Interview-Driven Workflow Construction and Repair in Galaxy

Workflow Debug Report: Ebola Virus Infection RNA-seq Analysis

Failure Classification

Evidence Captured

Diagnosis

  1. Salmon Step: Succeeded completely on both samples (SRR5085167, SRR5085168), generating valid transcript abundance tables (quant.sf) via biocontainers.
  2. DESeq2 Step: Succeeded in reading Salmon quant.sf files via tximport, but failed when instantiating the DESeqDataSet object.
  3. Cause of Failure: The workflow’s tool_state for DESeq2 set design_formula: "~ Genotype + Timepoint + Condition". However, the test dataset in sample_metadata.tabular only provided 2 samples (SRR5085167 and SRR5085168), both of which had Genotype=WT and Timepoint=24. In DESeq2, every variable in the design formula must have at least two levels across the sample set; otherwise, the design matrix contains invariant columns that cannot be estimated.
  1. Set the workflow’s default design formula in galaxy-workflow.gxwf.yml to ~ Condition (the primary biological contrast: Mock vs Ebola infection).
  2. Expand the test fixture in ebola-test/test-data/ to include 4 samples (2 Mock replicates and 2 Ebola replicates), providing sufficient statistical degrees of freedom for DESeq2 dispersion fitting.
  3. Update galaxy-workflow.gxwf-tests.yml to include the 4 samples in reads_paired_collection and sample_metadata.tabular.