2026 AGENDA
Basel, 6 - 8 October 2026
Schedule
Create your personal agenda –check the favourite icon
How AI is changing for drug discovery at Novartis
Semantic Layer or Silent Failure: Evidence-Based Rules for Grounding Agentic Workflows in Drug Discovery
The use of agentic AI in R&D
BioTechX Connect
1 hour. For Partners. Optimal Efficiency
A dedicated power-hour of pre-scheduled 1:1 meetings designed to solve specific challenges
- The format: 4x15 minute meetings
- The Goal: Rapid sourcing and partnership vetting
- The Match: 100% double opt- in (AI-Powered)
Chair's remarks
Chair's remarks
Chair's remarks
Chair's remarks
Chair's remarks
Chair's remarks
Chair's remarks
Chair's remarks
Chair's remarks
AI in Biotech: From Ambition to Value
An AI powered “neural system” that enriches operational decisions in Clinical Development
Beyond Data Models: Building AI-Ready Research Data Products at Scale
Building an AI Agent on top of small language models to automate Molecular Dynamics simulations
Connecting Life Sciences Data for Better Investment Decisions
Donor-derived cell-free DNA monitoring for personalized immunosuppression in transplantation
From Data to Design: Building Platforms for Medicinal Chemistry
Quality of life as a real-world endpoint
Using ML-enabled multi-omics bioinformatics to accelerate immunotherapy development
Breaking Barriers in Virtual Hit Identification: High Hit Validation Rates from Low-cost Exploration of Highly Synthesizable Ultra-Large Virtual Libraries
Virtual chemical space consisting of tens of trillions of molecules is now readily available. However, searching such spaces to identify biologically relevant hits and ensure diverse coverage of the space is not a simple task. As a result, much of the chemical space can often be left unexplored. To facilitate exploration of such large space, 2D searches are frequently utilized, due to computational restraints, but this is misaligned with the fundamentally 3D-nature of molecular recognition.
A range of technologies are now emerging that utilize advanced hardware, machine-learning and AI to rapidly explore ultra-large chemical spaces. Herein, we present low-cost workflows that combine the virtual exploration of our highly synthesizable, multi-trillion virtual library with rapid automated synthesis (2-4 weeks). In collaboration with our clients, such approaches have delivered outstanding confirmed hit rates (> 50%) following biological testing.
Enabling AI for modern drug discovery: from multimodal data to target and further
The Stack Beneath the Insight: How Raw Data Becomes Drug Program Insight with Data Products, a Knowledge Graph, and Agentic AI
In 2026 access to frontier AI is becoming a commodity. Now, the difference lies in how well the model is operationalized to answer questions sparking from drug programs. This is where every decision begins with evidence that was already recorded, omics, spatial, imaging, safety, trials, decades of literature, yet stays scattered, so each new scientific question restarts the same integration effort and ends in an answer no one can fully trace. This talk follows a single question on its journey from raw data to a defensible decision, told as a stack in which each layer earns the one above it.
The foundation is a data layer far beyond a pile of stored files. LLM-assisted ETL and metadata curation turn spatial, single-cell, imaging, omics and clinical inputs into FAIR, ontology-grounded data products, cutting manual harmonization by roughly 75%. These products don't sit in isolation: a biomedical knowledge graph is the connective core, a governed knowledge layer acting as a single source of truth for semantics and provenance, placing proprietary mechanisms in a broad biological context.
Over that foundation, an agentic AI layer, a supervisor coordinating MCP-connected specialist agents, quality-gated by a reviewer agent and deployed for multiple data consumers plans, reasons and justifies its way through multi-step questions. The result is a Drug Program Navigator: a queryable system compiling many orthogonal, source-cited lines of evidence you keep interrogating.
We ground it in two deployments: knowledge-graph-driven indication expansion, where GraphML surfaced non-obvious drug–disease–target links for validation; and multi-evidence target selection, adopted by 100+ scientists and accelerating 20+ programs. The shift is from "we need more data" to "we can trace, trust and act on what is in the data."
From Causal Knowledge Graphs to Cell-Level Predictions of Biological Activity in Single Cell and Spatial Datasets
Single-cell and spatial transcriptomics can resolve biological states at unprecedented granularity, but most analytical approaches depend on predefined features, such as marker genes, computationally-defined cell groups, or other fixed annotations, to define cell states. This constrains discovery to what's already been named, flattening the heterogeneity these technologies are designed to reveal. To address this, QIAGEN IPA single cell analysis includes Activity Prediction, a computational method that infers the predicted activity of thousands of upstream regulators, canonical pathways, and biological functions directly at the individual cell level, capturing context-dependent regulatory logic that feature-dependent scoring methods miss. The Activity Prediction embedding model is built from signed, directional causal regulatory facts curated within the QIAGEN knowledge base, generalized across cell types and conditions without requiring pre-defined marker sets or annotations. Because the Activity Prediction method is not signature-based, it complements group-based or clustering analyses, revealing populations, transitions, and regulatory dynamics that remain invisible to methods built on labeled features and predefined computational groupings. Applied across thousands of curated single-cell and spatial experiments, as well as customer-uploaded datasets, this approach surfaces cell subpopulations that elude conventional clustering, uncovers regulatory dynamics that predefined features can't capture, and reveals testable hypotheses for downstream analysis and wet-lab validation.
Increasing engagement with digital health
Making agentic workflows in drug discovery dependable with chemistry, pharmacokinetics, and safety data connectors
Reliable drug discovery agents require structured, harmonized, domain-aware data services that make scientific retrieval and reasoning repeatable. This presentation describes an approach combining controlled taxonomies, data science, domain expertise, and engineering to retrieve, normalize, and package scientific information for agentic workflows across chemistry, pharmacokinetics, and drug safety — with a focus on making these services repeatable, auditable, and suitable for scientific decision-making.
We demonstrate this through a pharmacokinetics sub-agent that connects chemical information with evidence from Elsevier's discovery, preclinical, and clinical sources, harmonizing results into a form agents can use consistently — showing how integrated data services enable agentic workflows grounded in heterogeneous evidence rather than unstructured retrieval alone.
Scaling the Complexity: Data Infrastructure Challenges for Macromolecules and Monomer Libraries
The AI-Readiness Gap: Why Most Lab Data Isn't Fit for Machine Learning (and How to Fix It)
Across biotech and pharma R&D, AI and machine learning pilots routinely stall — not because the models are wrong, but because the underlying lab data was never built to support them. This session lays out a practical framework for assessing and closing the AI-readiness gap in LIMS, ELN, and instrument data, so AI investments actually return value.
The Foundation for Trustworthy AI: From FAIR Data Principles to Enterprise-Ready AI in Pharma
Pharma faces a "scientific content crisis" due to poor data quality, non-FAIR data and governance gaps, hindering AI and regulatory compliance. This session offers an actionable blueprint: using FAIR principles for policy foundation and knowledge graphs for the operational layer. Attendees will hear strategic recommendations: investing in data infrastructure, adopting neuro-symbolic architectures to eliminate hallucinations, and building for machine actionability for audit-ready, trustworthy AI.
There’s no AI in 'Silo': Decentralized Clinical Trials, from Patients to Nations
The decisive advantage in Precision Medicine will not belong to whoever holds the largest data silo or trains the largest frontier model. It will belong to whoever maximizes access to patient data. Most of the industry is investing as if the opposite were true.
The evidence is already on the record. Sequencing costs have fallen five orders of magnitude in two decades, yet the cost of assembling a usable cohort has risen; in the UK, per-patient trial costs nearly tripled between 2018 and 2023. Data has never been cheaper, while accessible and genuinely usable datasets have never been more expensive. It took a decade and hundreds of millions in public funding to build the most utilized biobank, and over a billion in private capital to create an industry leader in precision oncology, because the hard part was never the data, it was the governance layer around it.
The same logic exposes the AI race: a proprietary model trained on datasets everyone can license is a commodity. A $300M deal for a consumer-genetics database was renewed for a fraction; the data was abundant, the signal was not. The window for a genuine solution has been open for a decade, especially for the well-funded digital health industry.
This talk sets out where durable advantage in health data research actually accrues, connecting the dots from persons to health systems, from patients to nations, and confronts the choices the industry has avoided to make precision medicine a reality for all patients, not just those who happen to sit in the right data bucket.
Unlocked & Validated: How Human-in-the-Loop LLMs Transform Unstructured EHR into High Quality RWD
Real-world data (RWD) is essential to biopharma R&D, but critical signals often remain locked in unstructured data types such as clinical notes, out of reach from standard analytics. Automation alone cannot reliably extract these features; success requires context, and context only emerges when clinical expertise is embedded throughout such workflows.
To capitalize on the phenotypic depth of its EHR-derived RWD, NashBio built a multi-layer LLM extraction system designed around clinical experts who informed extraction criteria, guided prompt and workflow refinements, and evaluated model output against source records to improve capture of clinically meaningful events. Applied to a 2,800 patient inflammatory bowel disease (IBD) cohort, the pipeline surfaced treatment response outcomes from each patient’s IBD clinic notes, spanning a predefined list of 25 medications. The result was more than 58,000 structured medication-response assessments, each substantiated by a verbatim quote from the attending healthcare provider and reviewable in context. This human-in-the-loop architecture achieved >90% accuracy on sampled review and 96% reproducibility – a level of rigor typically reserved for manual chart review – delivered at scale to advance more personalized medicine. Automation alone also missed important documentation patterns unique to specialty care and institutional practice.
NashBio's experience challenges the perception that AI eliminates the need for humans. This work reinforced the fact that, as extraction systems scale, human-in-the-loop matters more, not less; it is what keeps accuracy and context intact. We will also demonstrate how this framework extends to other applications relevant for biopharma R&D, including hepatology feature extraction and biomarker curation.
AI in Model-Informed Drug Development (MIDD)
Applications of biological foundation models to accelerate clinical trials
Evidence generation in early-stage drug development
From unstructured multi-center distributed healthcare data to Virtual Health Twins
Generative AI in service of data life cycle management
Human Gut Microbiome Health redefined
Predictive & Generative Modelling in Drug Design Opportunities and Challenges of Augmented Intelligence
Programmable medicine with digital twins
The Power of an R&D Unified Data Catalog with Domain Models: Accelerating Insights and Enhancing Data Governance
The Transformation Tax: Why Human Infrastructure Decides AI's Fate in Pharma R&D
Accelerating Drug Discovery with AI: Hybrid AI Innovation Across Public and Private Environments
Pharmaceutical and biotech organizations are leveraging Hybrid AI innovation across public and private environments to accelerate drug discovery by enabling secure, scalable access to complex scientific and clinical data. This distributed approach, combined with Hybrid AI capabilities that blend centralized model training with local AI inferencing for low-latency decision-making, helps shorten time to insight through faster analytics, improved data integration, and seamless collaboration across research and development teams, while maintaining data security, privacy, and regulatory compliance. By extending these capabilities across the value chain, biotech organizations can optimize discovery and development workflows through predictive modeling and real-time data analysis.
The Hybrid AI approach enables a balance between centralized intelligence and localized inference, improving responsiveness, compliance, and efficiency across environments. Together, these improvements strengthen operational agility and precision, ultimately helping organizations deliver measurable ROI across the drug development lifecycle.
Accelerating Hit Identification with AI
Augmented AI: Building Contextual Intelligence with Knowledge Graphs
Large Language Models are only as good as the context they receive. In this keynote, discover how knowledge graphs provide the semantic foundation for enterprise AI: powering GraphRAG, improving reasoning, reducing hallucinations, and delivering more accurate, explainable results. Through real-world examples from the pharma and life sciences industry, learn how organizations combine LLMs with knowledge graphs to accelerate research, enhance decision-making, and unlock greater value from connected data.
Beyond Trials: Engineering an AI-Driven Self-Synchronizing Clinical Data Hub
Every clinical data platform claims to unify ingestion, validation, and transformation. Few are engineered so that a change anywhere in the system automatically propagates wherever it is needed. This session presents a Clinical Data Hub built on that principle: a single, versioned platform that connects the entire data lifecycle, ingesting data from clinical systems and distributing standardized, audit-ready outputs to analytics environments and CTD submission packages, ultimately accelerating submissions to health authorities and supporting faster approvals.
One area where the challenge is particularly acute is vendor management. Every clinical operations team knows the cycle: a protocol amendment triggers a Data Transfer Agreement (DTA) update, which should in turn trigger a pipeline change, yet this process is rarely automated or reliable. Our solution closes that gap by linking automated vendor onboarding and DTA authoring directly to pipeline configuration, ensuring that contractual changes are reflected in data ingestion behavior with full versioning and an auditable trail. Because this underlying pattern, contracts that must drive downstream operational processes, is not unique to clinical operations, the architecture delivers measurable value far beyond the clinical trial environment.
Developed and engineered by EPAM, the solution combines deep pharmaceutical clinical operations expertise with award-winning software engineering, a combination designed to outlast the current generation of point solutions. Built-in semantic ontologies provide the platform with data integrity and continuity capabilities that legacy systems, many of which are now candidates for retirement, were never designed to support.
Data foundation and semantic layer for AI
Designed to bind, built to fail: closing the antibody developability gap
Evaluating digital health opportunities: An investor perspective.
Grounding AI agents in causal relationships
The Pharmacovigilance Signal Management Assistant: an AI tool for fast, efficient data review with near-human reliability
DCAM (Data management capability assessment model) framework
Enhancing discoverability with the SPHN Metadata Catalog
How Can AI Improve and Accelerate Clinical Trial Review?
ML in drug discovery
UniProt, Rhea and Chebi: Biological curation with a Chemical impact
What do building a start-up and choosing an AI tool as a pharmaceutical programmer have in common? Design Thinking
Roundtables
AI in Model-Informed Drug Development (MIDD)
AI-powered drug discovery for startups & mid-sized biotech. Build, buy, or partner
AI, Biomanufacturing and Workforce Physics: Building the AI-Enabled Bioprocessing Organisation
Change Management in SCE implementation
Generative AI adoption: drivers and barriers
How to bridge academia and industry
Invite Only: Building the Future of AI & Data Strategy in Life Sciences
Lost in the Numbers: Managing Data Overload?
Self-Driving Laboratories in Drug Discovery
When AI Scales in Pharma: Why Governance, Data Lineage and Operating Reality Matter More Than the Model
BioTechX Connect
1 hour. For Partners. Optimal Efficiency
A dedicated power-hour of pre-scheduled 1:1 meetings designed to solve specific challenges
- The format: 4x15 minute meetings
- The Goal: Rapid sourcing and partnership vetting
- The Match: 100% double opt- in (AI-Powered)
Chair's remarks
Chair's remarks
Chair's remarks
Chair's remarks
Chair's remarks
Chair's remarks
Chair's remarks
Chair's remarks
Chair's remarks
A FAIR layer for the INHERENT haemoglobinopathy patient registry
Bioinformatics & AI for nucleic acid therapeutics safety
Building an open scientific RWD platform
Building the AI-Assisted Regulatory Document Ecosystem
The regulatory document lifecycle is being transformed by AI, from early submission planning to dossier authoring and interactions with health authorities. In this talk, we will share our experience building an integrated ecosystem of AI solutions that support regulatory professionals across these activities. Based on an internally developed scenario-based planning platform, teams can explore submission strategies and generate draft content long before final study data are available. We will also showcase Scribbler, an AI-assisted authoring platform that supports document creation, review, quality control, and strategic assessment of regulatory documents, as well as emerging approaches for automating responses to health authority Requests for Information (RFIs) using knowledge from previous RFIs, Clinical Study Reports, and other regulatory sources. Together, these capabilities illustrate how AI can move beyond isolated writing assistance to become a trusted partner across the regulatory submission process.
Digital fundamendals of hybrid care
End-to-end AI Implementation in Dossier Submission
It’s all about change - are we AI ready?
Modernisation of the Statistical Computational Environment
Standardizing the Reaction Data Pipeline: Building a Unified Foundation for a Vibrant AI and Synthesis Ecosystem
Utilizing an organ-chip platform to improve digital twin models
A Practical Path to AI-Grounded Drug Discovery with Knowledge Graphs
LLMs in drug discovery suffer from hallucination and lack of semantic intelligence. This talk presents practical, knowledge-grounded AI applications using knowledge graphs for use cases like target identification and polypharmacy prediction. The core is the neuro-symbolic approach which uses a knowledge graph to transform probabilistic LLMs into reliable reasoning systems. Learn to accelerate your pipeline by connecting internal knowledge with the research ecosystem via persistent identifiers.
Accelerating recruitment with engaged real-world data cohorts
Bridging the gap between FEMtech and FEMhealth
Data Management in the era of AI - What it takes to deliver value at scale
From Paperless Labs to Defensible Decisions: Evidence Infrastructure for Scientific AI
Pharmaceutical research and development (R&D) laboratories have made major progress toward paperless workflows. Electronic lab notebooks, laboratory information management systems (LIMS), and chromatography data systems capture measurements digitally. Yet when a model flags an atypical impurity trend or recommends a formulation change, scientists often cannot trace that recommendation through instrument calibration, method validation, sample history, and study context to the original measurements. The data is digital; the evidence chain remains fragmented.
This keynote introduces evidence infrastructure: governed, bidirectional provenance created as data flows from instruments to scientific decisions. Grounded in the ICAD Principles (Integrate → Contextualize → Analyze → Decide), each integration enriches a scientific context graph. Typed relationships link analytical results to methods, specifications, stability protocols, batch genealogy, instrument state, and regulatory submissions. As these connections accumulate, the context graph becomes an operational knowledge graph whose conclusions remain traceable to source evidence.
The talk shows why evidence chains should be created during ingestion rather than reconstructed for each AI deployment, and how this supports scientific review, human oversight, and evolving transparency expectations. Attendees will leave with practical architectural patterns for assessing whether their digital lab produces accessible data alone or defensible evidence for scientific AI.
From PoC to GxP: Validating Agentic AI in Clinical Data Workflows
Self-Service Bioinformatics at Scale with AWS HealthOmics and Kiro
In the age of highly available, high throughput next generation sequencing, users want to leverage cloud infrastructure to orchestrate bioinformatics pipelines at immense scale. In the past, this has required the joint effort of multiple personas: the bioinformatician, the wet lab scientist, the cloud engineer, and the IT manager all working at different paces and with different priorities. Today, this can be achieved self-service by a number of personas. In this session, we will demonstrate how agentic tooling can be leveraged to provision, deploy, debug, and monitor bioinformatics pipelines powered by purpose-built infrastructure like AWS HealthOmics, compressing the time to science from months to hours.
The Intelligent Engine: Reproducible Workflows. Powerful Agents. Optimized Compute.
CIME (Cheminformatics Model Explorer): Integrating Machine Learning into an Interactive Human in the Loop Workflow
AI-Powered Regulatory Writing I: From Blank Page to First Draft
Talk I, From Blank Page to First Draft, introduces the challenges of regulatory writing and the motivation for AI-assisted drafting. Building on our PRINCE multi-agent framework, we demonstrate how approaches such as prompt engineering, draft reflection, and model customization enable the transition from fragmented source material to structured first drafts, while keeping expert oversight central to the process.
Beyond the App: Data, People, and the Real Foundations of Digital Health
Building a Trusted Research Environment to enable Biobank data analysis at scale and enable secondary use of internal clinical data
Exploratory rare variant analysis in neurodegeneration using AI within UK Biobank and clinical trials
FAIR business value framework
LLMs and knowledge graphs
Partnerships between industry and public sector
Strategic foresight for prioritisation of AI applications in RWE
The Backbone of Modern Medicine: Scaling NHS Data Infrastructure for Precision Healthcare
AI Applications: From Clinical Research To Operational Feasibility
AI-Powered Regulatory Writing II: The Harness Engineering for Deep Research
Talk II, The Harness Engineering for Deep Research, focuses on the system architecture required to support reliable, long-running workflows. Regulatory writing is inherently iterative, involving clarification, evidence retrieval, synthesis, drafting, and refinement. We highlight key design patterns in harness engineering and context engineering, including agent orchestration, tool integration, state management, and iterative reflection, to ensure robustness and adaptability.
Beyond Connecting the Dots: Defining AI-Ready Data and What It Unlocks
AI in pharma is having its "prove it" moment. The people closest to work have mostly figured out that the problem isn't the model, but the data underneath it. However, teams are still wrestling with what it means for data to be AI-ready. Most of the effort goes into connecting and organizing existing data. Paradoxically, far less time goes into defining which data characteristics drive high-quality results downstream. Data definitions that are scientifically valid and consistently applied, coupled with clear traceability, are what make agentic AI in R&D a reliable research partner rather than just another answer-generating tool. It's this foundation that unlocks an agent's real problem-solving ability, which we will walk through using examples drawn from real R&D work.
From FAIR Science to FAIR Manufacturing
Making pharma data usable across R&D, tech transfer and CDMO operations
FAIR principles are well established in scientific research, but their value extends far beyond the laboratory. Discover how a semantic layer and AI-ready data foundations help pharmaceutical companies and CDMOs improve tech transfer, cross-site comparability and data reuse from R&D to manufacturing.
From Invitation to Follow-up: What Swiss Breast-Screening Pathways Reveal About the Data We Need for Prevention
From Silos to Synergy: How an Ecosystem-Driven Platform Accelerates Scientific Discovery
Harnessing Agentic AI to Transform Unstructured Preclinical Antibody–Drug Conjugate Entities into Structured, Harmonized Insights
Making innovation simple: scaling what works
Part 1: Lessons Learned Building Connected Drug Discovery Workflows
A new strategic role for pharma CVC
Beyond the AI Model: How Context Layers Help Shape Agentic AI in Drug Discovery
In agentic AI for drug discovery, the model is only as useful as the biological context it can access, connect, and reason over. ETL pipelines and ELN/LIMS repositories store and organize records, but they are not designed to fully represent biology: data joins can remain syntactic, retrieval may rely on text similarity, and functional relationships, cross-modal linkage, reasoning provenance, and negative-result memory can remain fragmented. The data can sit static, and relations and semantics can stay incomplete. ReefIQ™, powered by HYFT® Technology, is MindWalk’s newly launched biological context layer for AI in drug discovery. It connects and contextualizes discovery data across sequence, structure, function, mechanism, pathway, and literature in one connected representation, creating queryable biological context designed to work with the AI infrastructure around it — whether a customer’s own AI models or MindWalk’s LensAI™ platform. In either configuration, ReefIQ provides the connected context, structured retrieval, and validation layer, while reasoning happens in the AI layer above it. When paired with LensAI™, MindWalk’s reasoning and application layer, that context can inform auditable, human-in-the-loop decision support across target discovery, pan-serotype biologic design, candidate diligence, and mechanism-aware variant interpretation. Ultimately, context can become more useful with each program and measurement as relationships within the data are refined over time.
From FAIR to Interoperability: How Open Standards Unlock Scalable Data Integration in the Lab
As FAIR data moves from aspiration to expectation and AI increases demand for accessible, machine-actionable data, interoperability has become a critical challenge. Laboratories are inherently multi-vendor environments, making open standards essential for scalable integration across instruments, software and technology stacks. The non-profit SiLA consortium brings laboratory users, vendors and integrators together to facilitate the development and adoption of open interoperability standards, with their potential demonstrated through real-world showcases. This talk shares lessons from across the SiLA ecosystem and explores how collaborative standards can enable scalable data integration, FAIR data and AI-ready laboratories.
Part 2: Agentic AI: Automating the Next Generation of Drug Discovery
Predictive Toxicology: Rational Digital Toxicology in the Cloud with a New AI-Accelerated Physics-Based Workflow
Semantic Data Product Architecture
What Stopped That Project? AI and Knowledge Graph Technology for informed pipeline decisions
Autonomous R&D: Transforming Data and Regulatory Complexity into a Strategic Decision Advantage
Build, Buy, or Partner: From Agentic AI Hype to Enterprise Value
As agentic AI moves from demos to decisions, pharma leaders must decide what to build, what to buy, and where partnership creates advantage. This panel cuts through hype to debate ownership, governance, validation, and the real sources of competitive moat.
Data Management & Readiness for AI
Data structures and FAIR processes
From Data to Diagnosis: The Power of Multi-Omics in Clinical Research
Gaining a competitive advantage in drug discovery through linked, multi-dimensional data at scale
Actionability of data in drug discovery depends on the completeness of underlying datasets and the analytical infrastructure to generate meaningful insights. This is amplified in the era of AI-driven interpretation, where models are only as powerful as the data they're trained on. As drug discovery teams adopt rapidly advancing approaches like machine learning for target identification, virtual cell modeling, and multiomic profiling, access to large-scale, diverse datasets with complete metadata has become a strategic imperative. This panel examines how hyperscale initiatives like the Alliance for Genomic Discovery (350,000+ whole genomes and 50,000+ linked proteomes) and the Illumina Billion Cell Atlas are providing the foundational data infrastructure for next-generation drug discovery. Panelists will bring expertise spanning functional genomics, machine learning, ADME, antibody developability and more to discuss what it takes to build AI-ready datasets, the importance and challenges of integrating across diverse datasets and infrastructures, and how both proprietary and pre-competitive collaboration are impacting today’s R&D landscape.
Improving patient care
Navigating AI Adoption in Clinical Trials
Create your personal agenda –check the favourite icon
BioTechX Connect
1 hour. For Senior Decision Makers. Optimal Efficiency
A dedicated power-hour of pre-scheduled 1:1 meetings designed to solve specific challenges
- The format: 4x15 minute meetings
- The Goal: Rapid sourcing and partnership vetting
- The Match: 100% double opt- in (AI-Powered)
Chair's remarks
From Hearing to Brain Health: Biomarkers as a Breakthrough in Early Neuro-Diagnosis and inflammatory diseases
AI at Scale: Charting the Future of Pharmaceutical Innovation
Abstract: The promise of AI to revolutionize drug discovery and development is undeniable. However, moving from isolated AI projects to enterprise-wide, value-driving capabilities presents a formidable challenge for even the most innovative Pharma organizations. The true test lies not in the algorithm, but in the ability to scale.
This panel brings together senior industry leaders to share their strategic perspectives on this critical journey. We will move beyond the hype to address the core operational, technical, and cultural questions that define AI readiness. Our discussion will explore actionable strategies for:
- Crafting a cohesive vision for AI that balances ambitious moonshots with tangible, near-term value.
- Designing effective operating models and data infrastructure to support AI at an industrial scale.
- Cultivating a data-literate culture and empowering the workforce to trust and adopt new AI-driven tools.
- Establishing robust governance and ethical guidelines that foster responsible innovation.
BioTechX Connect
1 hour. For Partners. Optimal Efficiency
A dedicated power-hour of pre-scheduled 1:1 meetings designed to solve specific challenges
- The format: 4x15 minute meetings
- The Goal: Rapid sourcing and partnership vetting
- The Match: 100% double opt- in (AI-Powered)
Chair's remarks
Chair's remarks
Chair's remarks
Chair's remarks
Chair's remarks
Chair's remarks
Chair's remarks
Chair's remarks
Chair's remarks
Chair's remarks
Building an agentic AI platform to improve efficiency across the value chain
ChronoTrack: Spatio-Temporal Representation Learning for Microvasculature-on-Chip Models
Designed In, Not Bolted On: Why Change Management Is the Missing Infrastructure in Clinical AI Deployments
As of late 2024, only 11% of pharmaceutical and biotech companies had fully implemented AI in clinical trial operations — despite years of investment and a technology landscape that has never been more capable. A 2025 McKinsey analysis puts the broader problem in stark terms: more than 80% of organizations report no measurable enterprise-level impact from generative AI. The technology isn't the bottleneck. The organizations receiving it are.
Across clinical operations, a predictable pattern is emerging: a tool clears validation, clears IT, clears legal — and then stalls at the team level. Adoption plateaus. Workarounds persist. The ROI case erodes. Emerging evidence now challenges the assumption that resistance is the culprit. The real barrier is structural: change management is routinely treated as a communication plan appended after build — a series of emails, a training session, a launch announcement. In clinical operations, where process discipline is a compliance requirement, that gap has consequences well beyond a missed adoption metric.
This session makes the case that organizational readiness is not a soft skill — it's a deployment requirement. Drawing on patterns across multiple AI and automation programs in clinical development, this talk presents three design principles that must be embedded at program inception, not activated at go-live:
- Stakeholder impact precedes requirements. Who absorbs the process change, and what does it cost them? That answer should shape the tool design, not follow it.
- Adoption metrics are success criteria, not afterthoughts. If you can't measure behavior change, you can't claim the benefit.
- Capability building is a workstream, not an event. Sustained adoption requires ongoing investment in people infrastructure — not a one-time training push.
Attendees will leave with a diagnostic lens for identifying where their current AI deployments are structurally at risk — and a practical framework for designing organizational readiness in from day one.
Don’t transform alone: What supporting 100+ early stage startups has taught us about accelerating innovation in pharma R&D
From Data to Answers: Exploring Agent-Driven Discovery in Cell Painting Screens
NHS secured data environments for RWE generation
Prevention instead of treatment, medicine 3.0
Reinvent Drug Development: How AI Agents are supporting clinical trial decisions
Swiss personalised health network
The rapidly changing landscape for precision health at a tertiary children’s hospital
AI-based multimodal clinical decision support in oncology - from screening to therapy
Analysis of Clinical Pharmacokinetic Data
Building an AI Scientist for a new generation of Pharma R&D
Pharmaceutical R&D requires scientists to tackle complex and diverse tasks in drug discovery and development. A new generation of AI scientists offers a way to support and increasingly automate aspects of this work. Owkin has developed K Pro, an AI Scientist designed to understand biology and grounded in multimodal patient data, domain-specific AI and deep scientific expertise.
In this session, we’ll explore what makes this approach different from general-purpose AI and how K Pro can support complex scientific tasks across the pharma R&D lifecycle. Through examples from our current work, we’ll show how AI Scientists can help researchers move from scientific questions to testable hypotheses faster, and discuss how these capabilities could evolve beyond individual projects to drive IP and asset development, scale across the enterprise, and ultimately shape a new generation of pharma R&D.
From Data to Decisions: Connecting Science, Data, and AI Across Biopharma R&D
Funding the Frontier, Governing the Risk: How MSK's AI Tech Fund Balances Innovation and Responsibility in Real-World Evidence Applications
Medical Value Algorithms for Diagnostics and Improvement of Care Along the Patient Journey
Orchestrating the Future of Clinical R&D: From Long-established Expertise to AI‑Driven Expert Agents
For years, digital transformation in clinical R&D has primarily focused on automating existing processes—often reinforcing long‑established functional silos such as data management, statistical programming, analysis, and reporting.
Artificial Intelligence represents a far more structural shift. Rather than simply accelerating current workflows, AI challenges how organizations are designed, governed, and held accountable.
In this joint session,Johnson & JohnsonandSASexplore how AI enables a move from siloed execution to anorchestration‑based model, where human experts leveragespecialized AI agents across data management, statistical programming, analysis, and reporting.
The discussion highlights how governed, auditable AI platforms make this model operational at scale, while addressing emergingrisks and regulatory expectations. A forward‑looking perspective on how health authorities approach AI adoption completes the session.
This session is for leaders looking to understand the fundamental shift in value, from manual production to judgment‑driven decision‑making, coordination, and transparency.
Regulatory-grade real-world evidence from unstructured clinical data
Roughly 40% of the clinical facts research needs never reach a structured data field: diagnoses, medication adherence, biomarkers, staging, social determinants, family history. Frontier LLMs can read that text, but at population scale they're expensive, non-deterministic, and hard to audit. This session shows how specialized medical language models extract and de-identify clinical facts at regulatory-grade accuracy: 98% F1 on PHI detection, and primary site, histology, and tumor staging extracted from unstructured pathology text at regulatory-grade accuracy (over 95%) – all at over 80% lower cost than current frontier models, with deterministic, reproducible output. Those facts become a governed, OMOP-standard real-world-evidence asset, with every value traced to its source note and every extraction carrying a confidence score. With that foundation in place, and a shared MCP boundary on top of it, use cases like cohort building, real-world evidence, clinical trial matching, and protocol design become far easier to build and to audit.
The Imaging Data Factory: Activating Imaging Data Across the Full R&D Pipeline
Medical imaging data is essential to AI-driven drug development, yet its full value is often hampered by curation, compliance, and workflow challenges. Flywheel provides a single platform that automates imaging pipelines to turn raw, scattered data into model-ready datasets — without compromising provenance or compliance. Leading pharmaceutical companies and academic reading centers rely on this approach to accelerate subject enrollment, endpoint analysis, and AI development, moving faster from data readiness to trustworthy results.
The Translational Stack: Leveraging real-world data and multiomics to decode GLP-1 early response
DNAnexus is a multi-omic analysis and collaboration platform that extends beyond core bioinformatics teams by connecting real-world data with translational research. In this study, we present an end-to-end biomarker discovery workflow executed on the platform using the Ovation GLP-1 dataset—a rich resource integrating robust clinical and phenotypic data with whole-genome sequencing (WGS). Specifically, our study investigates early GLP-1 receptor agonist response versus resistance, evaluated by the reduction in baseline HbA1c after 24 weeks of treatment.
Leveraging multiple analytical approaches—including traditional statistical genetics, an AutoML-based framework, and complementary analyses—we report preliminary findings characterizing phenotypic differences between early response and resistance cohorts. We highlight candidate gene signatures consistently identified across these methods and discuss their potential biological relevance. Ultimately, this work illustrates how combining real-world clinical data, multi-omic assets, and flexible computational tooling on a single platform can accelerate translational research—empowering teams across the discovery-to-clinic continuum to collaborate on complex therapeutic questions like GLP-1 response heterogeneity.
AI and Accountability in a GxP setting - Considerations on Automation vs. Risk
AI for bioinformatics
Beyond the Silos: Navigating the AI regulatory complexity
Biomedical Informatics Platform, (LOOP BMIP)
Building Responsible AI: Lessons for Life Sciences
As large language models and generative AI become increasingly embedded across biotech and life sciences, organisations need governance approaches that enable innovation while maintaining trust, transparency, and regulatory compliance. This presentation explores responsible AI principles, practical strategies for managing the unique risks of large language models, and how organisations can prepare for increasingly autonomous AI systems.
From Prediction to Validation: An Industry-Standard Approach to TCR-Mimic Antibody Development
Guy's Cancer Real World Evidence Programme - an opportunity for collaborative digital health.
Image analysis in cryoEM to rationalise drug discovery
Lighthouse Assistant: Exploiting FAIR Knowledge Graphs and Agentic Workflows for Clinical Study metadata management
Strategic view on digital innovation in clinical development
AI applications for patient risk certification in opioid use disorder
AI for Every Scientist: How Agentic AI, Grounded in the Scientific Record, Accelerates Discovery
General-purpose AI assistants can generate text, but they cannot traverse a scientific graph, cite source records, or take action directly within an R&D platform. This session explores what changes when agentic AI is built into the scientific data foundation.
We’ll look at how connected wet lab and dry lab data, no-code access to scientific models, and AI agents can help scientists work more efficiently. Examples include generating hypotheses grounded in internal data and published research, using reusable Skills to standardize workflows, and automating routine tasks such as report writing, sample transfers, and inventory checks.
The session will also demonstrate how an agent can connect to an external ontology using MCP to bring additional knowledge into Benchling, and how scheduled agents can flag expiring reagents before an experiment begins. Agentic AI is only trustworthy when it is grounded in well-structured scientific data and produces traceable, auditable results, turning AI from a black box into a collaborator scientists can rely on.
Enterprise Transformation with AI: From isolated use cases to enterprise value
A shared language and structured approach are key to align on outcomes, guide complex transformation work and move beyond isolated AI use cases toward measurable business value. The Enterprise Transformation with AI framework helps teams reshape how the enterprise works — so AI, people and workflows can perform together at scale.
From Retrieval to Reasoning: Where LLMs Are Headed in Life Sciences
How digital health is transforming our care system
The compute gap: Why healthcare AI’s biggest bottleneck isn’t the algorithm
Who wins when AI enters the trail? An investor’s view on the landscape
Engineering Certainty and Clinical Safety: From Probabilistic to Deterministic AI
From Data to Decisions: Building a Scalable AI Engine for Pierre Fabre Pharma R&D
How we are moving from fragmented data and an outsourcing model to an integrated, AI and agentic powered R&D engine. What works (and does not) along the way.
From Noise to Guidance: Rethinking Alerts in EMR Systems
Clinical decision support (CDS) alerts in EMR systems aim to support clinical decisions, yet many are viewed as noise: frequent, disruptive, and often necessary to override. In part, this is because existing CDS alert design recommendations have been shaped predominantly by principles of usability and human factors engineering, which tend to emphasise the limitations of human cognition rather than the complexities of human decision-making. This session presents a different approach: designing alerts not as expedient solutions, but as interventions to guide clinical decisions in ways better aligned with how humans actually make decisions.
Drawing on doctoral research in digital health, the session introduces a multidimensional model for designing CDS alerts with human decision-making considerations. The model defines and organises nine building blocks of CDS alerts and demonstrates how nine behavioural effects from the MINDSPACE framework for behaviour change can inform their design. Developed through design-science research and informed by several conceptual and empirical studies, including interviews with CDS alert designers and users, as well as digital health practitioners and academics, it offers a structured and behaviourally informed approach to rethinking alerts as more meaningful interventions to guide clinical decisions.
Harnessing AI for Global Health Impact: Why NGOs Must Unite to Lead
AI is transforming drug discovery and clinical development faster than any organization can navigate alone. For NGOs, this isn't just a technological shift — it's a defining moment.
DNDi has spent decades proving that partnership is the most powerful engine for impact: 14 new treatments, 6 deadly diseases defeated, millions of lives saved. Now, AI is supercharging that model. From intelligent compound screening to automated clinical documentation and real-time safety surveillance, the opportunities are immense — and they are accelerating.
But here's the hard truth: no NGO can capture this potential in isolation. The data, the talent, the infrastructure required to deploy AI at scale demand a new level of collaboration. The organizations that will lead the next era of global health innovation are those bold enough to build coalitions, join consortia, and co-create shared platforms with aligned partners.
This session makes the case that AI is not just a tool — it's the catalyst for a more connected, more ambitious NGO ecosystem. The future of equitable drug development won't be built by any single organization. It will be built together.
Roche approach to modernize shop floor operations with AI
This session explores how Roche is transforming manufacturing shop floor operations through AI-driven digitalization and intelligent process automation. The presentation highlights the Digital Operational Excellence Program (DOEP) and the deployment of a MuleSoft-based MCP (Model Context Protocol) architecture integrated with Tulip to modernize data capture, connectivity, and operational decision-making across manufacturing sites.
Attendees will learn how Roche is replacing manual paper-based shop floor logging with real-time digital process capture, enabling centralized data integration and conversational AI capabilities for manufacturing users. The session willexplainhow AI-powered insights, streamlined workflows, and interoperable systems accelerate operational excellence, reduce manual effort, and improve manufacturing agility at scale.
Key topics include:
- AI-enabled shop floor digitalization
- MCP server deployment and enterprise integration architecture
- Real-time manufacturing data orchestration using MuleSoft and Tulip
- Conversational AI interfaces for manufacturing operations
- Operational efficiency gains and business impact
- Governance, security, and deployment considerations in regulated environments
The adoption gap: why digital transformations fail (and how to fix it).
BioTechX Connect
1 hour. For Partners. Optimal Efficiency
A dedicated power-hour of pre-scheduled 1:1 meetings designed to solve specific challenges
- The format: 4x15 minute meetings
- The Goal: Rapid sourcing and partnership vetting
- The Match: 100% double opt- in (AI-Powered)
Chair's remarks
Chair's remarks
Chair's remarks
Chair's remarks
Chair's remarks
Chair's remarks
Chair's remarks
Chair's remarks
Accelerating discoveries in diagnostic biomarkers using AI solutions
AI Driven Endpoints: Redifining Clinical Trials for the next decade
AI influenced nano technology in drug discovery and development
AI, you, and the future of work
What types of work will AI reshape? How do we adopt AI? The 3 Es of of AI Adoption, Human Elements: Employees & Leaders, Case Study: A Tale of 2 Swedish Companies (Klarna & Ikea) - Conclusion: What can and should we do now for the future.". It is currently something about ppm.
An Investigator’s Perspective on Practical Impact Beyond the Hype
Artificial intelligence is rapidly entering clinical trial design and execution, yet its value depends on how well it addresses the real-world challenges faced by investigators, sites, sponsors, and patients. This presentation explores AI in clinical trials from the investigator’s perspective, focusing not on technical algorithms but on practical clinical and operational impact.
The session will examine where AI can meaningfully support trial feasibility, patient identification, eligibility screening, recruitment, retention, risk-based monitoring, endpoint assessment, data quality, and safety oversight. It will also distinguish realistic current applications from hype, while addressing key limitations including bias, poor data quality, lack of transparency, regulatory expectations, and the risk of over-automation.
Attendees will leave with a clear framework for evaluating AI-enabled trial solutions, collaborating effectively with sponsors and technology partners, and adopting AI in ways that improve efficiency while preserving patient safety, data integrity, scientific credibility, and investigator judgment
ELNs and Academia - a difficult story
Garbage In, Genius Out? Why AI-Ready Data Is the Real Bottleneck in Drug Discovery
Garbage in, genius out is not how it works! AI in drug discovery fails less often on model architecture than on the data beneath it!
The challenge continues to be data being fragmented across systems, inconsistently annotated, and often invisible to the people building the models. This talk discusses what "AI-ready" requires around ontology harmonisation, metadata completeness, semantic integration, and curation quality you can measure. Excelra’s experience in data and bioinformatics provides a practical lens for identifying when data readiness accelerates discovery and when the promise is overstated.
Innovating towards clinical outcomes in the AI and digital health world
Rhythms in the gut – A new target to prevent and treat diseases?
Target discovery using LLM
AI Blueprint for Lifesciences: AI-Driven Drug Discovery and Molecular Design
Beyond the Pilot: How Agentic AI Is Already Running Across Pharma — From R&D to Commercial
A candid look at different agentic AI systems already live inside global pharma organizations. Dr. Grace Lomax, Chief Solution Officer at Globant and Javier Jiménez, Chief Medical Officer at PharmaMar, share how agentic AI is narrowing 8,000 drug-combination candidates down to a ranked top 10, 15x faster, with 90%+ retrieval accuracy, while a parallel system has been running in production for 15 months, managing 10+ commercial brands across two markets. A real, measurable look at what happens when AI moves past the pilot stage.
Designing Agent-Ready Regulatory and Pharmacovigilance Processes: Data Foundations, Compliance, and Applied Use Cases
Embedded Calibration Markers for Lateral Flow Assay Data Quantification using Smartphone in Point-of-Care Diagnostics
End to end evidence generation ecosystems in neuroimmunology powered by federated learning
From Data Flood to Decision: Why Peer-Reviewed Science is the Missing Foundation for Trusted AI in R&D
Lessons in building a scientist-first AI protein engineering platform
Multimodal Data Assets - Optimizing Drug Development with Actionable Data Insights
Reimagining clinical ops planning with AI - a computational twin approach
Strategy for image platform
A novel technique for sensitive detection of disease-specific T cells
Current T-cell diagnostics predominantly rely on cytokine release assays, which may exhibit reduced sensitivity in immunocompromised patients due to impaired effector T-cell function. To address this limitation, we developed ProliSpot, a novel immune-monitoring platform that measures antigen-specific T-cell proliferation at the single-cell level using fluorescence imaging and automated analysis.ProliSpot combines the biological sensitivity of proliferation-based assays with the scalability and standardization required for routine clinical diagnostics. Following antigen stimulation, proliferating T cells are quantified through automated image acquisition and analysis, providing a direct measure of antigen-specific cellular immunity. The platform has been developed as a user-friendly kit format and is currently being translated towards an IVDR-compliant diagnostic workflow.As a first clinical application, we developed ProliSpot-TB for the detection of latent tuberculosis infection (TBI). Preliminary studies indicate that ProliSpot-TB detects TB-specific immune responses in a higher proportion of immunocompromised individuals than conventional interferon-gamma release assays (IGRAs), addressing a major unmet need in tuberculosis prevention.Beyond tuberculosis, the ProliSpot platform has potential applications in infectious diseases, vaccine evaluation, immune monitoring, and personalized medicine. This presentation will describe the technology, automation strategy, clinical validation pathway, and opportunities for broader implementation of proliferation-based immune diagnostics.
Architecting the AI Lifecycle: Data Infrastructure, Model Metadata, and Synthetic Data Management
Clinical trials feasibility suite
Clinical-Grade Digital Mindset – Agility & Resilience for Human+AI teams in pharma
Federated Learning Interoperability Platform: Unlocking Real-World Medical Imaging for AI
MLConfGen - Transforming Hit Discovery with Generative AI
Multimodal Transcriptomic Analysis Reveals Multiomic Biomarkers for Disease
Physics-Informed Large Language Models for Biologics: Applying Nuclear Engineering Rigor to AI Safety and Reliability
Redesigning the Falls Pathway from the Inside Out: AI, Home Care, and NHS in a Single Integrated Pathway
Unlocking Germany's Health Data Potential beyond clinical trials
17:25 Big Systems or Best-Fit Solutions? Evaluating Value, Impact and ROI in Healthcare Innovation
AAV Vector-mediated DMD Treatment: Advancements and Challenges
AI in Pathology: from pixels to precision
Beyond Adoption: The gaps in Digital Transformation no one speaks about
Getting people to adopt new technology is only one part of transformation. What happens when the tools change the work itself, shift where decisions are made, redistribute expertise and accountability, and challenge the structures around them?
Drawing on complex enterprise transformations, Amruta explores the gaps that emerge beyond adoption, from changing roles and capabilities to governance, decision rights and ways of working, and asks a bigger question: are we transforming technology, or are we transforming the organization around what technology has made possible?
Importance of real-world evidence for rare diseases
Real life examples of AI in clinical developments
Research Hospitals’ Role in Digital Health Technology Transfer: New Frontiers
Speed Meets Quality: How Roche Leverages GenAI for 10-Minute Clinical First Drafts
The Disease Field Problem: Multilingual Disease Normalisation for a National Sample Catalogue
Swiss Biobanking’s national biosample catalogue, NExT, makes over 100 biobanks visible to researchers but few of them are willing to actually upload their sample data. Depositing is unpaid work, and because disease information is entered as free text across four national languages, the deposited data cannot be consistently queried, aggregated or compared anyway. Effort in, no value out.
This talk presents what Swiss Biobanking is doing to invert that calculation: a curation interface, built into the deposit workflow, that maps free-text disease mentions onto a standard ontology as they are entered.
Behind it sits Messier, an open-source, fully on-premise entity-linking system. It contributes no new algorithm; it makes an existing one deployable under Swiss data-protection law, where cloud LLMs are disqualified by design.
Digital Transformation in Biomedical Research: Challenges and Opportunities
Enable compliant preclinical discovery data sharing between organizations
Reimagining Clinical Site Selection with AI: Analytics and Process Modernization
The evidence foundation that delivers AI agents with the best accuracy and scalability
AI agents can synthesize information rapidly, but drug discovery requires evidence that is mechanistically precise, traceable, and reproducible. We benchmark LLM-only workflows against the same models connected to knowledge graphs through MCP. Across representative discovery tasks, we compare answer quality, coverage, adherence to biological constraints, provenance, and inference cost. The results demonstrate how curated knowledge can provide the evidence backbone needed to make agentic drug discovery more reliable and scalable.
AI in clinical trials is it all a bit shit?
Beyond the Hype: AI, Supercomputing, and Data Platforms Driving Biotech Success with HPE and AMD
Bridging AI and Open Source: Advancing Data Science for Clinical Quality by IMPALA consortium
Data Integration: Generating Insights with FAIR Data Principles
The Future of Bioinformatics: Challenges, Opportunities, and Innovation
Your LLM is Monolingual – Your BioTech Company Isn’t. Now What?
Biotech companies operate across multiple languages and cultures, yet the LLMs supporting their R&D, clinical and regulatory workflows are still built largely on English-only biomedical corpora, including models trained exclusively on PubMed abstracts. At the same time, multilingual clinical NLP research shows uneven data availability and inconsistent model performance across languages, raising important questions about how reliably AI can support global evidence extraction and documentation.
This panel opens a discussion on what happens when multilingual organisations rely on monolingual models – and what teams can do about it. Where do gaps, risks and inefficiencies emerge in cross-site collaboration, terminology alignment, documentation practices and knowledge sharing – and where might new opportunities arise? Bringing together perspectives from AI development, clinical and regulatory operations as well as linguistic diversity management, we explore what it takes to make LLMs more reliable and usable across global teams.
Create your personal agenda –check the favourite icon
Introductory remarks
The evolution of Medical Affairs: the drivers of future change, innovation and technology use.
Investment trends in AI
BioTechX Connect
1 hour. For Senior Decision Makers. Optimal Efficiency
A dedicated power-hour of pre-scheduled 1:1 meetings designed to solve specific challenges
- The format: 4x15 minute meetings
- The Goal: Rapid sourcing and partnership vetting
- The Match: 100% double opt- in (AI-Powered)
Chair's remarks
Chair's remarks
Chair's remarks
Chair's remarks
Chair's remarks
Chair's remarks
Chair's remarks
Chair's remarks
Judges Remarks
Application of Artificial Intelligence and Machine Learning Techniques to Enhance Early Detection and Diagnosis of Niemann-Pick Class C1 Disease and Associated Liver Dysfunction
Beyond the Document: Joining narrative, data and lineage to answer our most critical questions
40–70% of regulatory submission content is repeated across studies — yet re-authored from scratch every time. At GSK, we're rebuilding our content architecture around governed, reusable components rather than static documents: data and narrative are structured, traceable, and composed on demand, not copy-pasted and reconciled by hand. This session outlines our vision for Structured Component Management (SCM) — a three-layer model (Demand, Supply, Layout) that turns submission content into an inspection-ready, event-driven asset rather than a recurring manual burden.
No vendor today fully solves this: it's a systems problem, not a tooling gap. That's why we're opening this up. We'll share what we've learned, and where we're still working it out — then turn to the room. If you're wrestling with the same shift from document-centric to data-centric submissions, we want to hear how you're approaching reuse, governance, and interoperability, and start building the cross-industry community this problem needs.
Takeaways for attendees:
- A practical model for moving from document-centric to component-centric content
- An open invitation to shape a peer community tackling data-centric submissions together
Digital Transformation and AI how to apply it effectively
Discovery and engineering of next-gen immunotherapies with EMLy Co-pilot
Evaluating Data Utility in Anonymization, Federated Approaches, and OMOP-CDM
Background:To estimate remaining data utility, we evaluated three data strategies: Anonymization, Federated Approaches, and OMOP-CDM transformation.
Methods: CDISC-SDTM Data from a retrospective HER2+ breast cancer study (73 variables) were anonymized and mapped to OMOP-CDM. Using DataSHIELD, we tested a federated approach by splitting SDTM and OMOP databases into three samples. Statistical analyses (descriptive statistics, regression methods, survival analyses) for each method were compared against the raw CDISC-SDTM gold standard, focusing on information loss, consistency, and reproducibility.
Results: None of the anonymization methods successfully reproduced all statistical analyses. The federated approach demonstrated good consistency but showed decreased accuracy in multivariate models due to database variability. Conversely, CDISC-SDTM was successfully mapped to OMOP-CDM, showing high statistical concordance.Conclusions: Whilst data was successfully mapped to OMOP, utility was reduced when further privacy preserving methods were applied. A trade-off has to be found between privacy and usefulness of data.
Keeping AI Safe: Data Hygiene, Supply Chain Defense, and Leakage Prevention
Quantum and AI to discover new medicines
When Regulators Come Knocking: Building Agentic AI That Can Answer
Where Knowledge Graphs Create Value: From Published Knowledge to Company Data
Challenges and Solutions for Raw Data Management in Pharmaceutical Companies
Application of Quantum Computing for Protein modelling and ensemble
Demystifying Knowledge Graphs: Ontologies, Methodology and Federated Governance
From Clinical Evidence to Societal Impact: How Real-World Data Accelerates Access to Innovation
From FAIR Data to Agentic Science: What Happens When Data Gets an Agent?
AI is reshaping scientific research, but success depends on more than access to data. While FAIR principles have improved data sharing and reuse, AI and autonomous agents require data that is contextualised, interpretable, trusted, and actionable.In this session, Bruno Fievet explores the evolution from FAIR data to AI-ready and agent-executable knowledge. He examines the additional foundations needed to support trusted AI, scientific agents, and explainable decision-making at scale, including semantics, governance, processes, and knowledge representation.Attendees will gain a practical perspective on how life sciences organisations can prepare their data ecosystems for the next generation of AI-driven and agent-enabled scientific discovery.
LLMs at the University of Zurich
Sorry, Who Are You Again? — Personal Branding and Authority Beyond the Company Badge
The (R)evolution of Nanocyclix: A Data-Driven AI/ML Platform for Novel Kinase Inhibitors
Unleashing Innovation: From Data to Discovery with AMD + HPE
Genome2Protocol and The Rare Diseases Case: Between the Data We Have and the Patients We're Missing
Agentic AI use-cases clinical quality
Beyond Clinical Trials: AI, Real‑World Evidence, and the Future of Pain Care
From Innovation to Revenue: Fixing the Commercial Gap in Science-Led Companies
Strong innovation doesn’t automatically lead to commercial success. Many science-led companies struggle to translate capability into revenue. This session introduces a platform thatidentifies commercial gaps and turnsinnovation into a clear, execution-ready growth strategy.
Generating drug candidates in the chemist-in-the-loop framework
HepSAFE: Integrating Clinical, Literature, and Experimental Evidence for DILI Prediction
Improving Clinical Trial Retention with AI-Powered Dropout Risk Prediction
Making Knowledge Graphs Interoperable: Architectures for Agents, Integration and Reproducibility
Use cases transforming pharma with applied behavioural science
Beyond Scale: Data- and Compute-Efficient Foundation Models for Phenotypic Screening
Agentic Competitive Intelligence: How to Walk the Trust–Coverage–Speed Tightrope and Stay Upright
This talk presents a practitioner’s view on agentic competitive intelligence built on LLMs, grounded in real deployments of a competitive intelligence agent. I examine the core tension between trustworthiness, coverage, and response time, and show how to balance these forces on a tightrope — keeping agents credible, comprehensive, and timely enough for industrial decision workflows.
Agent-orchestrated Insights across Clinical Development
From Awareness to Adoption: Driving Change Beyond the Go-Live
Most transformations do not fail because people were not informed. They fail because awareness was mistaken for adoption.
This session reframes change as a behavioral journey, not a go-live activity. It explores how Organizational Change Management can help people move from understanding achange, to engaging with it, toconfidently working in a new way. Drawing on experience across complex enterprise IT transformations, the session focuses on designing change interventions that reduce friction, create relevance, and sustain momentum beyond implementation.
Mapping therapeutic hotspots at scale to develop better drugs
The Digital Transformation Deadlock: Why Your "Integrated" (PPM) Platform is Killing Your Strategy
Why Digital Twins and RL for Biology Are Starved for Real-World Data — and What That Data Actually Needs to Look Like
Most digital twin / RL-for-biology work today trains on synthetic, sparse, or single-modality data (imaging-only, or genomics-only cohorts).Longitudinal, multimodal,linkedreal-world data (EHR + NGS + pathology + pharmacy + labs, tracked over time per patient) is the actual bottleneck for state transitions in a digital twinorreward signal in an RL formulation. This is the exact problemthat OmicsBank is solving at global scale.
From Nature to Industry: Unlocking the potential of Fungal biocompounds for skin health and beyond
How to Accelerate Discovery with AI-assisted Automation
Mapping the Path to Clinical Implementation of Multi-omics
Pierre Fabre medical care digital transformation journey
An engineered cell culture device to reproduce human tissue oxygenation for improved preclinical prediction
Women in Leadership Keynote Panel
Get involved at BioTechX Europe
Created by
Terrapinn is proud to be a member of isla.
Working together to build sustainable events