GSM8006584 names its antibody KMT2A-C1 in its title, its characteristics field and its own filename, and KMT2A-C2 in its description - so keying the antibody off the free-text description gives patient TB-92-0488 two C2 libraries and no C1

Audited by comparing the antibody named in `!Sample_title` against the antibody named in `!Sample_description` for all 171 records. **Fifteen records differ; fourteen of the fifteen are formatting only** (`AFF1_1` vs `AFF1-1`, `K4me3` vs `H3K4me3`, a double space). **One is a content disagreement** [E1]: ``` GSM8006584 title TB920488_KMT2A_C1_(220412_JS_Hs_032922_JS2165) desc AutoCUTnRUN for KMT2A-C2 in TB-92-0488... chars sample: Mouse KMT2A-C1 SantaCruz H10; TB-92-0488; KMT2A-AFF1... file GSM8006584_TB920488_KMT2A_C1.hg38.bw ``` **Three independent fields say C1, one says C2, and the adjacent record GSM8006585 is a genuine C2** with the matching `Mouse KMT2A-C2 Millipore 05-765` reagent. The description is the typo. **What it would cost.** TB-92-0488 is one of the eight KMT2A::AFF1 patients. A pipeline keying the antibody off `!Sample_description` assigns it two C2 libraries and no C1, so the N-over-C denominator for that patient is built from one reagent instead of two - and **nothing errors**, because two C libraries are still present. The oncoprotein score for one of eight index-fusion patients would quietly become a different measurement from the other seven. **The generalisable half:** when a deposit's fields disagree, **count how many independent fields support each reading before choosing**. Here the vote is 3-1 and the losing field is the only free-text one. A single-field read - in either direction - would have been a coin flip dressed as a lookup.