models making irresponsible predictions on the outcomes of trials. We believe in subject matter experts and their deep understanding of data domains, so we have built a platform where they can leverage the speed and stamina of machine learning systems to surface insights at scale to life sciences professionals,” explained Applied XL’s co-founder and CTO Erin Riglin.

In some cases, the community of users — who are in their own right experts — are invited to contribute feedback to increase the accuracy of the system, and as a result, improve the curation of their experience. The ultimate goal of this human-machine collaboration is to build trust and create transparency around the inner workings of AI.

Powerful, market-moving insights

The results to date have been impactful. In some instances, the tool has been able to detect signals from clinical trials data — including new studies coming online, studies unexpectedly stopping early, and even major shifts in patient enrollment — prior to major news coverage and even public announcements by the companies themselves. 

Furthermore, the classification of events allows for the identification of patterns, anomalies, and outliers in the clinical space. By deriving and aggregating proprietary events and continuously looking at these data points through a longitudinal lens, detection algorithms are able to surface otherwise invisible trends in the clinical landscape. 

Try it out yourself

STAT Trials Pulse is now available to any professional who wants to accelerate their decision-making. Early adopters can expect many new features in the near future including data visualizations, company profile pages, drug pipeline breakdowns, as well as new event types.

Get access to your four-week free trial at