SPIREX - Improving LLM-based relation extraction from RNA-focused scientific literature using graph machine learning

Aug 5, 2026

Relation extraction from scientific literature aligned with a domain ontology is a major challenge in natural language processing, especially for precision medicine applications. Large language models (LLMs) offer new solutions, but extracted triples can suffer from errors such as hallucinations. SPIREX extends the SPIRES-based framework for extracting triples about RNA molecules by combining schema-constrained prompt formulation with graph machine learning over an RNA knowledge graph (RNA-KG).

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages11

Reading Time28 min

AccessFree to download

Abstract

Introduction

RNA knowledge graph (RNA-KG)

Motivation for relation extraction from biomedical text

Limitations of supervised RE and LLM hallucinations

Motivation for LLM-KG integration

SPIREX: Improving LLM-based relation extraction from RNA-focused scientific literature using graph machine learning Emanuele Cavalleri University of Milan, Italy [emanuele.cavalleri@unimi.it](emanuele.cavalleri@unimi.it) Mauricio Soto-Gomez University of Milan, Italy [mauricio.soto@unimi.it](mauricio.soto@unimi.it) Ali Pashaeibarough University of Milan, Italy [ali.pashaeibarough@unimi.it](ali.pashaeibarough@unimi.it) Dario Malchiodi University of Milan, Italy [dario.malchiodi@unimi.it](dario.malchiodi@unimi.it) Harry Caufield Lawrence Berkeley National Lab, USA [jhc@lbl.gov](jhc@lbl.gov) Justin Reese Lawrence Berkeley National Lab, USA [justaddcoffee@gmail.com](justaddcoffee@gmail.com) Christopher J. Mungall Lawrence Berkeley National Lab, USA [CJMungall@lbl.gov](CJMungall@lbl.gov) Peter N. Robinson Charité University, Berlin, Germany [peter.robinson@bih-charite.de](peter.robinson@bih-charite.de) Elena Casiraghi University of Milan, Italy [elena.casiraghi@unimi.it](elena.casiraghi@unimi.it) Giorgio Valentini University of Milan, Italy [giorgio.valentini@unimi.it](giorgio.valentini@unimi.it) Marco Mesiti University of Milan, Italy marco.mesiti@unimi.it ABSTRACT Relation extraction from scientific literature to align with a domain ontology is a well-known challenge in natural language processing, particularly critical in precision medicine. The advent of large language models (LLMs) has enabled the development of new and effective approaches to this problem. However, the extracted relations can be prone to problems (e.g., hallucination) that must be minimized. In this paper, we present the initial development of SPIREX, an extension of the SPIRES-based system designed to extract triples from scientific literature involving RNA molecules. Our system leverages schema constraints in the formulation of LLM promptsand utilizes graph machine learning on our RNA-based knowledge graph, RNA-KG, to assess the plausibility of the extracted triples. RNA-KG comprises more than 12.5M edges representing various types of relationships involving RNA molecules. VLDB Workshop Reference Format: Emanuele Cavalleri, Mauricio Soto-Gomez, Ali Pashaeibarough, Dario Malchiodi, Harry Caufield, Justin Reese, Christopher J. Mungall, Peter N. Robinson, Elena Casiraghi, Giorgio Valentini, and Marco Mesiti. SPIREX: Improving LLM-based relation extraction from RNA-focused scientific literature using graph machine learning. VLDB 2024 Workshop: LLM+KG. VLDB Workshop Artifact Availability: The experiments have been realized by using the following datasets: (schema, docs, and manual annotations) [https://zenodo.org/records/11393776](https://zenodo.org/records/11393776); (RNA KG): [https://zenodo.org/doi/10.5281/zenodo.10078876](https://zenodo.org/doi/10.5281/zenodo.10078876) . This work is licensed under the Creative Commons BY-NC-ND 4.0 International License. Visit [https://creativecommons.org/licenses/by-nc-nd/4.0/ to view a copy of](https://creativecommons.org/licenses/by-nc-nd/4.0/ to view a copy of)[ ](https://creativecommons.org/licenses/by-nc-nd/4.0/ to view a copy of)[this license. For any use beyond those covered by this license](this license. For any use beyond those covered by this license), [obtain permission by](obtain permission by)[emailing info@vldb.org. Copyright](emailing info@vldb.org. Copyright) is held by the owner/author(s). Publication rights licensed to the VLDB Endowment. Proceedings of the VLDB Endowment. ISSN 2150-8097 . 1 INTRODUCTION Ribonucleic acid (RNA) is essential in the central dogma of molecular biology, functioning as the intermediary between DNA and proteins, the fundamental building blocks of life. Beyond its traditional role in protein synthesis, RNA is involved in various cellular processes, including gene regulation and catalysis, emphasizing its critical importance in understanding the complexities of biological systems. RNA-KG [8] is an ontology-based knowledge

FAQs

What problem does SPIREX address in biomedical NLP?

SPIREX targets relation extraction from scientific literature so the extracted relations align with a domain ontology, a challenge that is crucial for precision medicine.

How does SPIREX use LLMs to extract RNA-related triples?

It formulates LLM prompts using schema constraints to guide the extraction process toward ontology-compatible triples.

How does graph machine learning improve the quality of extracted relations?

SPIREX uses graph machine learning on the RNA knowledge graph (RNA-KG) to assess the plausibility of the extracted triples, helping reduce issues like hallucination.

Aug 5, 2026

Upload by Seraphina