GATHER: Convergence-Centric Hyper-Entity Retrieval for Zero-Shot Cell-Type Annotation

2026-05-07

AI Summary

Existing knowledge graph–based zero-shot single-cell type annotation methods suffer from poor scalability and high large language model (LLM) invocation costs when handling hyper-entity queries composed of dozens to hundreds of genes, as they rely on local, gene-by-gene reasoning. This work proposes GATHER, a novel topology-centric hyper-entity retrieval framework that operates on a self-constructed cell-centric knowledge graph (VCKG). GATHER identifies convergence nodes reachable from multiple input genes through global, multi-source graph traversal, leveraging these compact, information-rich nodes as collaborative evidence to replace conventional local inference paradigms. Notably, the method eliminates LLM involvement during retrieval and achieves exact-match accuracies of 27.45% and 59.64% on the Immune and Lung datasets, respectively, while requiring only a single LLM call per sample—substantially outperforming existing KG-RAG approaches that demand 2 to 61 calls.

Abstract

Zero-shot single-cell cell-type annotation aims to determine a cell's type from a given set of expressed genes without any training. Existing knowledge-graph-based RAG approaches retrieve evidence by expanding from source entities and relying on iterative LLM reasoning. However, in this setting each query contains tens to hundreds of genes, where no single gene is decisive and the label emerges only from their collective co-occurrence. Such hyper-entity queries fundamentally challenge local, entity-wise exploration strategies, which reason from individual genes, leading to poor scalability and substantial LLM cost. We propose GATHER (Graph-Aware Traversal with Hyper-Entity Retrieval), a convergence-centric retriever tailored to hyper-entity queries. It performs global multi-source graph traversal and identifies topological convergence points -- nodes jointly reachable from many input genes. These convergence nodes act as high-information hyper-entities that capture entity synergy. By incorporating node- and path-importance scoring, GATHER selects informative evidence entirely without LLM involvement during retrieval. Instantiated on a self-constructed cell-centric biological knowledge graph (VCKG), GATHER outperforms strong KG-RAG baselines (ToG, ToG-2, RoG, PoG) on two datasets (Immune and Lung), achieving the highest exact-match accuracy (27.45% and 59.64%) with only a single LLM call per sample, compared to 2--61 calls for KG-RAG baselines. Our results demonstrate that convergence nodes compress multi-entity signals into compact, high-information evidence that conveys more per item than multi-hop paths, providing an efficient global alternative to local entity-wise reasoning.

Problem

• zero-shot cell-type annotation
• hyper-entity queries
• knowledge graph retrieval
• single-cell genomics
• entity synergy

Innovation

• hyper-entity retrieval
• convergence-centric retrieval
• zero-shot cell-type annotation
• knowledge graph reasoning
• global graph traversal