A proposal

An EMS variant field recorded against the animal, and a mating rule written on it

The material on this page is drawn from the Global Cat Health Survey 2026, which is where its sources and its declarations of interest are set out in full.

This is a proposal put to registries by the compiler of this directory. It is not a description of anything any registry does today. Every figure in it is recomputed from the register on each build, and each one says what it counts.

What is proposed

A variant field in EMS, recorded against the individual animal and separate from the eye-colour field. The eye-colour codes stay as they are and record what an observer sees. The variant field records which PAX3 variant a cat carries, identified or not, and how that was established. Matings are then gated on the field rather than on eye colour.

1. The variant codes

The codes already live in the GCCF's GEMS are adopted unchanged. Two numbering systems for these variants are already in circulation and they disagree, so no third numbering is proposed. The field carries a DBE prefix, because a bare number is not a variant name.

161 is not withdrawn, and that is deliberate. It is a live code and cats are registered under it. The suffix refines it: a registry that ignores the suffix still reads 161, so nothing already registered becomes wrong, and a registry that reads it gains the line. A code list that asks a registry to withdraw a code it has just introduced asks for the one thing a registry will not do.

2. The x suffixes, for a variant that is not identified

An unidentified variant is still a variant. Each suffix stands for one, named after the founder cat the line is named for, on the rule this directory applies throughout: resolve every variant name to a named founder cat before relying on it. "The Altai gene" names two different variants in the source material, and a code that is not tied to a founder repeats that fault.

The suffixes are frozen and are never reassigned. They were given in order of the cats showing the trait when the series was written. A line that grows keeps its number and a new line takes the next free one, because renumbering would invalidate every registration already written in the scheme, which is the fault this whole proposal exists to avoid.

22 lines in this register hold a blue or odd eyed cat in which no variant has been identified. The 6 above are those that are dated, carry registry codes and cannot be resolved to an identified variant. The register supplies the rest, and each would take the next free suffix.

An unresolved record does not receive a suffix, and the difference is the whole basis of the series. It stays at bare 161. Three cases in this register: Siberian, Altai or Barnaul unconfirmed, 95 records of which 36 showing; Altai and Barnaul, both suspected, 3 records of which 0 showing; Highlander, Olynx (C11), 10 records of which 5 showing. None of them is an unidentified variant. Each is a record nobody has been able to place.

3. The candidate form, for a variant identified but not established in this cat

A cat can carry a variant that has a test, without anybody knowing which one. That state is the opposite of an x suffix. An x suffix marks a variant no test can name. This marks a variant a test names today, on a cat nobody has swabbed.

116 cats in this register are in that state: carriers with no result of their own and at least one identified variant standing in their recorded ancestry. 65 have two or more different identified variants above them, so each of those may be a compound heterozygote. The pairings are 65 with DBE^ALT and DBE^CEL, 5 with DBE^ALT and DBE^RE, 5 with DBE^CEL and DBE^RE. These are the cheapest tests in the register: one swab moves a cat from a candidate form to a variant code, or to clear.

The Topaz is the case the scheme was written for. It was founded deliberately on two identified variants at once, DBE^CEL (Roxi) and DBE^ALT, and no Topaz cat has been genotyped. Its 18 foundation records carry no variant, because attribution belongs to the individual animal and is established by genotype, and 22 cats descend from them in this register.

These are not the same animals as the recommendation to test the cats whose status cannot be seen, and the two do not compete. That recommendation is about cats that show nothing, where the question is whether the cat carries at all. This is about cats that already show the trait, where the question is which variant. 114 of the 116 here show a blue or odd eye, so the two sets barely meet. A registry has to answer both questions and they are answered by different swabs.

A breed of the colourpoint series receives no code. A breed of that series is not a group of unknown variant unless a known variant has been bred into it.

4. The evidence modifier

A cat reads BSH n 11 61 DBE 163t. A latent carrier reads BSH ns 09 DBE 161 x1 p, with no eye code at all, which is the case no existing scheme can write down.

5. The status, and the mating rule

Carrier to carrier is prohibited. Carrier to unknown is not permitted. Carrier to clear is permitted.

6. The scheme applied to this register

A code list nobody can apply is an opinion. This is the proposal written against every record in the directory, computed from the fields already held. The field appears on every cat page and in the public export, so a reader can see what the scheme would say about the animal in front of them.

A cat can stand below two determinants, and the field says so

Where more than one determinant stands in a cat's recorded ancestry the field names all of them, separated by a vertical bar, and does not choose. 114 records in this register read that way, and 27 of them carry an identified variant and an unsequenced line at the same time.

That last group is why the rule exists. Coding such a cat for the identified variant alone points a breeder at a test, and a clear result on that test then reads as not a carrier while a second determinant, for which no test exists, is still standing above the animal. The field has to be able to say that no available test resolves it. A registry that ignores the second token still reads the first, so nothing already registered becomes wrong.

The case that produced this rule. A breeder holding both lines made the cross Seymour (DBE-ALT) x Igor (DBE, variant not identified). The daughter of that mating stands below an unsequenced line on her sire side and an attributed variant on her dam side, and reads DBE 161 x1|163p. Both daughters of the cross were born deaf, on the breeder's own account, on a pedigree that also carries white spotting in three places, so nothing here is attributed to the variants. Before this rule the same cat read DBE 163p, which named one determinant of the two and pointed at a test that could not settle her.

The scheme writes a field on 1130 of the 4225 records in this register, 26.7 per cent, and leaves the rest unknown. Every one of those fields is computed from what the register already holds, at build time, and none of it is stored: a cat moves from p to t the day a result is entered for it.

This is the test the suffix scheme has to pass and a withdrawal of 161 would fail. Withdraw 161 and there is no field at all for the largest group in the register, the carriers standing on no named line, so the column would be empty for most of the directory. The suffix keeps them coded and adds the line where a line is known.

Why this and not something else

Containment has failed

Six variants have been identified, and the trait is recorded in 35 of the 52 breeds in this register. 22 named lines carry a blue eye in which no variant has been identified, and 21 breeds are reached only by such a line. There is no boundary left to draw around the trait.

The harm is a mating, not a cat

A heterozygote is a healthy animal. The two kittens born deaf carried one copy each of DBE^CEL (Roxi) and DBE^ALT. The kitten that died at birth with limb contractures and an abnormal head is presumed to have carried one copy each of DBE^CEL and DBE^RE. The DBE^ALT homozygotes described in the literature were white and, in the founding Altai programme, "most often deaf". Every documented harm is a mating.

To those is now added Seymour (DBE-ALT) x Igor (DBE, variant not identified), reported on 30 August 2026 by a breeder who held both lines and made the cross deliberately. Both daughters of it were born deaf. It is the second crossing of two determinants in this register whose outcome anybody has reported, and both of the two produced deaf kittens.

A single copy of one variant does not do this. Nine genotyped heterozygous carriers of DBE^ALT were BAER tested at Glasgow and all nine heard normally in both ears. The kittens in both reported crossings were not simple carriers of one variant; they were the product of two determinants, which is the state the literature associates with the harm.

That carries an inference about the unsequenced line, and it is worth more than the deafness. A compound heterozygote is two different alleles at one locus, and the published case is the PAX3 compound: two white deaf kittens carrying DBE^CEL and DBE^ALT, one on each chromosome. A cross to a second line producing the same outcome is evidence that the second determinant is also at PAX3, since a variant elsewhere would give a double heterozygote at two loci rather than a compound at one. All six identified determinants are PAX3 variants, so the prior was already high. The test that settles it is a PAX3 screen on one cat of the line, not a whole genome and not a variant panel, and a panel is the wrong instrument because it reports only the variants already named and returns clear on a seventh. The white these animals carry stays on the record as an unexcluded alternative, and the PAX3 read is what distinguishes them.

The rule does not depend on which way that resolves. Carrier to carrier was proposed because a compound heterozygote cannot be predicted, not because any particular pairing has been shown to be harmful, and a rule written on identified variants would not have reached this mating at all: no test for the Igor line exists to have been required.

A breed cannot hold it, and a colour class cannot either

A breed cannot hold it. A breed constituted by a variant that is lethal in the homozygote cannot be bred to itself, so its stud book cannot close and it exports the variant in every generation by the terms of its own standard. That is the Céleste, set out in case study one of the survey.

A colour class cannot hold it. Registry classes are written on coat. A latent carrier has no blue eye at all: it carries the variant, cannot be found by looking, and is not caught by a rule written on eye colour. A class written on eye colour also catches the cats whose blue eye a source attributes to something else: to dominant white, to white spotting, or to the colourpoint allele.

A breed and a colour class are both classifications of cats. This is not a property that classifies a cat. A carrier is an ordinary animal of its breed and its coat, and the variant bears on nothing until two cats are put together.

The same problem, already solved: blood group

A type B queen bred to a type A sire loses the litter to neonatal isoerythrolysis. She is a healthy cat, and the harm exists only in the pairing. No registry created a type B breed or a type B colour class. The blood group is recorded against the individual animal, it travels with her when she is sold, and one mating is avoided.

Why one unknown code is not enough, and why a suffix answers it

Applied to this register the scheme identifies 1039 carriers. The variant is identified in 90 of them and not in the other 949. A rule written on identified variants alone reaches fewer than one carrier in ten. Bare 161 holds the other nine without separating them, and the suffix separates them without taking 161 away.

• A line can be reassigned as a block. If the Igor variant is sequenced, every cat coded DBE 161 x1 moves to a variant code in one operation.
• A cohort can be defined for the laboratory. The recommendation to trace the descendants of the lines in which no variant has been identified needs a defined population, and a suffix defines it.
• Two kinds of unknown separate. A variant nobody has sequenced, which no purchase resolves, and a variant identified but not yet established on this cat, which a test already detects. Bare 161 conflates them and the suffix does not.

Why carrier to carrier, and not a test requirement

Carrier to carrier prohibits three matings and a test reaches only the first: two carriers of one identified variant, the homozygote; two carriers of different identified variants, the compound heterozygote, which is where the deaf and the dead kittens are recorded; and two carriers of unidentified variants, or one identified and one unidentified. That last is the Igor to Marusya cross, which no laboratory can predict, and the Kaena to Lazuli mating that founded the Céleste programme with a variant on each side. The rule does not require the variant to be identified, which is why it reaches the 22 lines for which no test exists.

Until 30 August 2026 this page said that nobody had reported such a cross. That is no longer true and the sentence is withdrawn. A breeder holding both lines made the cross Seymour (DBE-ALT) x Igor (DBE, variant not identified), and reports that both daughters of the mating were born deaf. Neither line has ever been sequenced, so no test on either parent existed to be required, and no laboratory could have predicted the pairing. The cause of the deafness is not established and this register does not claim it: the animals carry white spotting, which is the ordinary route to congenital deafness in the cat, and a two-line cross and an ordinary explanation stand side by side here as they do everywhere else. What is not in doubt is that the mating this rule prohibits has now been made and reported. The pedigrees and the breeder's account are on the records of Solomiya Josephine, her unnamed sister and King_Pride Linda in the directory.

Why a blue eye is not by itself a carrier record

A blue eye already explained is not evidence of a variant. Where a cat's own code carries a pointing factor, 31 Burmese, 32 Tonkinese or mink, or 33 colourpoint, its blue eye has an ordinary cause. Where a registry codes both, two causes of a pale eye stand on one animal and the record cannot separate them. 5 cats in this register are in that position, and the scheme leaves them unknown rather than coding them either way. Only a genotype separates the two.

Why a negative test does not clear a cat

A negative panel narrows the risk. A cat with a line in which no variant has been identified standing in its recorded ancestry cannot be moved out of unknown by any result now available, because the panel does not cover that line. 2 of the 12 wild-type cats in this register are in that position, so a scheme treating a negative panel as clearance is already wrong about 2 animals. The test for it is the upward trace this directory already runs.

What the scheme requires, and what it does not do

A latent carrier keeps its breed, its colour class and its show career, and gains one field. No new category is created for these cats to sit in.

Status can be derived from the pedigree where no test exists. This directory computes obligate carriers, obligate pairs and cats standing on the path above a pair from recorded parent links alone. A registry holding pedigrees can compute the same, and it is the only mechanism available for the breeds reached only by lines in which no variant has been identified.

A carrier may be bred only to a cat that has been tested, so the cost of not testing falls on the breeder who wants to use a carrier.

Unknown to unknown is uncontrolled, and on this register that is 3160 animals of the 4225 it holds. The proportion falls as the tested population grows. It cannot be closed for the 22 lines in which no variant has been identified.

No registry currently records a status of this kind against an individual animal, carries it on the pedigree and the sale documentation, and publishes it in a form another registry can read. The variants have crossed breeds and borders, so a status held inside one registry does not see the mating that matters.