| We studied tumour biopsies from five women with HR-positive, HER2-negative metastatic breast cancer in the PROMISE study, all treated with first-line palbociclib plus letrozole (a single-arm study with no untreated control). Each patient was biopsied once before starting treatment and once at disease progression, giving ten samples. Our groups are the levels of the `group` column: baseline is the biopsy taken before treatment, and progression the biopsy taken at disease progression on treatment. The `treatment` column holds the same values; it records the biopsy occasion, not a treatment assignment. | |
| The two biopsies are defined by clinical events rather than by a time on a common scale: the interval from baseline to progression differs between patients and is not recorded, and there is no time column. The `patient` column identifies the patient (the sample name without its _baseline or _progression ending), so each patient contributes exactly one baseline and one progression biopsy, both from the same person. Each biopsy was made into its own library and sequenced once, so there are no technical replicates. No batch or library-preparation information is recorded, and we know of no other variable that matters; the remaining columns are GEO annotations. | |
| The data are bulk RNA-seq of frozen tumour tissue, given as gene-level raw read counts from the MAP-RSeq pipeline (GRCh38, Ensembl gene IDs). | |
| Our question is which genes differ in tumour expression at disease progression compared with baseline in the same patients, that is, the progression biopsies compared with the baseline biopsies. All ten samples are part of this question. | |