What are complementarity-determining regions, and how are they numbered?

What is a complementarity-determining region?

An antibody recognises its target with its variable domains, and within each variable domain the binding happens at three short stretches of sequence: the complementarity-determining regions. A set of CDRs forms the paratope, the antigen-binding site. Four framework regions, FR1 to FR4, sit between and around them and make up the scaffold of the domain. Read from the start of the domain, the order is always FR1, CDR1, FR2, CDR2, FR3, CDR3, FR4. The classical CDR definitions are an approximation of the paratope, not a measurement of it.

A heavy-chain variable domain has three CDRs, called H1, H2 and H3, and a light-chain variable domain has three more, L1, L2 and L3. A heavy and a light domain together make one binding site with six CDRs, and a typical IgG has two sites, so twelve CDRs. A nanobody (VHH) is a single variable domain taken from the heavy-chain-only antibodies of camelids, so it has three CDRs and no light chain: its whole binding site is three loops on one domain. We compare the formats in our note on VHH, scFv and IgG formats.

CDR1 and CDR2 lie within the V region. CDR3 includes part of V, all of D (in heavy chains only) and J, and it is the most variable of the three.

Why a CDR needs a numbering scheme

The sequence alone does not say where a CDR begins. Loops differ in length from one antibody to the next, so the 30th residue of one antibody is not the 30th residue of another. A numbering scheme fixes that by giving every residue of a variable domain an address: position 52 is the same place in every antibody, a loop that is longer than usual takes extra positions with a letter, such as 52A, and a CDR becomes a range of addresses instead of a stretch of letters.

That is what lets you compare antibodies position by position, name a mutation without ambiguity, or cut the CDRs out of a sequence. Extracting a CDR means numbering the sequence, then reading off the residues whose numbers fall inside the range that your scheme gives to that CDR.

Three schemes, three sets of boundaries

Kabat, Chothia and IMGT were built for different purposes.

• Kabat comes from aligning sequences and looking for where they vary: its CDRs are the hypervariable stretches (Wu and Kabat, 1970). Residues beyond the usual length are labelled with letters, such as 52A or 100B.
• Chothia is based on structure: it places the CDRs where three-dimensional structures show the antigen-binding loops (Chothia and Lesk, 1987). It moves the point where insertions occur in the first heavy-chain loop from position 35 to position 31, so the Kabat and Chothia numbers diverge where a CDR-H1 is longer than usual, and its CDR-H1 runs from position 26 to 32.
• IMGT numbers every variable domain the same way, heavy, light and T-cell receptor chains alike, continuously from 1 to 128, on the basis of an alignment of germline V genes. It has its own definitions of the framework regions and CDRs (FR-IMGT and CDR-IMGT), and it avoids insertion codes except between positions 111 and 112 in a long CDR3.

On a heavy chain, the three schemes put the CDRs at these positions:

• CDR-H1: IMGT 27 to 38, Kabat 31 to 35, Chothia 26 to 32
• CDR-H2: IMGT 56 to 65, Kabat 50 to 65, Chothia 52 to 56
• CDR-H3: IMGT 105 to 117, Kabat 95 to 102, Chothia 95 to 102

A lettered position belongs to the CDR of its number: in Kabat, 52A is in CDR-H2 and 100A in CDR-H3. The ranges above are those of the AbNumber library, which applies each scheme's published definition. The IMGT ones also match the region definition that ANARCI itself uses.

The same sequence, three sets of CDRs

Take the nanobody in the figure, chain E of PDB entry 3OGO. Its CDR1 has eight residues in IMGT (GFPVNRYS), seven in Chothia (GFPVNRY) and five in Kabat (RYSMR). It is one stretch of sequence, and the three schemes start and stop at different places. The human antibody, the heavy chain of adalimumab (PDB 3WD5), does the same: GFTFDDYA, GFTFDDY and DYAMH.

CDR3 shows the other kind of difference. In IMGT it begins right after the conserved cysteine at position 104 and ends before the tryptophan at position 118. In Kabat and Chothia it begins at position 95, two residues later. For the human heavy chain, that makes the CDR3 14 residues long in IMGT (AKVSYLSTASSLDY) and 12 in Kabat and Chothia (VSYLSTASSLDY). The A and K are not missing from the second version: Kabat and Chothia number them 93 and 94 and count them as framework.

The longer loop also shows how insertions are written. IMGT fits 14 residues into its 13 positions by adding one, which ANARCI writes 112A. Kabat and Chothia add four, 100A to 100D.

For these two sequences, Kabat and Chothia give every residue the same number. They differ in where they draw the CDRs, not in the labels. Neither sequence has inserted residues in CDR-H1, which is where the two numberings diverge.

How the numbers are computed

The numbering in the figure was produced with ANARCI, an open-source tool that numbers antibody and T-cell receptor variable domains in the most widely used schemes. It aligns each sequence to profile hidden Markov models (with HMMER), then converts the alignment to the scheme you ask for. We ran ANARCI 1.3 once on the two sequences and saved its output with this page: the figure shows stored results, and nothing is computed when you open it. ANARCI numbers only the variable domain, so the methionine and His tag of the nanobody and the constant domains of the heavy chain are left out.

Both sequences are public. The nanobody is chain E of PDB entry 3OGO (Kubala et al., 2010), and the human heavy chain is chain H of PDB entry 3WD5 (Hu et al., 2013).

Numbering runs on your device, not on ours, and no sequence is sent to a server.

Where this matters for design

Every design step that touches a loop starts from a numbering. When we humanise an antibody, we number the molecule by IMGT and map the residues that support the loops and the interface, and only then move the framework towards human germlines: that is how our antibody humanization work starts. A nanobody has no light chain, so its three loops are its whole binding site, and a boundary that is two residues out, as in the CDR3 above, changes what you call the binding site. If you want a VHH designed against an epitope you choose, that is our nanobody discovery service.

Sources

• Wikipedia, Complementarity-determining region (consulted 28 October 2026).
• Dondelinger M, et al. Understanding the significance and implications of antibody numbering and antigen-binding surface/residue definition. Frontiers in Immunology 9, 2278 (2018). doi:10.3389/fimmu.2018.02278
• Wu TT, Kabat EA. An analysis of the sequences of the variable regions of Bence Jones proteins and myeloma light chains and their implications for antibody complementarity. Journal of Experimental Medicine 132, 211-250 (1970). doi:10.1084/jem.132.2.211
• Chothia C, Lesk AM. Canonical structures for the hypervariable regions of immunoglobulins. Journal of Molecular Biology 196, 901-917 (1987). doi:10.1016/0022-2836(87)90412-8
• Al-Lazikani B, Lesk AM, Chothia C. Standard conformations for the canonical structures of immunoglobulins. Journal of Molecular Biology 273, 927-948 (1997). doi:10.1006/jmbi.1997.1354
• Lefranc M-P, et al. IMGT unique numbering for immunoglobulin and T cell receptor variable domains and Ig superfamily V-like domains. Developmental and Comparative Immunology 27, 55-77 (2003). doi:10.1016/S0145-305X(02)00039-3
• Hamers-Casterman C, et al. Naturally occurring antibodies devoid of light chains. Nature 363, 446-448 (1993). doi:10.1038/363446a0
• Dunbar J, Deane CM. ANARCI: antigen receptor numbering and receptor classification. Bioinformatics 32, 298-300 (2016). doi:10.1093/bioinformatics/btv552
• Prihoda D, et al. BioPhi: a platform for antibody design, humanization, and humanness evaluation based on natural antibody repertoires and deep learning. mAbs 14, 2020203 (2022). doi:10.1080/19420862.2021.2020203. Its AbNumber library: github.com/prihoda/AbNumber.
• Kubala MH, et al. Structural and thermodynamic analysis of the GFP:GFP-nanobody complex. Protein Science 19, 2389-2401 (2010). doi:10.1002/pro.519. PDB entry 3OGO.
• Hu S, et al. Comparison of the inhibition mechanisms of adalimumab and infliximab in treating tumor necrosis factor alpha-associated diseases from a molecular view. Journal of Biological Chemistry 288, 27059-27067 (2013). doi:10.1074/jbc.M113.491530. PDB entry 3WD5.