D2H2

D2H2 aggregates curated transcriptomics datasets from the Gene Expression Omnibus (GEO) and provides analytical modules for differential gene expression, single-gene queries, hypothesis generation from bulk RNA-seq signatures, and GPT-driven automated tool invocation for diabetes omics research.

Key Features:

• Curated Transcriptomics Datasets: Hosts hundreds of diabetes-relevant transcriptomics datasets curated from GEO.
• Data Visualization and Analysis: Provides dataset-associated visualization options, differential gene expression analysis, and single-gene query capability.
• GPT Chatbot Integration: Uses GPT to parse free-text queries and invoke specified bioinformatics tools via their APIs based on provided tool and workflow information.
• Hypotheses Generation Module: Randomly selects gene sets from precomputed bulk RNA-seq signatures, identifies highly overlapping gene sets by comparison with gene sets extracted from PubMed Central while accounting for abstract dissimilarity, and uses GPT to speculate on potential explanations for observed overlaps.

Scientific Applications:

• Gene Expression Studies: Enables differential gene expression analyses to identify genes associated with diabetic conditions.
• Pathway Analysis: Facilitates exploration of gene sets and pathways implicated in diabetes.
• Hypothesis Generation: Identifies overlapping gene sets across studies to support formulation of hypotheses about genetic interactions and regulatory networks in diabetes.

Methodology:

Integration of curated GEO transcriptomics datasets with analysis modules; GPT-based natural language processing to parse free-text queries and automate tool invocation via APIs; hypothesis generation by comparing randomly selected precomputed bulk RNA-seq signature gene sets with gene sets extracted from PubMed Central while considering abstract dissimilarity and using GPT to generate explanatory speculation.