anindya64 commited on
Commit
338d0ed
·
verified ·
1 Parent(s): 5241ffa

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +19 -4
README.md CHANGED
@@ -18,11 +18,9 @@ configs:
18
  path: data/test-*.parquet
19
  ---
20
 
21
- # UniRef50 Shard Index
22
 
23
- This dataset contains the original UniRef50 FASTA shards plus a viewer-friendly file/shard index. The full sequence data is stored as 61 `.fasta.zst` shards and the per-record metadata JSONL is very large, so the default Dataset Viewer table indexes repository files instead of expanding all 60,315,044 sequence records.
24
-
25
- Use the original `sequences/.../shard-*.fasta.zst` files for complete FASTA records. Use the default Parquet table for Dataset Viewer previews, source discovery, file sizes, and download patterns.
26
 
27
  ## Splits
28
 
@@ -124,3 +122,20 @@ with shard.open("rb") as f, dctx.stream_reader(f) as reader:
124
  ## Preparation
125
 
126
  The normalization script used to create the Parquet files is included at `scripts/prepare_uniref50_dataset.py`.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
18
  path: data/test-*.parquet
19
  ---
20
 
21
+ # UniRef50
22
 
23
+ UniRef50 is one of three clustering levels in the UniProt Reference Clusters (UniRef100 / UniRef90 / UniRef50). It is produced by clustering UniRef90 cluster representatives with MMseqs2 using a 50% sequence identity threshold and an 80% length overlap, yielding a substantially condensed view of UniProtKB plus selected UniParc records. UniRef50 is the workhorse search and training database for protein ML: AlphaFold2 and most protein language models (ESM, ProtTrans, etc.) use UniRef50 or its close cousin UniRef90 as the primary diverse protein corpus, since the redundancy reduction gives a much flatter sample of sequence space than raw UniProtKB.
 
 
24
 
25
  ## Splits
26
 
 
122
  ## Preparation
123
 
124
  The normalization script used to create the Parquet files is included at `scripts/prepare_uniref50_dataset.py`.
125
+
126
+
127
+ # Citation
128
+
129
+ ```
130
+ @article{suzek2015uniref,
131
+ title = {{UniRef} clusters: a comprehensive and scalable alternative for improving sequence similarity searches},
132
+ author = {Suzek, Baris E. and Wang, Yuqi and Huang, Hongzhan and McGarvey, Peter B. and Wu, Cathy H. and {UniProt Consortium}},
133
+ journal = {Bioinformatics},
134
+ volume = {31},
135
+ number = {6},
136
+ pages = {926--932},
137
+ year = {2015},
138
+ publisher = {Oxford University Press},
139
+ doi = {10.1093/bioinformatics/btu739}
140
+ }
141
+ ```