Bridging Large Language Models and Single-Cell Transcriptomics in Dissecting Selective Motor Neuron Vulnerability

TL;DR

This paper introduces a novel framework that combines gene annotations and large language models to generate biologically meaningful cell embeddings from single-cell RNA sequencing data, enhancing interpretability in cell analysis.

Contribution

It presents a new multimodal approach that integrates gene descriptions with language models to improve cell-type clustering and vulnerability analysis in single-cell transcriptomics.

Findings

Enhanced cell clustering accuracy

Improved interpretability of cell vulnerability

Effective integration of biological annotations with language models

Abstract

Understanding cell identity and function through single-cell level sequencing data remains a key challenge in computational biology. We present a novel framework that leverages gene-specific textual annotations from the NCBI Gene database to generate biologically contextualized cell embeddings. For each cell in a single-cell RNA sequencing (scRNA-seq) dataset, we rank genes by expression level, retrieve their NCBI Gene descriptions, and transform these descriptions into vector embedding representations using large language models (LLMs). The models used include OpenAI text-embedding-ada-002, text-embedding-3-small, and text-embedding-3-large (Jan 2024), as well as domain-specific models BioBERT and SciBERT. Embeddings are computed via an expression-weighted average across the top N most highly expressed genes in each cell, providing a compact, semantically rich representation. This multimodal strategy bridges structured biological data with state-of-the-art language modeling, enabling more interpretable downstream applications such as cell-type clustering, cell vulnerability dissection, and trajectory inference.