Reference panels flip ancestry labels

An individual’s inferred ancestry can change dramatically depending on which ancient reference samples and modeling choices are used: the author’s 23andMe data fits a four‑person Roman/Italic source as '100% Roman' under one panel but reveals different mixtures under alternative comparisons. This shows ancestry assignments are not immutable facts but outputs of specific datasets and statistical tests. — Public and political claims that rest on genetic ancestry (identity, entitlement, migration narratives) are fragile because they depend on reference selection and model design.