Reading Your Blood Type From Your DNA

Oak
genomics wgs blood-type abo variant-calling
A giant translucent tree in fog, sap glowing amber below a gap in a bead-like chain and blue above it, with grazing beasts beneath.

Your ABO type is determined almost entirely by three positions in exons 6 and 7 of the ABO gene on chromosome 9q34.2. Your RhD status is usually determined by whether you carry zero, one, or two copies of the entire RHD gene on chromosome 1p36.11. If you have a whole-genome BAM or CRAM, which are the standard file formats holding your aligned sequencing reads, you can read both in about ten minutes. If instead you have a consumer genotyping array, you can often read ABO and almost never read RhD reliably. Neither result is a substitute for serological typing before a transfusion, and no hospital will accept one.

The variants that determine ABO

Before looking at any data, it helps to know what you are looking for. The ABO gene encodes a glycosyltransferase that adds a sugar to the H antigen on the red cell surface. Which sugar it adds, or whether it works at all, is set by a handful of coding changes. Three of them carry most of the information.

  • O (the common allele): c.261delG, rs8176719, a single-base deletion in a run of guanines in exon 6. It shifts the reading frame and produces a truncated, nonfunctional enzyme. Two copies of a 261delG-bearing allele gives type O in the large majority of O individuals.
  • A vs B: four amino acid differences separate the A1 and B transferases (residues 176, 235, 266, 268), but only two change sugar specificity in practice: c.796C>A (Leu266Met, rs8176746) and c.803G>C (Gly268Ala, rs8176747). A-type sequence at both positions gives A, and B-type sequence at both gives B. Someone heterozygous at both positions with no 261delG on either haplotype is AB.
  • A2: c.1061delC (rs56392308), a deletion near the C-terminus that extends the protein and weakens transferase activity. Serologically A2 is still type A, but it matters for subgroup interpretation and for the rare A2 individuals who make anti-A1.

There are also rare O alleles that lack 261delG entirely (O03 and relatives), and these are the main way a genotype-only call goes wrong in the A/O direction. Most of them are rare and population-specific, and deep whole-genome sequencing of population cohorts is how they were cataloged in the first place 1. If your ancestry is not well represented in the ISBT allele tables, treat a predicted A or B with more caution than a predicted O.

Calling ABO from a CRAM

The practical route is to pull the region and call it directly rather than trusting a joint-called VCF, a variant call file produced across many samples at once, which may have filtered out the indel you care about. The following three commands do the pileup, the calling, and the normalization:

bcftools mpileup -f GRCh38.fa -r chr9:133,255,000-133,276,000 \
  -a AD,DP -d 1000 -q 20 -Q 15 sample.cram \
| bcftools call -m -Ou \
| bcftools norm -f GRCh38.fa -m -any -Ov -o abo.vcf

Verify the coordinates against your own reference build before you trust them. On GRCh37 the gene sits near chr9:136,125,000-136,151,000, and rs8176719 moves by roughly 2.87 Mb between builds. The bcftools norm step matters because a deletion inside a homopolymer has several equivalent left- or right-aligned representations, and a naive string comparison against dbSNP will miss it.

With the calls in hand, look at the four positions by hand in IGV, the Integrative Genomics Viewer. There are three things worth checking, in this order:

  1. Depth at rs8176719. Single-base deletions in homopolymer runs are the worst case for Illumina chemistry. If allelic depth is lopsided (say 42 reference, 8 deletion), that is more likely a PCR slippage artifact than a real heterozygote. A clean het should sit near 50/50 with 20x or better total depth, meaning each position is covered by at least twenty independent reads.
  2. Coverage uniformity across exons 6 and 7. Library preparation changes local coverage more than most people expect, and GC content plus fragmentation method drive dropout in specific clinically relevant genes 2. If your exon 7 coverage is 8x while your genome median is 30x, the A2 and A/B calls are soft.
  3. Phase, meaning which variants sit together on the same inherited chromosome. c.796 and c.803 are 7 bp apart, so any 150 bp read spans both and read-backed phasing is trivial. c.261 and c.803 are about 540 bp apart, which no short read spans. Run whatshap phase --reference GRCh38.fa --indels and accept that the 261-to-803 link will often come out unphased. That ambiguity is what separates “B/O, phenotype B” from the rarer configurations, and it is the single best argument for long reads if you care about the answer.

RhD is a copy-number question

RhD behaves differently from ABO, and the difference shapes how you measure it. RHD and RHCE are adjacent paralogs in opposite orientation, about 97% identical in their coding sequence. They are flanked by two nearly identical segments called Rhesus boxes. The most common cause of RhD-negative status in European-ancestry people is an unequal crossover between those boxes that deletes the entire RHD gene. The primary measurement is therefore depth rather than a variant call, and a tool such as mosdepth will produce it:

mosdepth --by RHD_RHCE.bed --fasta GRCh38.fa -Q 0 -t 4 rh sample.cram

Compute the ratio of mean RHD depth to mean RHCE depth, and separately to your genome-wide median. Expect roughly 1.0 for two RHD copies, 0.5 for one, and near zero for a homozygous deletion.

Two of the flags deserve comment. Do not filter on MAPQ, the mapping quality score, here. Reads from the shared sequence get MAPQ 0 because the aligner cannot choose between the paralogs, and dropping them flattens the very signal you want. Instead, restrict your depth windows to intervals containing paralogous sequence variants. These are the positions where RHD and RHCE differ, so that reads can be assigned to one gene or the other.

Non-deletional RhD-negative phenotypes exist and are common outside Europe, which limits what a depth ratio can settle. The RHD pseudogene, a 37 bp duplication at the intron 3/exon 4 boundary plus a nonsense change, is frequent in African-ancestry populations and looks copy-number normal. Asian-type DEL (c.1227G>A, exon 9) produces a serologically D-negative red cell that is genomically D-positive, and people with it are often typed as negative by routine serology. A depth ratio of 1.0 therefore does not close the question.

What a genotyping array can and cannot tell you

Most people who have any genetic data at all have array data rather than a genome, so it is worth being precise about what that buys you. 23andMe and AncestryDNA chips include rs8176719 on some versions, plus tag SNPs such as rs505922 in intron 1 that correlate with the O haplotype. That gets you a decent ABO guess.

It does not get you RhD. Arrays measure hybridization intensity at fixed probes, and a whole-gene deletion between two near-identical paralogs is exactly the thing probe-based platforms handle badly. Cross-hybridization of probes to paralogous sequence is a documented, systematic failure mode of Illumina bead arrays rather than a rare accident 3. Any consumer report that hands you an Rh sign with confidence is inferring it from ancestry-correlated tag SNPs rather than measuring it.

H antigen, secretor status, and the rare systems

ABO and Rh are the systems everyone knows, but several neighboring genes change how those antigens are presented, and a few are worth a look while you are in the data. ABO antigens are built on the H antigen, made by FUT1 on chromosome 19. People homozygous for inactivating FUT1 variants have the Bombay phenotype: they type as O on routine testing but carry anti-H and can only receive blood from other Bombay individuals. It is rare globally and notably less rare in parts of India.

The neighboring gene FUT2 controls secretion of ABH antigens into saliva and mucus. The nonsense variant rs601338 (c.428G>A, Trp143Ter) is homozygous in roughly a fifth of European-ancestry people, making them non-secretors. It is a single line in your VCF and has downstream associations with gut microbiome composition and norovirus susceptibility.

The ISBT recognizes more than 40 blood group systems beyond ABO and Rh, and several of them are single-variant calls that are already present in data you have:

  • Vel-negative, from a 17 bp deletion in SMIM1
  • Duffy-null, from the DARC/ACKR1 GATA-box variant rs2814778
  • Kell, from KEL c.578C>T

If you are running these, use the ISBT allele tables as ground truth rather than a blog post.

One caution outranks everything above. Do not use a genomic prediction to make any transfusion decision. Blood banks type serologically and crossmatch every unit for a reason, and discrepancies between genotype and phenotype are the whole point of that workflow. If you are pregnant and Rh-negative, or facing surgery, that conversation belongs with a clinician.

Questions people also ask

Can two type B parents have a type O child? Yes, if both parents are B/O heterozygotes. Each passes the O allele with probability 1/2, so a quarter of their children are expected to be type O.

Can a child have a blood type different from both parents? Routinely. An A/O parent and a B/O parent can produce O, A, B, or AB children, three of which differ from both parents.

How did I end up with O-negative? You inherited a nonfunctional ABO allele from each parent, usually 261delG. Separately, you inherited a deleted or inactive RHD from each parent. The two genes are on different chromosomes and segregate independently.

What is the rarest blood type? Rhnull, in which no Rh antigens are expressed at all, arising from inactivating variants in RHAG or RHD/RHCE. Fewer than a hundred people have been documented worldwide. Bombay phenotype and Vel-negative are also very rare.

What is the oldest blood type? The A-type transferase sequence most closely resembles the form shared with other primates, and the O alleles look like independent loss-of-function events that arose on A backgrounds multiple times. A is therefore the better answer than O, and the ABO polymorphism itself predates our species.

What does Rh-negative mean? You lack the D antigen on your red cells, most often because both copies of RHD are deleted. It has no known effect on day-to-day health. It matters for transfusion compatibility and for pregnancy management, both of which are clinical questions.

Oak builds longitudinal molecular profiles of individuals: whole-genome sequencing, RNA sequencing, proteomics, blood biomarkers, and continuous glucose data, integrated into one model of you. Build your profile.

Footnotes

  1. Masao Nagasaki, Jun Yasuda, Fumiki Katsuoka, et al. Rare variant discovery by deep whole-genome sequencing of 1,070 Japanese individuals. Nature Communications, 2015. https://doi.org/10.1038/ncomms9018 ↩

  2. Vanessa Process, Madana M.R. Ambavaram, Sameer Vasantgadkar, et al. Optimization of DNA Fragmentation Techniques to Maximize Coverage Uniformity of Clinically Relevant Genes Using Whole Genome Sequencing. Diagnostics, 2025. https://doi.org/10.3390/diagnostics15182294 ↩

  3. Ruth Pidsley, Elena Zotenko, Timothy J. Peters, et al. Critical evaluation of the Illumina MethylationEPIC BeadChip microarray for whole-genome DNA methylation profiling. Genome biology, 2016. https://doi.org/10.1186/s13059-016-1066-1 ↩