![(You)-[:COME_TO]->(Booth:
624)
alexander.jarasch@neo4j.com](https://image.slidesharecdn.com/20241009biotechxneo4jalexanderjaraschkristofneysfinal-241014084703-20c61aa2/85/Neo4j-Public-Graph-Slides-BioTech-X-Basel-2024-63-320.jpg)
Neo4j Public Graph Slides - BioTech X Basel 2024
- 1. From Data to Discovery: Using Generative AI and Knowledge Graphs to Unlock Hidden Insights in Biotech Dr. Alexander Jarasch Technical Consultant Pharma & Life Sciences Kristof Neys Director Graph Data Science and GenAI
- 2. Anti-counterfeiting Fraud detection Circular payments Anti-money laundering Account/identity control Threat detection Adaptive & Intelligent access control Zero trust Reputation scoring Route planning & optimization Risk Analysis Inventory planning Real-time shipment tracking Product pricing Product inventory Product recommendations New Product introductions Product Customizations Bottleneck identification Process improvement Process monitoring Process automation Recommendations Personalization Dynamic Pricing Intelligent Ads (Targeting) Customer Loyalty Programs Customer Offers Churn prevention Skills management Talent learning & development Career mgt Job search Orgs (who is who) Finance Network & Security Infrastructure Suppliers & Partners Product Customer Employees Process Your business is a graph
- 3. Neo4j Inc. All rights reserved 2024 3 Integrates with All Your Data Patents / Literature Omics RWD / Clinical trials Open Target Omics High-throughput screening Ontologies / Terminologies Ontologies / Terminologies Internal Data External Data
- 4. Neo4j Inc. All rights reserved 2024 4 Connect Data + Metadata + Ontologies Patents / Literature Open Target Omics Ontology 1 (external) Ontology 2 (internal) Terminology 1 (internal) Internal & External Ontologies
- 6. The State of Generative AI
- 7. The Good The State of Generative AI
- 8. The Good The State of Generative AI
- 9. The Good The State of Generative AI The Bad
- 10. The Good The State of Generative AI The Bad
- 11. The Good The Bad The State of Generative AI The Ugly 😱
- 12. The State of Generative AI The Good The Bad The Ugly 😱 GenAI Alone != Right Outcomes 🤯
- 13. Challenges with GenAI: Stochastic Parrot? ● Lack of enterprise domain knowledge ● Inability to verify answers ● Hallucination ● Ethical and data bias concerns ● and more 13 Neo4j Inc. All rights reserved 2024 GenAI PARROT
- 14. 14 Managing AI risk is the biggest barrier to scaling AI initiatives1 Skepticism: Over half of business leaders are skeptical in adopting GenAI.2 Neo4j Inc. All rights reserved 2024 Explainability: Over 80% of executives worry about non-transparent nature of GenAI could result in poor or unlawful decisions.2 Reliability: Inaccuracy and hallucination are two of the most-cited risks of adopting GenAI technology at all levels of an organisation.3 1. Deloitte’s State of AI in the Enterprise 2. BCG’s Digital Acceleration Index Study 2023 3. McKinsey: The state of AI in 2023
- 15. 15 Neo4j Inc. All rights reserved 2024 How can enterprises use domain-specific knowledge to rapidly build accurate, contextual, and explainable GenAI applications? Problem Statement
- 16. Why RAG? And what is it anyway…
- 17. Retrieval Augmented Generation: The ability to dynamically query a large text corpus to incorporate relevant factual knowledge into the responses generated by the underlying language model
- 18. Neo4j Inc. All rights reserved 2023 18 RAG augments LLMs by retrieving up-to-date, contextual external data to inform responses: Retrieve - Find documents of interest for the user question Augment: Combine the user question with the relevant documents Generate: Feed enhanced prompt to an LLM and obtain answer Retrieval Augmented Generation Database of Truth RAG is becoming an industry standard
- 19. Why RAG With Vector Databases Fall Short Similarity is insufficient for rich enterprise reasoning Neo4j Inc. All rights reserved 2024 19 1 3 2 4 Only leverage a fraction of your data: Beyond simple “metadata”, vector databases alone fail to capture relationships from structured data Miss critical context: Struggle to capture connections across nuanced facts, making it challenging to answer multi-step, domain-specific, questions Vector Similarity ≠ Relevance: Vector search uses an incomplete measure of similarity. Relying on it solely can result in irrelevant and duplicative results Lack explainability: The black-box nature of vectors lacks transparency and explainability
- 20. Can Knowledge Graphs help?
- 21. Recap a Knowledge Graph A knowledge graph is a structured representation of facts, consisting of entities, relationships and semantic descriptions 21 Neo4j Inc. All rights reserved 2024 Name: “Her2” receptor: True Domain: “IV” Name: “Her2 positive Breast Cancer“ ICD10: C50 TARGET OF OVEREXPRESSED T R E A T S B I N D S Approved: Sep 1998 Gene Diseas e Drug HAS TARGET Name: “Trastuzumab“ ATC code: “L01FD01” Brand_name: “Herceptin”
- 22. Knowledge Graphs–New & Improved! NOW WITH VECTORS!
- 23. Name: “Her2” receptor: True Domain: “IV” Name: “Her2 positive Breast Cancer“ ICD10: C50 Name: “Trastuzumab“ ATC code: “L01FD01” Embedding: Brand_name: “Herceptin” TARGET OF OVEREXPRESSED T R E A T S B I N D S Approved: Sep 1998 Gene Diseas e Drug HAS TARGET Now with Vectors! Vectors as Node properties = Vector Search + Graph Traversal 23 Neo4j Inc. All rights reserved 2024
- 24. Neo4j Inc. All rights reserved 2023 24 Neo4j - Vector Database Capabilities Vector Search Data Science Knowledge Graph ● Find nodes using an implicit similarity search in the vector index* and enrich with additional explicit relationships from the knowledge graph ● Hybrid Search with text ● Create vectors of network information using node embeddings Now a top 10 vector database on LangChain.
- 25. Neo4j Inc. All rights reserved 2023 25 By 2025, 50% of generative AI initiatives will have improved reliability and transparency by combining deep learning foundation models with knowledge graphs or other composite AI elements. Technological Implications of Generative AI, August 2023 Impact Radar for GenAI (2024)
- 26. From RAG to GraphRAG
- 27. GraphRAG Technique for richly understanding text datasets by combining text extraction, network analysis, LLM prompting and summarization into a single end-to-end system
- 28. Neo4j Inc. All rights reserved 2024 A Neo4j Knowledge Graph combined with LLM’s obtains some unique improvements: Accuracy - Obtain better answers compared to plain vector searches Specificity: domain specific, factual knowledge on your subject Explainability: Provide the user with more reasoning on how the results were obtained. Security: Role Based Access Control Retrieval Augmented Generation Evolving From RAG to GraphRAG
- 29. We are not making this up… Neo4j Inc. All rights reserved 2023 29
- 30. You need a better R…
- 31. Neo4j Inc. All rights reserved 2023 31 Data Science on Graphs: Graph Data Science… Vector Search Graph Data Science Knowledge Graph Bring the context of your connected data into a format that other pipelines can ingest. The Largest Catalog of Graph Algorithms Graph Vector Embeddings for Machine Learning
- 33. Neo4j Inc. All rights reserved 2023 33 GraphRAG with Neo4j Find similar documents and content Identify entities associated to content and patterns in connected data Improve GenAI inferences and insights. Discover new relationships and entities Unify vector search, knowledge graph and data science capabilities to improve RAG quality and effectiveness Vector Search Graph Data Science Knowledge Graph
- 34. An example… Patent Summary GenAI App 34 Neo4j Inc. All rights reserved 2024
- 35. Why a KG Matters in a Patent GenAI App 35 Challenges Outcomes Time consuming to TO FOLLOW Knowledge base to collect, store and retrieve domain-specific information Repetitive and manual tasks to synthesise the content Drive efficient, accurate, contextual and explainable way to streamline Claim extraction Non-standard structure of Patent making it difficult to do data modelling Flexible storage that’s adoptable to the varying structure of an Claims Neo4j Inc. All rights reserved 2024
- 36. Anatomy of a Patent Document 36 Patent A Intro Detailed Description Claim Abstract Background Subsection 1 Subsection 2 Subsection 1.1 Content Content Content Content Content Content Content Content Content Content Neo4j Inc. All rights reserved 2024 Examples
- 37. Anatomy of a Document 37 Patent A Intro Detailed Description Claim Abstract Background Subsection 1 Subsection 2 Subsection 1.1 Content Content Content Content Content Content Content Content Content Content Neo4j Inc. All rights reserved 2024 Examples
- 38. Patent Document as a Graph 38 Patent A Intro Detailed Description Claim Content Chunk Content Chunk Content Chunk Background Abstract Content Chunk Content Chunk Content Chunk Subsection 2 Subsection 1 Content Chunk Content Chunk Subsection 1.1 Content Chunk Content Chunk Content Chunk Vector Embedding Vector Embedding Vector Embedding Vector Embedding Vector Embedding Vector Embedding Vector Embedding Vector Embedding Vector Embedding Vector Embedding Vector Embedding Neo4j Inc. All rights reserved 2024
- 39. Patent Document as a Graph 39 Neo4j Inc. All rights reserved 2024 Patent A Claim Detailed Description Intro Content Chunk Content Chunk Content Chunk Claim 2 Claim 1 Content Chunk Content Chunk Content Chunk Background Abstract Content Chunk Content Chunk Summary Content Chunk Content Chunk Content Chunk Vector Embedding Vector Embedding Vector Embedding Vector Embedding Vector Embedding Vector Embedding Vector Embedding Vector Embedding Vector Embedding Vector Embedding Vector Embedding
- 40. Knowledge Graph as the Knowledge Base Document in a KG Knowledge Graph 40 Neo4j Inc. All rights reserved 2024
- 41. Knowledge Graph as the Knowledge Base Document in a KG Knowledge Graph 41 Neo4j Inc. All rights reserved 2024
- 42. Why Neo4j KG Matters in Patent GenAI App? 42 Challenges Outcomes Time consuming to read previous Knowledge base to collect, store and retrieve domain-specific information Repetitive and manual tasks to synthesise the content Drive efficient, accurate, contextual and explainable way to streamline Patent related work Non-standard structure making it difficult to do data modelling Flexible storage that’s adoptable to the varying structure of an Patents Neo4j Inc. All rights reserved 2024
- 43. Accurate, Contextual and Explainable 43 Neo4j Inc. All rights reserved 2024 GenAI App What is the broadest Claim of Patent A?
- 44. Accurate, Contextual and Explainable 44 Neo4j Inc. All rights reserved 2024 GenAI App What is the broadest Claim of Patent A?
- 45. Accurate, Contextual and Explainable 45 Neo4j Inc. All rights reserved 2024 What is the broadest Claim of Patent A? GenAI App User Question Embedding Model
- 46. Accurate, Contextual and Explainable 46 Neo4j Inc. All rights reserved 2024 GenAI App Embedding Model User Question Vector Embedding What is the broadest Claim of Patent A?
- 47. Accurate, Contextual and Explainable 47 Neo4j Inc. All rights reserved 2024 GenAI App Embedding Model User Question Similarity Search using Neo4j Vector Index Vector Embedding What is the broadest Claim of Patent A?
- 48. Patent A Claim Detailed Descript ion Intro Content Chunk Content Chunk Content Chunk Claim 2 Claim 1 Content Chunk Content Chunk Content Chunk Backgrou nd Abstract Content Chunk Content Chunk Summary Content Chunk Content Chunk Content Chunk Vector Embedd ing Vector Embedd ing Vector Embedd ing Vector Embedd ing Vector Embedd ing Vector Embedd ing Vector Embedd ing Vector Embedd ing Vector Embedd ing Vector Embedd ing Vector Embedd ing Accurate, Contextual and Explainable 48 Neo4j Inc. All rights reserved 2024 Similarity Search using Neo4j Vector Index Claim Related to Protein
- 49. Patent A Claim Detailed Descript ion Intro Content Chunk Content Chunk Content Chunk Claim 2 Claim 1 Content Chunk Content Chunk Content Chunk Backgrou nd Abstract Content Chunk Content Chunk Summary Content Chunk Content Chunk Content Chunk Vector Embedd ing Vector Embedd ing Vector Embedd ing Vector Embedd ing Vector Embedd ing Vector Embedd ing Vector Embedd ing Vector Embedd ing Vector Embedd ing Vector Embedd ing Vector Embedd ing Accurate, Contextual and Explainable 49 Neo4j Inc. All rights reserved 2024 Contextual Knowledge Retrieval within Neo4j KG Reduces Obesity Deploys Semagluti de US clinical trials
- 50. Patent A Claim Detailed Descript ion Intro Content Chunk Content Chunk Content Chunk Claim 2 Claim 1 Content Chunk Content Chunk Content Chunk Backgrou nd Abstract Content Chunk Content Chunk Summary Content Chunk Content Chunk Content Chunk Vector Embedd ing Vector Embedd ing Vector Embedd ing Vector Embedd ing Vector Embedd ing Vector Embedd ing Vector Embedd ing Vector Embedd ing Vector Embedd ing Vector Embedd ing Vector Embedd ing Accurate, Contextual and Explainable 50 Neo4j Inc. All rights reserved 2024 Knowledge Retrieval to aid in Explainability Reduces Obesity Deploys Semaglutide US clinical trials
- 51. Patent A Claim Detailed Descript ion Intro Content Chunk Content Chunk Content Chunk Claim 2 Claim 1 Content Chunk Content Chunk Content Chunk Backgrou nd Abstract Content Chunk Content Chunk Summary Content Chunk Content Chunk Content Chunk Vector Embedd ing Vector Embedd ing Vector Embedd ing Vector Embedd ing Vector Embedd ing Vector Embedd ing Vector Embedd ing Vector Embedd ing Vector Embedd ing Vector Embedd ing Vector Embedd ing Accurate, Contextual and Explainable 51 Neo4j Inc. All rights reserved 2024 Fine Grained Access Control to prevent unwarranted Knowledge Retrieval Response from ABC Company Signed by John Smith He’s the General Manager
- 52. Accurate, Contextual and Explainable 52 Neo4j Inc. All rights reserved 2024 GenAI App Embedding Model User Question Similarity Search using Neo4j Vector Index Vector Embedding What is the broadest Claim of Patent A?
- 53. Accurate, Contextual and Explainable 53 Neo4j Inc. All rights reserved 2024 GenAI App Embedding Model User Question Vector Embedding Similarity Result Similarity Search using Neo4j Vector Index What is the broadest Claim of Patent A? Some LLM
- 54. Accurate, Contextual and Explainable 54 Neo4j Inc. All rights reserved 2024 GenAI App Embedding Model User Question Vector Embedding Similarity Result Similarity Search using Neo4j Vector Index Some LLM What is the broadest Claim of Patent A?
- 55. Accurate, Contextual and Explainable 55 Neo4j Inc. All rights reserved 2024 GenAI App Embedding Model User Question Vector Embedding Similarity Result Similarity Search using Neo4j Vector Index The claim of Patent A is that it achieves a reduction of obesity levels through use of semaglutide as evidenced in US clinical trials Similarity + Contextual Result Some LLM What is the broadest Claim of Patent A?
- 56. How about Graph Data Science? 56 Enrich the measure of relevancy using graph algorithms. ● Page Rank to understand the important of parts of documents, genes, pathways ● Link Prediction to find hidden relationships that further contextualise the results, predict target-> disease relationships ● Community Detection to group related parts of documents for more focused knowledge retrieval Neo4j Inc. All rights reserved 2024
- 57. Try it yourself…-Neo4j Graphbuilder
- 58. But does it REALLY work??
- 59. 59 Neo4j Inc. All rights reserved 2024
- 60. Neo4j Inc. All rights reserved 2023 60
- 61. Let’s Wrap up…
- 62. GraphRAG enables you to..: ● To leverage structural information across entities to enable more precise and comprehensive retrieval ● To perform advanced Graph analytics to enhance retrieval ● To have an accurate conversation with your data that is explainable 62