Understanding Maintenance and Support in a Community-Driven Scientific Workflow Ecosystem: A Cross-Space Study of Galaxy

Abstract

The study investigates maintenance and support in the Galaxy scientific workflow ecosystem across three complementary spaces: GitHub issues, GitHub pull requests, and the Galaxy Community Help Forum. The final dataset includes 11,762 GitHub issues, 52,203 pull requests, and 6,235 Community Forum discussions collected from Galaxy-related repositories and support spaces. The study addresses three research questions: RQ1: What types of maintenance and support concerns emerge across Galaxy GitHub issues, pull requests, and Community Forum discussions? RQ2: What factors are associated with resolution outcomes and resolution time across these artifact types? RQ3: How are maintenance and support artifacts connected across repository-centered development spaces and community-centered support spaces? For RQ1, the package includes the preprocessing and topic-modeling workflow used to analyze maintenance and support concerns. BERTopic was applied separately to issues, pull requests, and forum discussions using Sentence Transformer embeddings, UMAP dimensionality reduction, HDBSCAN clustering, and class-based TF–IDF topic representation. The resulting models identified nine issue topics, 14 pull-request topics, and 14 forum topics. For RQ2, the package contains scripts and analysis outputs for examining factors associated with resolution outcomes and resolution time. The analyses include descriptive statistics, Fisher’s exact tests, Mann–Whitney U tests, effect-size estimation, multivariable logistic regression, Kaplan–Meier survival analysis, log-rank tests, and Cox proportional hazards models. Issues, pull requests, and forum discussions are analyzed separately to reflect their distinct resolution processes. For RQ3, the replication materials include scripts for identifying explicit cross-artifact references, generating candidate semantic connections, manually validating candidate links, and analyzing recurring technical signals across development and support spaces. Candidate semantic connections were generated using TF–IDF representations, cosine similarity, temporal filtering, and top-ranked candidate selection. The package is intended to support transparency, reproducibility, and further research on software maintenance, scientific workflow systems, developer collaboration, community support, and cross-space traceability. Researchers may use the materials to reproduce the analyses reported in the paper, inspect the derived datasets and statistical outputs, or extend the study to additional software ecosystems and communication channels.