In autoimmune drug development, uncertainty is the norm. Each program demands enormous time, money, and patience - often more than a decade and over $2 billion to bring a single therapy to market. Yet most never make it past Phase 2.

Why? Because early translational decisions are made with incomplete information. The biology is complex, patient symptoms are subjective, and data lives in silos - spreadsheets, PDFs, registries, trial reports, that rarely talk to each other. What should help scientists learn faster often ends up slowing them down.

That’s the challenge we explored in our recent webinar, De-risking Autoimmune Clinical Trials with Agentic AI, led by Sachin Kumar Gupta from Elucidata and Dr. Devan Moodley, a Translational Scientist based in Boston. Together, they showed how Agentic AI systems can make clinical trials not just smarter but more human - by connecting molecular biomarkers with the voice of the patient.

Listening to What Data Doesn’t Say

Every clinical trial collects two kinds of signals:

• Biomarkers - objective, lab-based measures such as CRP, IL-6, or TNF.
• Patient-Reported Outcomes (PROs) - how patients describe fatigue, pain, mobility, or daily function.

Traditionally, the first group drives decision-making, while the second sits quietly in the background. But ignoring PROs means ignoring the patient’s own window into disease progression and treatment response.

“The insight we could gain from how patients actually respond to treatment is often lost in translation,” said Dr. Moodley.

He’s right. These reports are unstructured - buried in PDFs or surveys - and every study measures them differently. Without structure, there’s no way to compare fatigue in one trial to fatigue in another. The result is an underused data asset hiding in plain sight.

Why Autoimmune Trials Are the Perfect Test Case

Autoimmune diseases such as lupus and rheumatoid arthritis are notoriously complex. No two patients look alike. Symptoms fluctuate, biomarkers overlap, and every dataset seems to speak a different language.

If an AI system can bring order here - linking what patients feel with what biologists can measure, it can work in other diseases and conditions too.

That’s why we tested Polly Xtract, Elucidata’s Agentic AI capability, on autoimmune data. The goal was simple: turn messy, scattered documents into structured, traceable evidence that connects biomarkers and PROs in a single view - helping make trial protocol development better and faster.

What Polly Xtract Revealed

When Polly Xtract extracted and analyzed autoimmune studies, clear patterns surfaced:

• In rheumatoid arthritis, HAQ-DI (a measure of physical function) consistently tracked inflammation markers like CRP and ESR.
• Fatigue scores correlated with IL-6, a cytokine central to disease activity.
• Some PROs changed weeks before biomarkers did - offering early signs of treatment response.

These insights show how linking the biological and experiential can reshape trial design. Instead of waiting months for lab results, teams could use validated PROs to make faster, patient-friendly decisions without compromising scientific rigor.

Making AI Evidence That People Can Trust

One of the biggest barriers to adopting AI in translational research is trust. Regulators and clinical teams want to know where each number comes from.

“The FDA and other regulators are ready for it. What they want is transparency - an ability to see what’s under the hood,” said Dr. Moodley.

Polly Xtract was built with that in mind. Each extracted value is stored with typed metadata and a resolvable citation that points to the page, table, or figure it came from. Ambiguity isn’t hidden, it’s surfaced and clearly marked for expert review.

That combination of speed and transparency makes AI adoption practical in R&D, not theoretical.

Efficiency Is the New Innovation

Manual curation of 20–30 clinical publications can take weeks or months. Polly Xtract can process the same volume in hours, producing structured outputs ready for modeling or submission.

“The fact that I can go from six months of work to two days of work, and do it reliably, is hugely valuable,” said Dr. Moodley.

Faster analysis means earlier go/no-go decisions, lower R&D costs, and more capital-efficient programs. It also lets scientists focus on asking better questions instead of wrestling with file formats.

The Broader Lesson

Agentic AI isn’t here to replace scientists. It’s here to help them see connections that were previously invisible - between the biological and the experiential, the measurable and the meaningful.

When unstructured data becomes AI-ready data, the patient’s voice stops being anecdotal and becomes part of the scientific record. Trials become not just faster, but more humane and decisions become grounded in real evidence rather than assumptions.

As Sachin Gupta summarized:
“Patient data that was previously overlooked can now drive faster, cheaper, and more patient-friendly trials. PROs make the patient voice a measurable scientific variable, not an afterthought.”

‍