Why Sex-Chromosome QC Requires Ploidy Awareness
A genotype call is meaningful only in relation to the number of chromosome copies expected at that site. Autosomal SNPs are usually represented as diploid genotypes, but the usual X- and Y-chromosome patterns depend on chromosome region and on the sample’s sex-chromosome complement. If a pipeline assumes the wrong ploidy, it can produce misleading calls, missingness, or apparent heterozygotes.
Sex-chromosome QC uses these expected patterns to check whether genotype data are broadly consistent with recorded sample metadata and with the analysis pipeline’s assumptions. A mismatch is a flag for investigation, not a diagnosis of a sample’s biology or identity. Recorded sex metadata, sex assigned at birth, gender identity, and inferred sex-chromosome patterns are distinct concepts; genotype data alone do not establish a person’s gender.
Expected Patterns Under Common Conventions
For many datasets, the starting convention is that samples recorded as XX have two X copies and no Y copy, while samples recorded as XY have one X copy and one Y copy. These are useful working expectations, not a complete account of human sex-chromosome variation. The expected call pattern also depends on whether a variant lies in a PAR or in a non-pseudoautosomal region, often abbreviated non-PAR.
| Region | Common XX convention | Common XY convention |
|---|---|---|
| X non-PAR | Diploid X calls; heterozygous calls can occur | Haploid X calls; one allele copy per site |
| X/Y PAR | Usually treated as diploid | Usually treated as diploid, with corresponding X and Y regions |
| Y non-PAR | Usually no Y call expected | Haploid Y calls may be present |
A diploid biallelic genotype is often written as 0/0, 0/1, or 1/1, where 0 and 1 identify the two alleles in the variant record. A haploid call has only one allele, such as 0 or 1. Some software or file formats represent a haploid call using a diploid-looking homozygous form, such as 0/0 or 1/1. The format’s ploidy rules must therefore be checked before interpreting the displayed genotype. A haploid allele should not be counted twice merely because a file stores it in a two-field form.
The PAR boundaries depend on the reference genome build and the dataset’s annotation. A site mistakenly classified as non-PAR may appear to violate the expected ploidy pattern even when the underlying call is reasonable. Before interpreting QC summaries, verify the genome build, chromosome labels, PAR coordinates, and the software’s handling of haploid calls.
Worked Example: Interpreting Non-PAR X Calls
Suppose a hypothetical sample recorded as XY has calls at 100 well-called X non-PAR SNPs. The file uses one allele per haploid call. At 98 sites it records either 0 or 1, and at 2 sites it records a diploid-looking heterozygous call, 0/1. What should a QC analyst conclude?
Under the common XY convention, X non-PAR sites are expected to have one allele copy. The observed haploid-call fraction is
Thus 98% of these calls have the expected one-allele representation, while 2% are represented as heterozygous diploid calls. Those two calls merit review: possible explanations include a ploidy or file-conversion issue, an incorrectly annotated region, or a calling problem. The 2% figure alone does not identify which explanation is correct, and it is not sufficient to reclassify the sample.
The site coordinates should first be checked against the PAR annotation. If they are confirmed to be non-PAR, inspect the caller and downstream conversion settings to determine whether haploid genotypes were represented consistently. The calculation describes these 100 selected, well-called sites; it does not by itself establish a genome-wide error rate.
Simple Sample-Level Summaries
A practical review starts with a small set of summaries calculated separately by chromosome region. For each sample, count called and missing genotypes on X non-PAR, X PAR, and Y non-PAR. On X non-PAR, summarize the fraction of calls with a haploid-compatible representation, using the software’s documented convention. On Y non-PAR, summarize the number or fraction of called sites. Where the data format supports it, also summarize heterozygous calls in regions expected to be haploid under the working convention.
For example, a call fraction can be defined as the number of called sites divided by the number of eligible sites in the region:
“Eligible” should mean sites that passed the chosen site-level filters and were included in the summary. A low fraction can reflect poor sample-level data, but it can also arise from assay design, regional coverage, filtering, or a mismatched ploidy model. Compare samples processed with the same platform and pipeline, and interpret summaries in light of how many informative sites were available. Do not treat a particular cutoff as universal.
Worked Example: Comparing Y Call Fractions
In a hypothetical dataset, two samples recorded as XY are evaluated at the same 200 eligible Y non-PAR SNPs. Sample A has 182 called sites; Sample B has 41 called sites. Calculate and compare their call fractions.
For Sample A,
For Sample B,
Sample A has a 91.0% Y call fraction and Sample B has a 20.5% fraction among these eligible sites. Under the working XY convention, Sample B’s much lower fraction is a QC flag. Possible causes include a sample or metadata mismatch, poor data quality, or differences in assay and filtering behavior. Because the samples share the same set of eligible sites, their summaries are directly comparable in this example. Even so, the comparison does not prove that either sample has a particular chromosome complement; the sample identity, pipeline settings, and underlying calls need review.
Using Discordance as a QC Signal
A sample’s X and Y patterns can be compared with its recorded metadata, but no single summary is decisive. For example, an XY-recorded sample with many heterozygous calls on X non-PAR may indicate a ploidy-setting or calling issue, or it may reflect an unusual biological or technical circumstance. An XX-recorded sample with substantial Y non-PAR signal may prompt checks for sample swaps, contamination, or metadata problems, but the observation alone does not distinguish among them. Conversely, weak Y calls do not automatically contradict XY metadata if the assay has limited Y coverage or the calls are heavily filtered.
The check is strongest when independent indicators agree: non-PAR X representation, Y non-PAR call fraction, sex-chromosome missingness, and the consistency of calls across batches or technical replicates. The scope here is to recognize and investigate sex-chromosome discordance; other QC analyses are needed to characterize broader heterozygosity patterns, contamination, or batch effects.
Worked Example: Investigating a Metadata Mismatch Flag
Consider a hypothetical sample recorded as XX. Among 120 eligible X non-PAR sites, 116 have diploid calls and 4 are missing. Among 80 eligible Y non-PAR sites, 0 are called. Calculate the X non-PAR call fraction and describe what the pattern does and does not show.
The X non-PAR call fraction is
So 96.7% of eligible X non-PAR sites have calls, and the observed Y non-PAR call fraction is \(0/80=0\). This pattern is consistent with the common XX convention in this dataset: diploid X calls are present and no Y calls were observed. It is not proof of a particular sex-chromosome complement, because the result depends on site eligibility, assay sensitivity, sample identity, and the call-generation pipeline. In particular, “no Y calls observed” means no calls among these 80 eligible sites, not that every possible Y sequence was tested and absent.
If this sample had instead shown many Y calls, the appropriate next steps would include checking sample identifiers and metadata, confirming the chromosome and PAR annotations, reviewing the Y-call filters and assay design, and comparing the pattern with other samples processed in the same batch. Any conclusion about a discrepancy should be made only after those checks.
A Practical Review Sequence
Check the genome build, chromosome naming, PAR boundaries, and whether each site was assigned to PAR or non-PAR correctly.
Read the file-format and software rules for haploid calls. Determine whether haploid genotypes appear as one allele or as a homozygous diploid-looking pair.
Compare X non-PAR representation, X and Y missingness, and Y non-PAR call fractions using eligible sites and consistent filters.
Check sample identifiers, metadata, assay coverage, and pipeline settings. Report a discordance flag and the evidence behind it, not an unsupported biological conclusion.
Why These Checks Matter—and a Common Pitfall
Sex-chromosome ploidy errors can affect genotype encoding, missingness summaries, and downstream analyses. A pipeline that treats haploid sites as diploid may generate implausible heterozygous calls or count an allele twice. A pipeline that assigns PAR sites the wrong ploidy can create apparent inconsistencies. Separating these issues from sample-level metadata checks helps prevent technical artifacts from being interpreted as properties of a person.
Biological variation also matters. The XX and XY patterns are common conventions, not an exhaustive model of sex-chromosome complements, mosaicism, or other biological circumstances. For that reason, a QC workflow should distinguish a data-quality or metadata flag from a clinical interpretation and should restrict access to sensitive sample-level results appropriately.
Check Your Understanding
Answer each question using the region, ploidy convention, and evidence specified.
- Under the common XY convention, what genotype representation is expected at an X non-PAR biallelic SNP, and why can a diploid-looking 0/1 call deserve review?
- Why should X/Y PAR sites not automatically be evaluated using the same ploidy expectation as non-PAR sites?
- A sample has 45 Y calls among 150 eligible Y non-PAR sites. Calculate its call fraction and give one reason that this result alone cannot establish a sample’s chromosome complement.
- A file stores a haploid allele as 1/1. What must be checked before counting allele copies or calling this a diploid genotype?
- Give two technical or metadata explanations for an apparent disagreement between recorded sex metadata and sex-chromosome calls, and state one cautious way to report the finding.