UniRef

UniRef, short for UniProt Reference Clusters, is a collection of non-redundant protein sequence clusters produced by the UniProt consortium. It combines sequences from UniProtKB and the UniParc database to reduce redundancy and speed large-scale analyses. UniRef provides three levels of clustering: UniRef100, UniRef90, and UniRef50. In each cluster, sequences that meet the corresponding identity threshold are grouped together, and a single representative sequence is assigned to the cluster. UniRef100 contains all identical sequences (100% identity) across the source data; UniRef90 groups sequences with at least 90% identity to the seed; UniRef50 uses a 50% identity threshold. Each cluster includes metadata and a mapping to member entries in UniProtKB and related resources.

The clusters are generated by hierarchical clustering algorithms applied to the protein sequences in UniProtKB and UniParc, and they provide a compact, non-redundant representation of protein sequence space. The representative sequence serves as a proxy for the whole cluster in many analyses. UniRef identifiers link to the cluster’s metadata and to the list of member sequences, enabling traceability to original UniProt entries.

Applications include speeding up sequence similarity searches, reducing database size for large-scale analyses, and facilitating functional annotation transfer across related proteins. UniRef is widely used in bioinformatics workflows, proteomics pipelines, and comparative genomics studies.

Access and formats: UniRef data are publicly available from the UniProt website and FTP site in flat files and downloadable formats. Users can retrieve representative sequences, access cluster mappings, and download member lists. The resource is updated periodically to align with updates to UniProtKB and UniParc.