Antibody numbering and CDR definitions

Antibody variable domains share a conserved framework with hypervariable loops threaded through it. Numbering schemes exist so that the same structural position carries the same number across every antibody, no matter what insertions or deletions the loops contain. There are several schemes, they disagree, and mixing them is a reliable way to produce confusion that survives into a report.

The schemes in use

Kabat came first and is based on sequence variability. Its CDR definitions are sequence-derived, and it handles insertions with letter suffixes, so a long CDR-H3 runs 100, 100A, 100B and onward. Much of the older literature, and many legacy liability tables, are in Kabat.

Chothia is based on structure. It defines the loops by their structural conformation rather than by variability, so the Chothia CDR boundaries differ from Kabat, most visibly at CDR-H1. The Martin scheme, often called enhanced Chothia, fixes inconsistencies in how Chothia handles some insertions.

IMGT is a unified scheme built on a structural alignment of all immunoglobulin and T cell receptor domains. It gives every position a fixed number with a defined gap pattern, which makes it the most portable across species and across VHH, Fab and TCR. Most modern tools default to it.

AHo is another structure-based scheme, designed for consistent spacing and widely used in engineering work and in some framework libraries.

Why it matters in practice

The CDR boundaries differ between schemes, so "the CDRs" is ambiguous unless you say which definition you used. That matters whenever a rule depends on being inside or outside a CDR, which covers most developability work: a deamidation motif that you are willing to edit in the framework but not in the paratope, a humanization exercise that grafts loops onto an acceptor framework, or a liability report that counts motifs by region.

It also matters for anything position-based. A statement like "the liability at H55" carries no information unless the scheme is named. The same residue can be H52a in Kabat and H57 in IMGT.

Working rules that save time

• Pick one scheme as the project's internal standard and convert at the edges. IMGT is a reasonable default for new work.
• Number with a tool, not by eye. ANARCI is the standard open option and handles VHH as well as paired chains.
• Record the scheme in the data, not in the method section. Column names like `cdr3_imgt` survive being copied into somebody else's spreadsheet; a caption does not.
• Expect CDR-H3 to break your assumptions. It varies most in length and conformation, it is where insertion codes pile up, and it is where the schemes differ most in what they include.
• For VHH, check the framework 2 positions. The hallmark substitutions that make a heavy chain soluble without a light chain sit there, and they are a common source of numbering confusion when a VHH is treated like a VH.