An embedding-based statistical framework is developed that transforms gene and function descriptions into vector representations, enabling statistical testing of gene-gene and gene-function relationships and quantitative prioritization of de novo functional hypotheses inferred by LLMs.
Abstract
Emerging large language models (LLMs) can infer gene functions directly from gene lists, enabling hypothesis generation without predefined gene sets. However, these LLM-derived predictions are qualitative, and principled statistical validation is lacking. Here, we develop an embedding-based statistical framework that transforms gene and function descriptions into vector representations, enabling statistical testing of gene-gene and gene-function relationships and quantitative prioritization of de novo functional hypotheses inferred by LLMs. We benchmark seven state-of-the-art embedding models using curated and retrieval-augmented literature-derived gene descriptions across diverse biological contexts. OpenAI’s text-embedding-3-large and Google’s gemini-embedding-001 perform best, capturing gene-gene functional relationships in 88.7-92.5% of Gene Ontology biological processes and approximately 98.6% of canonical pathways. In gene-function association analyses, these models achieve high sensitivity (95.2-98.4%) and specificity (72.7-84.3%). Through contamination analysis and evaluation using experimentally informed protein assembly gene sets, our framework distinguishes biologically meaningful LLM-inferred hypotheses from noise, outperforming confidence-based inference and conventional enrichment analysis. We further develop the open-source R package DEGEmbedR and demonstrate its utility for interpreting a drug perturbation-derived differentially expressed gene (DEG) signature lacking significant conventional enrichment results. Together, these results establish LLM-derived embeddings as a quantitative foundation for functional genomics and the statistical validation of LLM-based gene function inference. Whilst large language models can infer gene functions from gene lists, these predictions lack validation. Here the authors develop a framework to transform gene and function descriptions into vector representations, enabling statistical testing of gene-gene and gene-function relationships.
The Gene Ontology (GO) is a public resource that describes gene functions and characteristics through a structured vocabulary of standardised terms. It currently contains annotations for over 1.5 million gene products, each linked to one or more GO terms. In this study, we propose integrating this GO-based semantic structure into machine learning systems for medical diagnostics. This approach serves a dual purpose: first, to prioritise genes that are semantically relevant to a given clinical task, thereby refining model input; and second, to enable the analysis of biologically predefined gene sets, which may reveal novel mechanisms underlying disease.
Evaluated across 16 benchmark data sets spanning diverse medical domains, our GO term-informed gene selection method generally outperformed models trained on full gene sets. Further analysis of individual GO terms not only enhanced classification performance but also identified high-performing, task-specific gene subsets that were overlooked during initial gene selection.
Our findings demonstrate that Gene Ontology can be effectively leveraged for semantics-aware gene selection in clinical machine learning. Moreover, systematically evaluating individual GO terms offers a scalable strategy to uncover new, testable biological hypotheses–revealing gene functions that might otherwise remain hidden when examining only broadly selected gene combinations.
Peter Eckhardt-Bellmann, N. Taha, Silke D. Werle et al.· BMC Bioinformatics· 0 citations
Across the included studies, stronger evidence for gLM utility was generally associated with biologically informed or task-aligned model design, including evolutionary alignments, motif-aware objectives, long-context architectures, RNA structural priors, population-aware representations, and domain-specific pretraining.
Mahinaz A. Mashhour, Manal Abdel Wahed, Mai S. Mabrouk· Biochemical and Biophysical...· 0 citations
This study systematically assessed the biological knowledge encoded in publicly available LLMs for structured phenotype assignment of microbial species, including state-of-the-art models such as Claude Sonnet 4 and the GPT-5 family of models.
Philipp C. Münch, Nasim Safaei, René Mreches et al.· Genome Biology· 0 citations
Genome-wide association studies and genomic prediction are fundamental for investigating complex traits, but they have different objectives and are rarely unified within a shared analytical framework. Although machine learning has broadened the applicability of both lines of research, their combined use in detecting gene-environment interactions remains underexplored. This study presents a novel statistical framework, iSVR, that incorporates gene-environment interaction terms into a support vector regression model, enabling both modeling of interaction effects and their statistical testing. By formulating a score test based on M-estimation theory within this framework, the iSVR facilitates robust detection of gene-environment interactions while accommodating complex genotype-phenotype relationships. Extensive simulations demonstrate that the iSVR effectively controls the type I error rate and attains competitive or improved statistical power relative to existing methods under the investigated scenarios. Application to soybean and GAW19 datasets further highlights the iSVR's ability to accurately predict trait values and identify significant gene-environment interactions. Collectively, these findings illustrate that the unification of association testing and predictive modeling within a common statistical framework provides a powerful approach to characterize the gene-environment interaction landscapes underlying complex traits.
Protein language models have been increasingly successful on tasks ranging from fitness prediction to functional design, yet what biological knowledge they acquire and where it is encoded within their internal representations remain underexplored. Through a high-resolution layer-by-layer interpretability analysis of 8 models from the ESM2 and AMPLIFY families on 22 concepts from human proteome annotations, we found that these models encode concepts of increasing levels of complexity along their depth: basic physicochemical properties and linear motifs are best captured by early-layer embeddings, secondary structure from subsequent layers, and domain-level semantics from middle layers. Principal component projections of these embeddings showed that they separate biologically meaningful protein groupings, and molecular-biology-inspired interventions demonstrated that pLM embeddings can discriminate phosphomimic-active from inactive mutants. Perhaps surprisingly, we observed that pretraining data and compute had a greater impact on the linear emergence of biological concepts than scaling up parameters. By revealing where biological knowledge is captured in pLMs and which choices shape its emergence, our work offers insights to develop more robust, biologically grounded protein language models.
Shawn T. Whitfield, Tom Marty, Robert M. Vernon et al.· bioRxiv· 0 citations
Gene Ontology (GO) enrichment analysis is a foundational tool for translating large-scale genomic data into biological insights, but typically yields hundreds of redundant terms that obscure overarching themes. Existing summarization tools rely on fixed similarity metrics (REVIGO, GOSemSim, clusterProfiler::simplify()), gene-overlap measures (Metascape), or static hierarchy mappings (GO-slim), and therefore cannot incorporate biological context. Manual curation provides context-aware grouping but is subjective and labor-intensive. A scalable, context-aware framework is needed to cluster GO terms into interpretable higher-order biological domains. Here we present LLMBDC (Large Language Model for Biological Domains Oriented Clustering of Gene Ontology), a training-free framework that leverages zero-shot semantic reasoning of LLMs with confidence scoring to cluster GO terms into BioDomains using only ontology information at inference time. Benchmarked across Alzheimer's disease (AD) and Fragile X syndrome (FXS) against six baseline methods including SapBERT, LLMBDC achieved substantially higher precision, recall, and clustering performance. Against ground-truth annotations, LLMBDC improved ARI from 9.7% to 73.3% (AD) and from 15.7% to 66.6% (FXS) over REVIGO, with corresponding NMI gains from 59.9% to 73.4% (AD) and 66.0% to 79.5% (FXS). A Cauchy combination test further confirmed that aggregated BioDomains retained statistically significant functional signals. LLMBDC provides a scalable, reproducible, and interpretable route to context-aware, system-level interpretation of GO enrichment results while preserving biological specificity.