Whole genome sequencing is becoming widespread, yet access, usability, and data ownership remain controlled by sequencing providers. But that doesn’t have to be. We’ll introduce novel architectures and tools that enables individuals to query their complete genomic data using natural language—directly from their personal devices, without uploading sensitive information to third-party servers. Users can ask health questions like “Do I have pathogenic variants related to hemochromatosis?” or questions related to kinship or traits. By shifting genomic analysis to the edge—leveraging modern mobile computing power and AI-driven natural language interfaces—we return data ownership to individuals while maintaining the analytical sophistication of centralized platforms.

Therapy-resistant prostate cancer requires sustained protein synthesis, but the regulatory bottlenecks that maintain this state remain poorly defined. Using AI-assisted integration of subcellular RNA-seq, quantitative proteomics, and patient cohort analysis, we identified ribosome export as a clinically relevant control point linking nuclear transport to oncogenic translation. XPO1 inhibition caused broad nuclear-cytoplasmic remodeling, whereas NMD3 depletion selectively constrained pre-60S ribosome availability with minimal transcriptomic disruption. By extending conventional differential expression analysis with compartment-resolved RNA-protein concordance mapping and residual modeling of protein output beyond RNA abundance, our framework prioritized ribosome export as a precision therapeutic vulnerability in advanced prostate cancer.

Based on the concept of competing endogenous RNA (ceRNA), this talk presents a new framework to analyze the potential contributions of microRNAs in response to ICI therapy in mesothelioma. We first constructed mesothelioma ceRNA networks based on two largest sets of publicly available gene and microRNA expression data and six public microRNA databases. We then used these networks to ascertain the contributions of differentially expressed microRNAs in mesothelioma patients’ survival outcomes with ICI.

Metabolic dysfunction and insulin resistance remain central challenges in cardiometabolic disease research, demanding integrative frameworks that bridge biology and bioinformatics. Leveraging a microphysiological liver model and bulk RNA-Seq, we developed a Nextflow-based bioinformatics pipeline integrating batch correction and statistical power modeling to ensure reproducible and interpretable results. This approach operationalizes design and computational rigor, transforming transcriptomic data into actionable insights to advance understanding of metabolic regulation and therapeutic targets.

Novel 3D IF (immunofluorescence) imaging technologies are shifting analysis of the tissue microenvironment from low-resolution 2D to high-resolution 3D approaches, allowing study of cellular interaction in unprecedented detail. This talk presents progress from displaying and analyzing spatial relations in tabular (2D) data towards scalable 3D multi-volume visualization and detailed image-based analysis of cellular features and cell-cell interactions. I will showcase different approaches and tools on healthy and cancerous tissue data.

• How are knowledge graphs supporting reliable scientific intelligence
• Lessons from early deployments
• Remaining challenges and barriers to scale​

• Learn how leading global pharma companies are applying quantum principles to real-world research. 
• Discover the quantum companies transforming traditional processes around drug development and drug discovery.
• Discuss how quantum, along with HPC and AI, is poised to help researchers tackle historically intractable problems and potentially find new treatments for diseases like cancer, Alzheimer’s, and diabetes.​

• Where are we today? Are AI tools truly integrated into workflows, or still operating in silos?
• What changes with agents? How do LLM-based agents shift drug discovery from prediction to decision-making?
• What is blocking autonomy? Data quality, validation, trust, or organizational readiness for AI-driven discovery?​

• Where are the highest-value collaboration opportunities between academia/TTOs and industry (biopharma, startups, AI/data, investors), and what makes them actually work?
• What types of partnerships are most effective in practice (licensing, co-development, startup creation, sponsored research), and how do both sides evaluate quality and fit quickly?
• What specific formats, access points, or structures at the Bio-IT event would actually drive follow-up, partnerships, and deal-making?

• How do CIOs enable rapid AI and data innovation without breaking governance, compliance, or data integrity frameworks?
• What separates pilot-stage innovation from scalable, production-grade systems? Where do most organizations fail?
• How do leaders manage emerging risks (model bias, data provenance, regulatory exposure) while still pushing competitive advantage?
  

• How real-world data is being used alongside clinical and experimental data
• Challenges in standardization, access, and regulatory acceptance
• Opportunities to improve outcomes, access, and long-term patient insights

• Why critical experimental knowledge is lost across ELNs, LIMS, and siloed teams, and the cost of reinventing work 
• How structured experimental memory makes past decisions, failures, and context reusable across teams
• How organizations can make scientific knowledge reusable for both humans and AI

• How to think strategically about IP early in biotech and AI-driven research
• Common pitfalls in patent and trademark applications and how to avoid them
• Navigating ownership, protection, and commercialization of innovations
  

• Why do current workflow systems fail most scientists in practice? 
• Can no-code and AI-driven workflows meet standards for reproducibility, validation, and clinical readiness? 
• What would a “production-ready” workflow ecosystem actually look like across teams and organizations?

• The gap between AI ambition and the reality of data foundations, platforms, and operating models
• What “AI‑ready data” actually means in practice—beyond pilots, dashboards, and proofs of concept 
• Where AI initiatives most often break down: data integration, data quality, metadata, and governance at scale​

Analyzing somatic mosaicism with DRAGEN using blood and sperm tissues, enhanced by a personalized reference genome for improved accuracy and detection. This approach will enable more sensitive detection of low-frequency variants and provide deeper insights into tissue-specific mosaic patterns.

Building scalable genomics analysis platforms traditionally requires significant infrastructure investment and specialized IT expertise, limiting access to well-funded institutions. We present a production-ready platform architecture using miniWDL and AWS Batch that enables research organizations to deploy enterprise-grade genomics infrastructure supporting complex genomics pipelines while achieving 30–50% cost reduction through intelligent spot instance orchestration. This platform democratizes genomics by providing researchers with self-service access to secure, reproducible, and cost-efficient large-scale analyses.

Current biological foundation models excel at static prediction but struggle to capture the dynamic complexity of disease. This presentation introduces an agentic AI architecture that models virtual cells by reasoning across longitudinal epigenomic, transcriptomic, and proteomic data. We demonstrate how this approach improves prediction of cellular response to perturbation and validate results against real-world clinical cohorts, highlighting a path toward more predictive, translationally relevant biological modeling.

AION Labs is a venture studio founded by Pfizer, AstraZeneca, EMD Serono, Teva, Amazon Web Services, and two venture capital firms, IBF and Amiti. In fewer than four years, the studio has launched nine companies and built a model that enables early-stage startups to co-develop cutting-edge technologies with pharma companies while refining their commercial offerings. In this talk, I will share opportunities for entrepreneurs, lessons learned from building this model, where the field is heading, and how to get involved.

Modern small-molecule drug discovery is constrained by practical screening limits, both experimental and computational. Identifying hits for novel targets typically requires extensive screening, which in practice is restricted to relatively small and chemically biased libraries. Even the most advanced computational approaches, including co-folding, do not scale to broad chemical exploration and therefore operate on libraries far below the few million compounds that represent a practical upper limit. As a result, discovery efforts stay close to familiar chemistry to control cost and timelines, limiting chemical diversity, reinforcing historical bias, and leaving many potentially better drugs undiscovered.

ProPhet addresses this limitation with an AI-driven discovery platform capable of screening billions of molecules at once, without relying on structural information or large-scale experimental screening. By making broad and diverse chemical exploration practical instead of risky, ProPhet enables teams to search far beyond standard chemical libraries without slowing development or increasing cost, opening new possibilities for targets previously considered out of reach.

In addition to scale, ProPhet provides early insight into selectivity and off-target binding, helping reduce late-stage toxicity risk and avoid unnecessary experimental investment. Together, these capabilities make a wider range of historically “undruggable” targets accessible in practice.

Here we will present the principles behind ProPhet’s screening approach, discuss its application across challenging targets, and share early validation results from ongoing collaborations with pharma, biotech, and academic partners.

Biology  has long been studied through the lens of the genome, transcriptome, and  proteome- yet a critical functional layer remains largely unmeasured.  Post-translational modifications govern how proteins actually behave: their  activity, localization, interactions, and degradation. With over 200 known  PTM types and millions of potential modification sites, this layer represents  an enormous and largely unexplored dimension of disease biology.    

Drug target discovery still fails at staggering rates. One root cause is rarely confronted head-on: the scientific literature we all depend on is a corpus of unvalidated, fragmented, and often contradictory findings. Every knowledge graph, every literature agent, every LLM-powered search tool treats published claims as equally reliable. They are not.

Human curation doesn't scale and rewards popularity over validity. General-purpose LLMs predict likely text but cannot detect contradictions or weigh evidence. Literature agents give flawed and rigorous studies equal standing. None of these approaches tell you whether a claim is actually supported by the data behind it. qed is the first platform purpose-built for evidence synthesis that scores the validity of scientific claims. The output is a scored, normalized evidence map that surfaces contradictions, quantifies confidence, and delivers probabilistic answers to therapeutic hypotheses grounded in scrutinized findings. 

Already validated with expert-level accuracy across 8,000+ labs, qed moves evidence synthesis from opinion-driven to evidence-driven, giving drug discovery teams an honest accounting of what the literature actually supports.