Written and maintained by CASRAI Editorial Board
Last updated
Short answer: CAZy is the Carbohydrate-Active Enzymes database at cazy.org, a specialist resource that classifies enzymes which degrade, modify or create glycosidic bonds, together with their carbohydrate-binding modules, into numbered sequence-based families. Lab users meet it as a lookup tool: a protein is placed in a family such as GH5 or GT2, and the family page lists characterized members.
What CAZy is and who maintains it
According to the CAZy homepage (read 2026-11-08), the database has been online since 1998 and covers genomic, structural and biochemical information on carbohydrate-active enzymes, often called CAZymes. It is maintained by the AFMB laboratory (CNRS, Aix-Marseille Université, INRAE). The homepage notes regular updates, so family counts change over time; the figures below are dated values from a published description of the database, not live totals.
The standard reference for the current curation model is Drula et al., “The carbohydrate-active enzyme database: functions and literature”, Nucleic Acids Research 2021;50(D1):D571-D577 (doi:10.1093/nar/gkab1045). Related reading on this site: lysozyme (a glycoside-cleaving enzyme) and DAVID bioinformatics for another annotation resource.
How families are assigned and curated
Per Drula et al. (2021), CAZy families are sequence-based, and each family is built around at least one biochemically characterized member, so there is no family of unknown function in CAZy. New GenBank sequences and PDB structures are scanned with BLAST and HMMER. Very confident hits (e-value below 1e-50 with 100% coverage) are assigned automatically; the rest go to a human curator. Functional information is taken only from peer-reviewed literature; preprints are excluded. Users can submit new characterizations with a sequence accession, the activity or EC number, and a PMID or DOI.
The six CAZy classes
The table gives the class definitions from cazy.org (read 2026-11-08) and the number of families reported by Drula et al. in December 2013 and September 2021.
| Class | Meaning (cazy.org) | Families, Dec 2013 | Families, Sept 2021 |
|---|---|---|---|
| GH, Glycoside Hydrolases | Hydrolysis and rearrangement of glycosidic bonds | 133 | 171 |
| GT, Glycosyl Transferases | Formation of glycosidic bonds | 94 | 114 |
| PL, Polysaccharide Lyases | Non-hydrolytic cleavage of glycosidic bonds | 23 | 42 |
| CE, Carbohydrate Esterases | Hydrolysis of carbohydrate esters | 16 | 19 |
| AA, Auxiliary Activities | Redox enzymes that act in conjunction with other CAZymes | 11 | 17 |
| CBM, Carbohydrate-Binding Modules | Adhesion to carbohydrates | 68 | 88 |
How to read a CAZy family ID
A CAZy identifier is a class prefix plus a number, for example GH5 or CBM20. Two points from the 2021 paper prevent most misreadings:
- Families are numbered, not named. The number carries no information about substrate or mechanism.
- A family is defined by sequence similarity, so one family can contain enzymes with different specificities. Family membership alone does not tell you the substrate.
Family pages list characterized members with downloadable accession lists, and the database links out to GenBank, PDB and UniProt identifiers. It also provides genome pages and Krona taxonomy charts.
Where it matters in lab work
Researchers use CAZy when annotating a genome or metagenome for carbohydrate-active genes, choosing candidate enzymes to characterize, or checking whether a protein of interest has a characterized relative. Sequences deposited in public archives (see Sequence Read Archive) and curated resources (see data curation) feed this kind of annotation work, and interaction resources such as the STRING database sit alongside it in typical bioinformatics workflows.
What CAZy can and cannot tell you
- Family is not function. A sequence-based family can hold several activities; confirm activity experimentally or from the characterized members listed on the family page.
- EC numbers are not always specific. Drula et al. note, for example, that EC 3.2.1.4 (cellulase) does not distinguish retaining from inverting mechanisms, and that “open” numbers such as 3.2.1.- exist. About 100 activities in roughly 400 proteins were awaiting EC numbers.
- Curation has limits. The curation team is small, and the authors state that they miss several and perhaps many biochemical characterizations. In September 2020, 96% of GH and PL families had a literature reference, versus 76% of characterized enzymes.
- Only peer-reviewed function counts. A recent preprint may not yet be reflected.
Common misunderstandings
- “GH means a hydrolase of one type.” GH covers hydrolysis and rearrangement of glycosidic bonds across many specificities.
- “A hit in CAZy proves activity.” Automatic assignment is by sequence; only the characterized members of a family carry experimental evidence.
- “The family counts are fixed.” They are updated regularly, so cite the date you checked.
FAQ
What is CAZy?
CAZy is the Carbohydrate-Active Enzymes database; CAZymes are the enzymes it classifies.
What is the difference between GH and GT?
Glycoside hydrolases hydrolyse or rearrange glycosidic bonds; glycosyl transferases form them (cazy.org).
Can I submit a new characterization?
Yes. The 2021 paper describes user submissions that include a sequence accession, the activity or EC number, and a PMID or DOI.
Sources
- CAZy, Carbohydrate-Active Enzymes database, cazy.org, read 2026-11-08.
- Drula E, et al. Nucleic Acids Res 2021;50(D1):D571-D577, PMC8728194, doi:10.1093/nar/gkab1045, read 2026-11-08.
