UniRef is a set of protein clusters at three redundancy levels: UniRef100, UniRef90 and UniRef50. In the KBase Data Lakehouse it appears in two groups. One is the refdata_uniref* reference-data collections. The other is the kbase_uniref* datasets 1,2.
Collections in the KBase Data Lakehouse
The reference-data collections refdata_uniref50_2026_01, refdata_uniref90_2026_01 and refdata_uniref100_2026_01 each contain a cluster table. They hold UniRef protein clusters at the three redundancy levels, taken from a 2026-01 snapshot 1.
The platform’s protein and structure inventory lists 215,130,942 uniprot proteins. It lists 475,217,233 UniRef100, 188,848,220 UniRef90 and 60,315,044 UniRef50 clusters as of 2026-01. It also lists 241,070,489 AlphaFold predicted structures (kescience-alphafold) and ~253K experimental PDB structures (protein-data-bank). These are inventory counts reported directly by the source 1.
A separate group of datasets is named kbase_uniref50, kbase_uniref90 and kbase_uniref100. These are partial or sample datasets, not the full UniRef releases, which would be hundreds of millions of clusters. A cluster lookup against them should not assume completeness 2.
Query Caveats
To join UniRef100 to UniProt, you must strip the UniRef100_ prefix, for example REPLACE(uniref100, 'UniRef100_', '') = uniprot_accession. This is a per-row string operation, so for large joins it may help to materialize a mapping table first 2.
Tensions
The two sources describe UniRef coverage differently. The data atlas reports full-scale 2026-01 cluster counts and names the refdata_uniref* collections 1. The pitfalls digest warns that the kbase_uniref* datasets are partial samples, not full releases 2. The two sources name different table groups, so both statements may hold for their own tables. Neither source says how the refdata_* and kbase_* collections relate. Before assuming a query is complete, check which collection it uses 1,2.
Related
- kbase-uniref is the data page for the
kbase_uniref*datasets 2. - BERDL Data Atlas — Inventory, Topic Map, and Cross-Reference Synergies is the source of the inventory counts and the snapshot date 1.
- pitfalls is the source of the join and completeness caveats 2.