3rd edition of GraphTalk pharma & life sciences returns on 2 July 2026 in Munich, again as a hybrid event.
Join us to explore how leading organizations are moving beyond AI experimentation and into production-ready, reliable AI systems powered by context graphs.
AI fatigue isn’t about the model, it’s about the foundation. In pharma and life sciences, fragmented data limits the impact of GenAI. Without connected, explainable context, AI remains stuck in pilots.
Knowledge graphs change that. They unify data across scientific literature, omics, clinical trials, and real-world evidence, creating a structured reasoning layer that enables grounded, traceable, and up-to-date insights.
This year’s agenda follows the pharma product lifecycle, showcasing how GenAI and knowledge graphs drive impact across:
- R&D – accelerating discovery and target identification
- Clinical – enabling explainable, data-driven decisions
- Supply Chain – improving resilience and operational intelligence
Throughout the day, you’ll hear real-world use cases from leading organizations using Neo4j to bring context to AI and move from pilots to production.
You’ll leave with a clear understanding of where GenAI and context graphs can take your organization next and how to get there.
Choose your preferred experience: attend physically on the day, or stream online, whatever suits your schedule and interest.
This is a co-located event with two dedicated tracks—Pharma & Life Sciences and Manufacturing—running in parallel. You’ll have the opportunity to network with peers across both industries and explore how graph technology is driving innovation at scale.
In this keynote, we will explore how knowledge graphs unlock the natural connections within enterprise data, providing the context AI needs to deliver meaningful results. He will highlight how organizations leverage graph-powered contextual intelligence to transform decision-making, turn data into competitive advantage, and drive real business impact.
Biomedical knowledge graphs (KGs) encode multi-scale biological relationships and serve as a powerful foundation for early drug discovery. In this talk, we present two KG-driven approaches for target hypothesis generation and prioritization. First, KG reasoning methods leverage graph topology and local neighborhood structure to infer novel gene-disease associations, augmenting curated knowledge with predicted connections . Second, graph-based retrieval-augmented generation (GraphRAG) leverages the KG as a structured retrieval source, anchoring each retrieved fact to a verifiable graph relation. This structured grounding improves factual reliability and reduces hallucinations compared to standard RAG. Together, they demonstrate how knowledge graphs can serve as both a predictive engine and a grounding mechanism for AI-supported target discovery.
KWS Saat | From graph to gene: integrating Knowledge Graphs and AI for explainable candidate gene discovery
Identifying candidate genes for complex traits in agriculture remains challenging due to fragmented and heterogeneous biological data. We present an approach that combines knowledge graphs, link prediction, and generative AI to enable explainable gene discovery in plant breeding.
We describe how we construct an integrated knowledge graph linking genes, traits, and diverse biological evidence across crops, and how we apply both topology-based and machine learning–driven link prediction methods to prioritize novel candidate genes. To improve interpretability and accelerate hypothesis generation, we leverage generative AI and GraphRAG to traverse the graph and produce biologically meaningful explanations.
This framework connects predictive performance with transparent reasoning, supporting more informed and interpretable candidate gene discovery.
While biological systems are inherently interconnected, knowledge of these systems is both concentrated and fragmented. Model species, such as Arabidopsis, maize, and rice, are knowledge rich and well represented in public data repositories. Yet, when investigating species outside of these models, knowledge is frequently derived from defining similarities to model systems. Here, we discuss the process and outcomes of a proof of value project to connect this fragmented data landscape into a core knowledge graph. The talk will illustrate how the basics - identifier mapping & development of a core data model enables exploration and contextualization of experimental data. Finally, we will discuss how the core data model can serve as a data foundation for agentic systems to allow non-data scientists to explore their data and develop hypotheses.
Insel Spital Bern | From free text to Knowledge Graphs: visualizing Swiss medication and procedure mapping in Neo4j
We present a Neo4j-centered framework for harmonizing Swiss electronic health record medications and procedures. Heterogeneous, multilingual free-text entries were mapped to standard vocabularies using two medication workflows—OHDSI USAGI and a constrained large language model approach—and a cross-lingual retrieval pipeline for procedures. All mappings, classifications, and semantic relationships were ingested into Neo4j to support interactive graph visualization, validation, and gap analysis. The graph structure enabled exploration of semantic neighbourhoods, detection of inconsistent or missing links, and assessment of mapping quality through shortest-path metrics. Neo4j therefore served not only as a repository, but as an analytic and visual engine for clinical interoperability.
This session approaches patents from an analytics perspective, focusing on how patent data can be used to understand barriers to generic entry and their impact on patient access particularly in the context of delayed competition. It highlights the complexity and interconnected nature of patent datasets, where relationships across applications, families, and legal events are difficult to capture using traditional methods. The presentation demonstrates how graph technology can model these connections, enabling a more holistic analysis of patent lifecycles and uncovering hidden patterns that drive clearer, data-driven insights.
Enjoy a light lunch and continue the conversation with fellow attendees, speakers, and Neo4j experts in a relaxed networking setting.
-------------------
Streaming Sessions during lunch (virtual event only):
13:30 - 13:50 | Knowledge Graph of Drugs Data for Swiss Healthcare System
As part of a bachelor’s thesis in medical informatics at the Bern University of Applied Sciences in Switzerland, the speaker developed a knowledge graph that summarises medication data from several data sources in a Neo4j database. As a newcomer to graph databases, the speaker will show the advantages of Neo4j compared to relational systems and the challenges that arose during data modelling. The speaker will also discuss the special features of Neo4j Bloom/Explore for visual data exploration.
You will learn from the newcomer’s experience so that you do not make the same mistakes in your first projects with Neo4j. You will also be inspired by the flexible possibilities offered by graphs and Neo4j.
Christian Franke, Product Owner Data & Analytics, SwissDRG
13:55 - 14:15 | MedQGraph: TMKGs for AI-Driven Healthcare Insights
Clinical decision-making depends on Electronic Health Records (EHRs), but their complexity hinders efficient extraction and analysis. Traditional knowledge graph methods focus mainly on structured data and static relationships, limiting advanced querying. We present the Medical Record Knowledge Graph (MRKG), a scalable framework that transforms structured MIMIC-IV EHR data into an interpretable, queryable graph using Neo4j. MRKG integrates diverse clinical entities, diagnoses, procedures, and medications into a cohesive structure, enabling comprehensive exploration of patient history.
Isaac Ritharson, AI Researcher, Northeastern University
Boehringer Ingelheim | Supply Chain Insights, Without the Training Manual: Agents on a Pharma Context Graph
A pharmaceutical supply chain is hard to see as one connected whole: the data lives in separate systems — supply chain, quality, regulatory affairs, and finance — so a product's ingredients, who supplies and makes them, and how it's forecast to sell rarely line up in one place. Boehringer Ingelheim's Supply Chain Insights initiative changed that, unifying those sources into a single context graph — the business's first real overview of its internal and external supply chain. But connecting the data was only half the problem: the drill-down interface needed training, leaving both quick answers and big-picture questions out of reach for most users. This talk is about closing that gap — an agentic layer that lets anyone interrogate the graph in plain language, from tracing one product's ingredients and suppliers to spotting where the network is over-concentrated or exposed to a live geopolitical or natural-disaster event. What resonated most in early demos wasn't the written answer but the subgraph itself — one product's entire slice of the network in a single interactive view, rather than something pieced together by hand. For any product manager, that alone would be hugely useful.
An industry standard for timely investigation closure is 30 days. We present a human-in-the-loop approach to reducing investigation times with AI and Context Graph data. (If you use tools like Veeva or TrackWise, your organization is already capturing the needed context.) Intuitively, lean manufacturing leaders know that shorter investigation timelines should benefit the bottom line, but where and how much? We quantify the financial value of shorter investigations and use knowledge graph simulations to avoid the bullwhip effect.
Discover how Neo4j combines knowledge graphs, agentic AI, and graph visualization to power the next generation of intelligent applications. Learn how to build context graphs that ground AI in trusted enterprise knowledge, enable GraphRAG and agentic workflows, and help users explore and validate insights through interactive visualization with Neo4j Studio. See how connected data becomes explainable, actionable intelligence.
