[Paper Review] Lomics: Generation of Pathways and Gene Sets using Large Language Models for Transcriptomic Analysis

Chun‐Ka Wong, Ali Choo+12 more|arXiv (Cornell University)|Jul 12, 2024

Genomics and Phylogenetic Studies4 citations

TL;DR

Lomics is a Python-based bioinformatics toolkit that leverages large language models (LLMs) to generate custom, biologically relevant gene sets and pathways tailored to specific transcriptomic research questions. It streamlines pathway selection, gene set generation, and validation through LLM-driven reasoning, JSON validation, and HGNC gene symbol checks, producing standardized .GMX files for downstream analysis with improved statistical power due to reduced multiple testing.

ABSTRACT

Interrogation of biological pathways is an integral part of omics data analysis. Large language models (LLMs) enable the generation of custom pathways and gene sets tailored to specific scientific questions. These targeted sets are significantly smaller than traditional pathway enrichment analysis libraries, reducing multiple hypothesis testing and potentially enhancing statistical power. Lomics (Large Language Models for Omics Studies) v1.0 is a python-based bioinformatics toolkit that streamlines the generation of pathways and gene sets for transcriptomic analysis. It operates in three steps: 1) deriving relevant pathways based on the researcher's scientific question, 2) generating valid gene sets for each pathway, and 3) outputting the results as .GMX files. Lomics also provides explanations for pathway selections. Consistency and accuracy are ensured through iterative processes, JSON format validation, and HUGO Gene Nomenclature Committee (HGNC) gene symbol verification. Lomics serves as a foundation for integrating LLMs into omics research, potentially improving the specificity and efficiency of pathway analysis.

Motivation & Objective

• To address the limitation of traditional pathway enrichment analyses, which rely on fixed, large gene set libraries that increase multiple hypothesis testing.
• To enable researchers to generate small, targeted gene sets and pathways aligned with their specific biological questions using LLMs.
• To improve the efficiency and specificity of transcriptomic data analysis by reducing noise and false positives from broad pathway libraries.
• To provide a reproducible, validated workflow for generating gene sets with explanations for pathway selection decisions.
• To integrate LLMs into systems biology workflows through a user-friendly, open-source toolkit (Lomics v1.0) for omics studies.

Proposed method

• Lomics employs a three-step pipeline: (1) LLM-based derivation of relevant pathways from the user's scientific question.
• It generates valid, biologically coherent gene sets for each selected pathway using LLM reasoning and gene symbol validation.
• All outputs are validated using JSON schema checks to ensure structural correctness.
• Gene symbols are cross-verified against the HUGO Gene Nomenclature Committee (HGNC) database to ensure nomenclature accuracy.
• The toolkit outputs results in .GMX format, a standard for gene set files used in pathway analysis tools.
• Explanations for pathway and gene set selections are generated by the LLM to enhance interpretability and reproducibility.

Experimental results

Research questions

• RQ1 Can LLMs effectively generate biologically relevant, custom gene sets tailored to specific transcriptomic research questions?
• RQ2 How can LLMs improve the specificity and statistical power of pathway enrichment analysis by reducing the size of gene set libraries?
• RQ3 To what extent can LLMs provide interpretable and justifiable explanations for pathway and gene set selections in omics studies?
• RQ4 Can an LLM-driven workflow maintain consistency and accuracy in gene set generation through validation mechanisms like JSON schema and HGNC checks?
• RQ5 How does the integration of LLMs into bioinformatics pipelines enhance the efficiency and usability of transcriptomic data analysis?

Key findings

• Lomics successfully generates small, targeted gene sets and pathways that are specifically aligned with the user's research question, reducing the number of hypotheses tested.
• The toolkit ensures high accuracy in gene symbol usage through integration with the HGNC database, minimizing nomenclature errors.
• All outputs are validated via JSON schema checks, ensuring structural integrity and reproducibility of the generated gene sets.
• The LLM provides human-readable explanations for pathway and gene set selections, improving transparency and interpretability.
• Lomics outputs are compatible with standard omics analysis tools through the .GMX file format, enabling seamless integration into existing workflows.
• The framework demonstrates the feasibility of using LLMs to enhance specificity and efficiency in transcriptomic pathway analysis.