Bio-IT 2026: Bridging the Gap Between Data and Insight

The promise of precision oncology has always been deceptively simple: matching the right treatment to the right patient at the right time. For decades, we were unable to deliver on that promise, not because we lacked biological data, but because the infrastructure to connect these elements with clinical decision-making in a coherent, reproducible, and timely way didn't exist.

Fragmented by Design, Costly by Default

Modern oncology generates extraordinary data, from whole-genome sequencing, spatial transcriptomics, single-cell RNA profiles, methylation landscapes, clinical registries, tumor microenvironment, and beyond. The modalities have multiplied faster than our ability to synthesize them and the result is a field that is simultaneously data-rich and insight-poor.

The workflow bottlenecks are well-known to anyone working in this space. Variant interpretation pipelines produce probabilistic outputs that rarely connect cleanly to clinical evidence. Tumor microenvironment analyses require manual annotation that doesn't scale. Literature review, still the backbone of clinical evidence synthesis, is time-consuming, inconsistent, and increasingly outpaced by publication volume. Furthermore, clinical trial design continues to rely on correlation rather than necessary causal inference.

“The field is simultaneously data-rich and insight-poor. The modalities have multiplied faster than our ability to synthesize them.”

Seven Talks. One Defining Problem

The current friction points in the field of oncology are precisely what the "AI-enhanced discovery and multimodal insights in precision oncology" session will address. I’m excited to chair this session at this week’s Bio-IT World Conference & Expo 2026, and I believe it will provide valuable insights into the latest approaches for overcoming these challenges.

This session has a lot in store, including:

The tumor microenvironment is finally becoming computationally tractable via a containerized deep learning pipeline (Sandeep Singhal, PhD, Associate Professor, Pathology, University of North Dakota);
Epigenomics is entering the foundation model era with the introduction of MethylFM, a transformer-based model trained on whole-genome bisulfite sequencing data to capture context-aware methylation patterns (Xiang Chen, PhD, Associated Member, Computational Biology, St. Jude Children's Research Hospital);
AI predictions are only as effective as their experimental grounding and remain probabilistic and often require further validation (Chris Dayton, CoFounder & CEO, Quality Assured AI and Marianna Weener, MD, PhD, Senior Researcher, Broad Institute of MIT and Harvard);
Systematic curation continues to outperform standard models like ChatGPT and OpenEvidence (Joe Jacher, MSC, CGC, Genomenon); Causal AI is redefining clinical trial design (Raviv Pryluk CEO & Co-Founder, PhaseV Trials, Inc);
Single-cell pathway analysis is maturing into a clinical tool (Joseph Pearson, Director, Global Product Management OmicSoft, QIAGEN); and
AI-driven molecular tumor boards are best built with the human-in-the-loop framework so to deliver measurable outcomes (Sanjay Jaiswal, Principal, Data & AI, Ernst & Young LLP).

From Data Generated to Decision Ready

What unifies every talk in this session is not a shared technology or a shared disease area. The focus is on a shared problem: the gap between generating biological data and producing credible, reproducible, and actionable insights that can change research outcomes or clinical decisions. This insight generation problem is the defining challenge of precision oncology today. While the data, compute, and AI capabilities are present and maturing rapidly, the field must now build the framework, validation infrastructure, and integration layers necessary to connect raw data to decision-ready conclusions.

“This insight generation problem is the defining challenge of precision oncology today.”

This is the work that matters. Not faster pipelines for their own sake, but pipelines whose outputs researchers and clinicians can trust, replicate, and act on. Not AI tools that replace biological expertise, but AI systems that amplify it and that surface the evidence a human might miss, validate the prediction a model alone cannot confirm, and synthesize across modalities in ways no individual researcher has the bandwidth to do manually.

What Getting This Right Actually Looks Like

The next couple of years in AI-enhanced precision oncology will be defined by three critical shifts from:

Probabilistic outputs to experimentally validated and clinically credible findings
Siloed data to integrated, harmonized analysis frameworks
Research-grade tools to production-ready systems

This is not a future state, rather it is what gets built right now, proving that the field has reached a significant inflection point. The challenge is no longer AI’s capability, but whether our infrastructure and institutional trust can keep pace with computational possibilities.

I will address these challenges in my talk “More Data, Less Insight: Why Scientific AI Needs a New Framework,” during the AI-Enabled Diagnostics and Multimodal Biomarkers session, in the afternoon of May 20. While the multi-omics data stack in drug discovery is richer than ever, pharma and biotech teams still struggle to generate defensible conclusions at the speed programs require. The talk will demonstrate how Eos, Mithrl's Scientific Decision Engine, runs validated multi-omics analyses on demand, orchestrates pipelines, QC, and statistical methods, while ensuring every result is reproducible, traceable, and auditable. By compressing time-to-insight from weeks to hours, Eos enables discovery teams to make go/no-go decisions with greater confidence.

The Conversation Continues at Booth 802

For those interested in exploring our solution, whether you're a computational scientist or a bioinformatician managing pipelines, an informatics leader looking to embed AI rigor into your discovery stack, or a research team seeking to accelerate analysis timelines, then please stop by the Mithrl booth, #802. We'll be providing live demonstrations of Eos, discussing real-world use cases from drug discovery programs, and engaging in the technical conversations that often exceed the limits of a standard presentation slot.

“The insight generation problem is solvable. We're building the solution, and we'd like to show you how.”

Session Talk: “More Data, Less Insight: Why Scientific AI Needs a New Framework”

Presenter: Vivek Adarsh, Co-Founder & CEO, Mithrl

Time/date: May 20, 2.30PM

Location: T7: AI-Enabled Diagnostics and Multimodal Biomarkers

Booth: #802

Demos: Daily (with free ice cream - we are excited to host you)

Precision Oncology Is Rich in Data but Poor in Insight. Will AI Finally Bridge the Gap?

Fragmented by Design, Costly by Default

Seven Talks. One Defining Problem

From Data Generated to Decision Ready

What Getting This Right Actually Looks Like

The Conversation Continues at Booth 802