Differential expression

limma-voom on the two replicated designs GanoDB holds. C/N starvation — treatment vs control at days 2, 4, 6 and 8, three biological replicates per arm. In planta vs axenic culture — PRJNA514399, three biological replicates each; this contrast was added in 2026-08 after the NCBI BioSample records showed the project is an infection experiment, not the six-wood-species panel it had been catalogued as.

The 7 and 21 dpi infection libraries remain n=1 per timepoint, so they carry no p-values and appear as TPM on each gene page rather than here. Their RM0 companion is the axenic inoculum reference, not an in planta sample — it assigns 49.2% of its reads to G. boninense, against 2.8% and 4.4% for the two genuine infection libraries.

Library yield varies more than fourteenfold across the resource and is shown on every row: 47–52% of alignments fall on G. boninense genes in the axenic C/N panel, 3.2–4.8% in the axenic arm of PRJNA514399, and 0.66–0.77% in planta. voom weights unequal library sizes, but a fold change resting on a few hundred thousand assigned reads is weaker evidence than one from the C/N panel. Shallow libraries are excluded from the coexpression network for the same reason, so correlations are not built out of near-empty columns. Neither contrast uses TMM normalisation — edgeR is not installed in this environment.

Database is empty — pipeline output not loaded yet.