Guides › IPD-IMGT/HLA releases

Deleted HLA alleles: why names are withdrawn and how to trace them

Updated 2026-10-30 · allele examples checked against IPD-IMGT/HLA 3.66.0 on 2026-10-30

An HLA allele name is deleted when IPD-IMGT/HLA withdraws it from the current allele list. The usual reasons are that the sequence was shown to be wrong or identical to an existing allele, that the name was assigned in error, or that the allele was renamed, for example because its expression suffix changed.

A deleted name is not the same as a fabricated one. It was real once, it appears in published typings, and in most cases it has a successor. The record of every deletion, with its reason, is Deleted_alleles.txt in the IPD-IMGT/HLA repository.

Where deletions are recorded

The repository README describes Deleted_alleles.txt as "a csv file detailing all deleted allele names, with reasons for the deletion", and adds that the list "also includes details of any suffix changes". Each row has an allele ID, the deleted name and a free-text description.

Two other files complete the picture:

• Allelelist.txt lists the names current in the release.
• Allelelist_history.txt lists, for every HLA ID, its official name in every release.

The HLA ID is what makes successors traceable. A name can change or disappear; the ID of the sequence behind it does not.

The reasons, with real examples

The descriptions in Deleted_alleles.txt are free text, but they fall into a few recognisable kinds. All examples below are quoted from the file in release 3.66.0.

A few of these need a word of explanation.

Never assigned. A*02:100 and B*15:100 were reserved and not used. The April 2010 update folded the old A*92 and B*95 rollover families back into A*02 and B*15 starting at A*02:101 and B*15:101, and said the :100 names were not to be assigned. They have no successor.

Partial of a new allele. B*07:02:06 was deleted in July 2021 when the sequence was shown to be part of a newly named allele, B*07:436. In the history file both names carry the same HLA ID, HLA02990: B*07:02:06 from release 3.00.0 to 3.45.0, then B*07:436. The protein-level name changed, not just a field.

Suffix changes. The suffix page says suffix assignments follow expression evidence and may be revised. When they are, the old name goes into the deletion file. A*24:02:03 gained a Q in April 2011 and is now A*24:02:03Q; A*02:01:14 gained a Q in March 2011 and lost it again in August 2026, so A*02:01:14Q is the deleted form in release 3.66.0.

Deleted, renamed and unknown are three different things

When a stored name is not in the current allele list, it is one of three things, and they need different handling:

The categories overlap in practice. A suffix change appears in the deletion file and also as a rename in the history. A deletion for "identical to" an existing allele names its successor only in the description text, not in a structured column.

How to find a successor

1. Look the name up in Deleted_alleles.txt. Read the description. Many name the allele the sequence was found to be, or the name it became.
2. Look the name up in Allelelist_history.txt. If it appears in any release column, note its HLA ID.
3. Read that ID's name in the current release column. That is the current name of the same sequence, if it still has one. C*04:09N and C*04:09L share HLA01451.
4. If the description names an allele that the history does not link, treat the description as the successor, and record that the link came from the text.
5. If there is no successor, as for A*02:100 or a sequence withdrawn as an error, keep the original string and mark the typing as unresolvable rather than substituting a neighbour.

Do not shorten a deleted name to find a current one. B*07:02:06 shortened to B*07:02 lands on a different two-field name from B*07:436, its actual successor.

Doing it through the API

GET /v1/allele/{name} returns status: "deleted" with a successor for deleted names, and chases legacy spellings of deleted names the same way. GET /v1/allele/{name}/history returns the HLA ID, every name it has had release by release, and the recorded reason a name was retired. For a list of stored names, POST /v1/translate maps each one from the release it was written in to the current one and labels each row, for example unchanged, renamed, deleted (with reason and successor) or unknown.

Every answer names the release it came from, currently 3.66.0. The release history tool runs the same lookups in the browser, and the guide on IPD-IMGT/HLA releases covers the other kinds of name change.

Frequently asked questions

Why are HLA alleles deleted?

Deleted_alleles.txt records reasons such as sequence errors, identity with an existing allele, being part of a newly named allele, a name assigned in error, and changes to an expression suffix.

Is a deleted HLA allele name the same as an invalid one?

No. A deleted name existed in an earlier release and often has a successor. A name that appears in no release at all is a data error or a fabrication.

How do I find the current name for a deleted HLA allele?

Look up its HLA ID in Allelelist_history.txt and read that ID's name in the current release, and read the reason in Deleted_alleles.txt, which often names the successor.

Does every deleted allele have a successor?

No. Names that were never assigned, such as A*02:100, and sequences withdrawn as errors without an identical allele have none.