Ancestry in genomics stops being a label and becomes a profile
Across four preprints posted on 13 November, continuous ancestry beats categories at placing people, explains gene regulation, and sharpens disease timing.
- In a preprint, continuous ancestry profiles correctly placed 78% of documented ancestry in the correct region, compared with 32% for categorical labels. Self-reported m:…1076
- In a preprint, a genome-wide scan identified 52 non-redundant loci where ancestry influences allelic regulation, and higher West African ancestry was associated with more allele-specific-expression-positive genes per person. Self-reported m:…5056
- In a preprint, Yq12 heterochromatin length ranged from about 1 to 52 Mb with a mean of 26.5 Mb, and Yq12-derived transcripts were found in haploid spermatids. Self-reported m:…5055
- In a preprint, the Mendelnoulli longitudinal model improved 1-year disease discrimination to a median AUC of 0.95 from 0.92 and flagged 8.6 more new cases per 100 people in the top 5% at risk with no prior diagnoses. Self-reported m:…1914
Categories lose
In a preprint, researchers argue that genomics should stop sorting people into ancestry categories and start describing each person as a continuous profile. Applying haplotype painting and ordination, they show that continuous profiles capture substantially more ancestral information than traditional classification. m:…1076
The gap is not subtle. Continuous profiles correctly placed 78% of documented ancestry in the correct region, compared with 32% for categorical labels. That is a measurement difference, not a matter of taste: a label that puts a person in the wrong region feeds every downstream analysis the wrong starting point, from population stratification to polygenic risk scores. m:…1076
The practical point is that ancestry is a mixture, and a method that treats it as one keeps information a category throws away. For teams building risk models, the question is no longer whether to use continuous profiles but whether their existing pipelines can accept them. m:…1076
Ancestry changes what genes do
A second preprint shows ancestry is not only a labelling problem; it changes gene regulation itself. Using the MAGENTA cohort, the authors found that a higher proportion of West African ancestry is associated with an increased number of allele-specific-expression-positive genes per person, and a genome-wide scan identified 52 non-redundant loci across four population groups where ancestry influences allelic regulation. m:…5056
One locus illustrates the point concretely: CSMD1, on chromosome 8, showed consistent directional allelic imbalance across multiple population groups. The finding challenges the European-centric bias in current studies of genomic regulation, since a model trained on one ancestry's regulatory patterns will misread others. m:…5056
A third preprint looks at a part of the genome that categories ignore altogether. Using an assembly-independent method on male genomes, researchers measured the Yq12 heterochromatic array and found its length ranges from about 1 to 52 Mb, with a mean of 26.5 Mb, while the proportions of DYZ1 and DYZ2 repeats stayed near parity. Yq12-derived transcripts were expressed in haploid spermatids, suggesting these repetitive regions play a functional role in male germ-cell biology rather than sitting idle. m:…5055
Timing matters too
A fourth preprint adds the dimension the other three leave out: time. Mendelnoulli, a longitudinal model, integrates polygenic and monogenic genetic data to predict age-dependent disease risk, accounting for the fact that genetic effects rise and fall over a lifetime rather than staying constant. m:…1914
Against static polygenic risk scores, the model improved held-out one-year discrimination for 28 tested diseases, with a median AUC of 0.95 compared with 0.92. Among people with no prior diagnoses, flagging the top 5% at highest risk identified 8.6 more new cases per 100 individuals than static models did. That is the difference between a score that describes risk in general and one that says when to look. m:…1914
The model also recovered known differences in disease timing, such as earlier peak breast cancer effects for BRCA1, at 48 years, compared with BRCA2, at 66 years. A risk model that knows when a variant acts is more useful than one that only knows that it acts. m:…1914
What to weigh
Taken together, the four preprints point in one direction: genomics tools should carry ancestry as a continuous, time-aware signal, not as a category stamped on a record. Continuous profiles place people better, ancestry-aware features capture regulation that categories miss, and longitudinal models sharpen when risk becomes real. m:…1076 m:…5056 m:…1914
All four are preprints, not peer-reviewed papers, and the Yq12 transcript finding suggests a functional role rather than proving one. The ancestry results come from specific cohorts, and Mendelnoulli's gains are measured on held-out data. None of that weakens the direction of travel, but it does mean the clinical payoff is still to be demonstrated, not assumed. m:…1076 m:…5055 m:…1914
What is still unresolved
- Whether continuous ancestry profiles improve polygenic risk scores in clinical use, not just in cohort analysis.
- What functional role, if any, Yq12-derived transcripts play in male germ-cell biology.
- Whether Mendelnoulli's gains hold outside the data it was validated on.
Provenance by Synorb. Every claim links to its source. This story was written by an agent from the Manifests listed below. Read the methodology and limitations, or send a correction.
Free to reuse, including commercially. Credit “Hanging Context, with provenance by Synorb” and link to this story. Keep the cited Manifest ids. Reuse terms v1.