Multi-Layer Knowledge Graph for CMC Process Development
From Document Silos to Process Intelligence: A Multi-Layer Knowledge Graph for CMC Process Development
This paper details a novel approach to structuring and querying complex, fragmented technical documents, which is relevant for anyone working with large, unstructured data in fields like pharmaceuticals or advanced manufacturing.
This research paper presents a modular agentic-AI platform that converts heterogeneous CMC process-development documents into a queryable, dual-layer knowledge graph.
A base knowledge layer builds a lexical graph with a Document-Section-Chunk hierarchy through lossless ingestion of various document types, while an intelligence layer extracts ontology-aligned entities and bridges cross-document concepts.
The system uses LLM agents to select the best retrieval path for each question.It was evaluated on 505 questions from 38 Sanofi reports, achieving a Tier-1 multiple-choice accuracy of 95% and a Tier-2 LLM-judge pass rate of 85%.
From Document Silos to Process Intelligence: A Multi-Layer Knowledge Graph for CMC Process Development
Chemistry, Manufacturing and Controls (CMC) process development generates an enormous body of technical information across a multi-stage, knowledge-intensive continuum from drug discovery to commercial manufacturing. This knowledge is traditionally fragmented across functions and heterogeneous formats, causing traceability gaps and significant knowledge-management costs during technology transfer and regulatory filing. We present a modular agentic-AI platform that converts a heterogeneous corpus of process-development documents into a queryable, dual-layer knowledge graph. A base knowledge layer builds a lexical graph with a Document-Section-Chunk hierarchy through lossless ingestion of digital, scanned, handwritten, and multilingual documents, while an intelligence layer extracts ontology-aligned entities and bridges cross-document concepts through a provenance-anchored domain graph. LLM agents operate across both layers, selecting the retrieval path best suited to each question. We evaluate the lexical layer with a novel three-tier protocol measuring the deployment-fidelity of a retrieval-augmented generation (RAG) system on proprietary data, demonstrated on 505 questions curated from 38 development reports of a Sanofi small-molecule program. Tier-1 multiple-choice accuracy of 95% signals strong platform reliability; the stricter Tier-2 LLM-judge pass rate of 85%, which degrades on comparative and corpus-wide questions, reveals a failure taxonomy that Tier-1 accuracy alone fails to capture. A router agent selects between layers according to question type. We anticipate this protocol will enable future designers of agentic platforms to assess their systems against nonpublic databases, and that graph-based architectures will see broader adoption in pharma as a means of transforming fragmented document repositories into structured process intelligence.
This report is categorized under AITOP's "Papers" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include knowledge graphs, RAG, LLM agents, pharma AI.
Source attribution: curated from arXiv cs.AI (@Reza Amirmoshiri, Faryad Sahneh, Yasser Jangjou).
