LA-MARRVEL: A Knowledge-Grounded and Language-Aware LLM Reranker for AI-MARRVEL in Rare Disease Diagnosis

2025-11-04

AI Summary

In rare disease diagnosis, clinical text-based phenotypic descriptions are highly complex, and candidate gene lists suffer from high false-positive rates; current workflows rely heavily on labor-intensive manual curation. To address this, we propose a knowledge-anchored, language-aware LLM re-ranking framework that integrates expert-curated contextual enhancements with a multi-round LLM voting consensus mechanism—enabling robust, interpretable gene prioritization and automatic generation of natural-language reasoning justifications. Our method synergistically combines AI-MARRVEL’s initial screening, structured genomic annotations, and semantic analysis of unstructured biomedical literature. Evaluated on a real-world patient cohort, it achieves a Recall@5 of 94.10%, outperforming AI-MARRVEL by 3.65 percentage points and significantly surpassing Exomiser, LIRICAL, and direct Claude invocation. The core contribution is a knowledge-guided, collaborative LLM inference architecture that jointly optimizes accuracy, robustness, and clinical interpretability.

Technology Category

• Machine Learning: Learning Preferences or Rankings
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Uses of LLMs and GenAI for marketplace design, bidding, and strategic interactions

Abstract

Diagnosing rare diseases requires linking gene findings with often unstructured reference text. Current pipelines collect many candidate genes, but clinicians still spend a lot of time filtering false positives and combining evidence from papers and databases. A key challenge is language: phenotype descriptions and inheritance patterns are written in prose, not fully captured by tables. Large language models (LLMs) can read such text, but clinical use needs grounding in citable knowledge and stable, repeatable behavior. We explore a knowledge-grounded and language-aware reranking layer on top of a high-recall first-stage pipeline. The goal is to improve precision and explainability, not to replace standard bioinformatics steps. We use expert-built context and a consensus method to reduce LLM variability, producing shorter, better-justified gene lists for expert review. LA-MARRVEL achieves the highest accuracy, outperforming other methods -- including traditional bioinformatics diagnostic tools (AI-MARRVEL, Exomiser, LIRICAL) and naive large language models (e.g., Anthropic Claude) -- with an average Recall@5 of 94.10%, a +3.65 percentage-point improvement over AI-MARRVEL. The LLM-generated reasoning provides clear prose on phenotype matching and inheritance patterns, making clinical review faster and easier. LA-MARRVEL has three parts: expert-engineered context that enriches phenotype and disease information; a ranked voting algorithm that combines multiple LLM runs to choose a consensus ranked gene list; and the AI-MARRVEL pipeline that provides first-stage ranks and gene annotations, already known as a state-of-the-art method in Rare Disease Diagnosis on BG, DDD, and UDN cohorts. The online AI-MARRVEL includes LA-MARRVEL as an LLM feature at https://ai.marrvel.org . We evaluate LA-MARRVEL on three datasets from independent cohorts of real-world diagnosed patients.

Problem

• Diagnosing rare diseases requires linking gene variants to clinical evidence manually
• Existing tools struggle to rank causal genes accurately in diagnosis pipelines
• Clinical decisions need interpretable gene rankings with integrated evidence reasoning

Innovation

• Knowledge-grounded reranking layer enhances gene prioritization
• LLM queries with ranked voting aggregate partial rankings
• Generates interpretable reasoning integrating phenotypic and variant evidence