| We co-cultured KYSE150 human esophageal squamous cell carcinoma cells for 48 hours with 60 µg/ml purified recombinant FadA, an adhesin of Fusobacterium nucleatum, or without FadA, and harvested all cultures at that single time point, so time is not a variable here. The nine cultures are named in `sample_id`: FadA_1 to FadA_3, and two separate control sets without FadA, 0M_1 to 0M_3 and NC_1 to NC_3. GEO records both control sets as treatment "control" and the protocol names PBS as the control, but does not say how 0M and NC differ. | |
| The `group` column pools the two control sets: "FadA_treated" (3 cultures) against "control" (the three 0M and three NC cultures); `treatment` repeats it. The `condition` column keeps the three labels from the sample names ("FadA", "0M", "NC"), and `genotype` holds placeholder labels from GEO ("WT1" for 0M, "WT2" for NC, "control" for FadA), not real genotypes. | |
| Each library comes from its own culture in a 6-well plate and was sequenced once, so no two samples share a culture and there are no technical replicates; the numbers 1 to 3 only count the cultures within each condition. The GEO batch field in `src_batch` holds the same value ("7/11/2025") for every sample, so there is no known batch variable. We have no other variable we consider relevant; the remaining columns are GEO annotations. | |
| The data are bulk mRNA-seq aligned with HISAT2 to GRCh38. The matrix holds raw gene-level read counts for 20,422 Ensembl genes plus 1,327 rows for novel or transcript-level entries. | |
| Our question is which genes are differentially expressed in FadA-treated cells compared with control cells, where the control is the average of the two control sets, 0M and NC, weighted equally. In `group` this corresponds to "FadA_treated" compared with "control", which holds three cultures from each control set; all nine samples are part of the analysis. | |