ROH Endogamy Detector

Runs of Homozygosity are stretches of your genome where you inherited an identical DNA sequence from both parents. They appear whenever your two parents share a common ancestor, whether that ancestor lived 5 generations ago or 1,500 years ago.

The single most important thing to understand:

Two parents from different continents (a Frenchman and a Chilean, for example) usually share almost no recent ancestors, so their child's ROH is very low, even though that child's ethnicity estimate looks dramatically mixed. Conversely, two parents from a small endogamous community (Ashkenazi, Sardinian, Tejano) share many deep founders, so their child's ROH is high, even though their ethnicity may look perfectly monotone.

So a "very mixed" ethnicity does not predict your ROH at all. Look at your two metrics separately.

FROH, fraction of your genome in ROH

This is the headline number. It is the proportion of your autosomal genome covered by homozygous segments. Reference ranges:

• 0 – 0.5%, outbred (most non-isolated humans)
• 0.5 – 1.5%, mild background structure (Eastern / Southern European, mixed colonial American)
• 1.5 – 3%, moderate endogamy (Sardinian, Finnish, Basque, South Italian)
• 3 – 6%, strong population endogamy (Ashkenazi, Druze, Romani, Mennonite)
• 6 – 12%, very recent consanguinity (1st or 2nd cousin parents) or extreme founder communities (Hutterite)
• over 12%, closer than 1st cousin or extreme isolation

SROH, total Mb of ROH

The same information as FROH but in absolute units. SROH = FROH × 2,881 Mb (the length of your 22 autosomes). Useful when comparing your numbers to published research, which often reports SROH directly.

nROH, number of distinct segments

How many separate ROH stretches your genome contains. The shape matters more than the count alone:

• Many short segments (20+ small ones, all under 5 Mb) → ancient drift or population bottleneck many generations back
• Several medium segments (10 to 15, mostly 5 to 15 Mb) → community-level endogamy from a closed founder population
• Few but long segments (3 to 8, several over 15 Mb) → recent shared ancestor between your parents

Longest ROH

The single longest segment. Length correlates with how recently the shared ancestor lived (recombination breaks segments down generation by generation). But interpretation depends on what surrounds it:

• A 25 Mb segment plus 4-5 other long segments → genuine 1st-2nd cousin parents
• A 25 Mb segment standing alone on a background of short segments → a specific deep ancestor (often a brick-wall) or a small-Ne population component (Native American), not 1st-3rd cousin parents

The tool now distinguishes these two cases automatically and adjusts the verdict accordingly.

Green, Outbred or mild background structure. No endogamy or consanguinity signal. The default human profile. Your parents come from genetically distant origins, or from large outbred populations.

Amber, Moderate endogamy, single long outlier, or background structure. Either a regional founder community somewhere in your ancestry (Sardinian, Finnish, South Italian), or one specific shared distant ancestor on top of an otherwise normal background, or many short segments from ancient drift. Cross-reference with admixture analysis.

Red, Strong endogamy or recent consanguinity. Either a strongly bottlenecked closed population (Ashkenazi, Romani, Druze), or genuinely close-relative parents (1st-3rd cousin range). The verdict text tells you which of the two patterns the tool detected.

"My ethnicity is super mixed but my FROH is tiny. Is something wrong?"

No, that is the expected pattern when your parents come from different continents or distant populations. ROH measures what your parents share, not how diverse you are. See Case A.

"My ethnicity looks plain but my FROH is high. Is something wrong?"

Not necessarily. Plain-looking colonial American, French Canadian (Québécois), Mexican Mestizo, or southern Italian profiles often have elevated FROH because the ancestors of those populations lived in semi-isolated communities for centuries with high cousin-marriage rates. See Case B.

"Multiple populations are showing as 4/4 Probable. Which one is mine?"

Probably none of them in the strict sense. When 4 or more historically distant populations all match perfectly (Romani plus Druze plus Hutterite plus Mennonite plus Bedouin, for instance), no real human descends from all of them. Your metrics simply happen to fall in a range where many bottlenecked populations overlap, and the tool flags this as admixture mimicry with an amber warning. The right reading is to focus on the verdict at the top, not the cards underneath.

"The verdict says '1st–3rd cousin parents' but I know my parents are not related."

The tool distinguishes a single long outlier (one long segment, otherwise normal background) from true consanguinity (multiple long segments). True 1st-2nd cousin offspring almost always produce 3 or more segments above 15 Mb, not one. If you only have one long segment, the tool will say so explicitly, that pattern more often reflects a specific shared deep ancestor, often hidden behind a genealogical brick wall four to eight generations back.

"23andMe and AncestryDNA give me different numbers. Which is right?"

Both are right at chip level. Differences come from the SNP coverage of each chip. The tool detects high-density and imputed files automatically and rescales the detection thresholds, so cross-platform results should be very close. Significant divergence (more than 30% difference in FROH) usually points to a file quality issue with one of the two uploads.

"Is 'Outbred' good or bad?"

Neutral. It just means your parents come from genetically distant origins. Most non-isolated humans are outbred. It does not mean your ancestry is "boring", it can mean your parents come from opposite sides of the world.

"What does the parental kinship / generations estimate mean?"

It is a rough estimate based on the longest ROH segment alone. It is reliable only when multiple long segments confirm recent consanguinity. When the longest segment is a single outlier on a background of mostly short segments, the estimate is unreliable, and the tool says so explicitly in the verdict text.

"How can I dig deeper into my own case?"

Three complementary checks: (1) run an admixture analysis (our Advanced Ancestry or Medieval & Modern Report) to see which populations actually contributed to your genome; (2) if both parents have tested, run a Gedmatch one-to-one between them, that is the cleanest single answer to "are my parents related"; (3) work back four to eight generations on both lines in your family tree, focusing on regions with known endogamy or colonial founder dynamics.

Consumer chips have a lower resolution limit. ROH segments below 1.5 Mb are unreliable on consumer chips because of variable SNP density. The tool ignores anything shorter.

No phasing. The tool uses unphased genotype data, which is what consumer chips provide. Lab-grade tools like hapROH use phased data and can detect deeper IBD signals, but require WGS data and specialized pipelines.

Population reference values are population-level averages. Individuals within any population vary widely. A "Probable" match on a population card means your metrics fall within the typical range, not that you definitely descend from that population.

Pericentromeric regions are excluded. Every consumer chip has SNP coverage gaps around centromeres, which produce false-positive ROH on naive callers. The tool excludes any candidate segment overlapping more than half of a pericentromeric ±3 Mb buffer. Your ROH counts here will therefore be slightly lower than from a tool that does not apply this filter, that is intentional and correct.

This tool gives you a SIGNAL, not a DIAGNOSIS. Real interpretation requires combining your ROH with: (1) admixture analysis from a separate tool, (2) genealogy if available, and (3) optionally a Gedmatch one-to-one between your parents. The verdict is a starting point for further investigation, not a final answer.