Huge thanks again to Camellia Rui, co-authors, and my amazing advisors, @charleston.bsky.social & @nmancuso.bsky.social for their support and guidance.

Applying HAMSTA to self-reported African American in PAGE, we estimated the h2γ and h2 for 20 quant. traits and found no significant test statistics inflation due to confounding biases, contrary to that genomic inflation factors and LDSC intercept depart from one for most traits

We also showed that HAMSTA is better at detecting biases in test statistics due to finer-scale structure than genomic inflation factor and LDSC/cov-LDSC (treating local ancestry as genotype). HAMSTA can also determine a significance threshold to achieve a better family-wise error rate.

In simulations, we showed that HAMSTA provides an approx. unbiased estimate for h2γ and is robust to finer-scale structure within a continental ancestry, unlike existing methods relying on individual-level data.

We create this tool HAMSTA github.com/tszfungc/HAM... for estimating heritability explained by local ancestry (h2γ) and evaluate stratification bias from AM summary statistics.

Admixture mapping studies have been conducted in admixed populations to identify trait-associated loci.  But currently, no methods have been developed for AM to diagnose the test statistics, or distinguish local ancestry effects from confounding biases