lhallee commited on
Commit
3effea8
·
verified ·
1 Parent(s): 7b48848

Canonical embedding sequences v1: full per-token rows for Atlas's 330,937 proteins only

Browse files
Files changed (2) hide show
  1. README.md +5 -5
  2. sequences.parquet +2 -2
README.md CHANGED
@@ -14,7 +14,7 @@ dataset_info:
14
  - {name: sequence, dtype: string}
15
  - {name: full, dtype: bool}
16
  splits:
17
- - {name: train, num_examples: 3567465}
18
  ---
19
  # Canonical embedding sequences
20
 
@@ -27,9 +27,9 @@ labels, pairs, source-membership columns or biological safety classification.
27
 
28
  | Quantity | Count |
29
  |---|---:|
30
- | Sequences | 3,567,465 |
31
- | Full per-token embeddings required | 330,783 |
32
- | Pooled only | 3,236,682 |
33
 
34
  ## Homology leakage
35
 
@@ -43,7 +43,7 @@ the whole inventory and supplies no model evaluation or leakage guarantee.
43
  3. Removed ASCII whitespace and uppercased letters without rewriting residues; deduplicated by SHA-256.
44
  4. Combined full-embedding requirements with boolean OR and verified all source coverage and written rows.
45
  5. Appended the proteins of the Atlas v3 interaction corpus the list lacked, and required full per-token embeddings for every one of its proteins.
46
- 6. Required full per-token embeddings for the proteins of the Atlas datasets only (PPI, PLI and CAMP: ppi_experimental_v1, atlas_v3_ppi, atlas_v3_balanced_ppi, bernett_gold_ppi_2, flock_ppi, pli_experimental_v2, annotation_experimental_v2), appending those the list lacked; every other protein needs pooled embeddings only (2026-10-19).
47
 
48
  ## Columns
49
 
 
14
  - {name: sequence, dtype: string}
15
  - {name: full, dtype: bool}
16
  splits:
17
+ - {name: train, num_examples: 3567544}
18
  ---
19
  # Canonical embedding sequences
20
 
 
27
 
28
  | Quantity | Count |
29
  |---|---:|
30
+ | Sequences | 3,567,544 |
31
+ | Full per-token embeddings required | 330,937 |
32
+ | Pooled only | 3,236,607 |
33
 
34
  ## Homology leakage
35
 
 
43
  3. Removed ASCII whitespace and uppercased letters without rewriting residues; deduplicated by SHA-256.
44
  4. Combined full-embedding requirements with boolean OR and verified all source coverage and written rows.
45
  5. Appended the proteins of the Atlas v3 interaction corpus the list lacked, and required full per-token embeddings for every one of its proteins.
46
+ 6. Required full per-token embeddings for the proteins of the Atlas datasets only (PPI, PLI and CAMP: ppi_experimental_v1, atlas_v3_ppi, atlas_v3_balanced_ppi, bernett_gold_ppi, bernett_gold_ppi_2, flock_ppi, pli_experimental_v2, annotation_experimental_v2), appending those the list lacked; every other protein needs pooled embeddings only (2026-10-19).
47
 
48
  ## Columns
49
 
sequences.parquet CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:a795299b7ed2609674221ba6c17e7785e7d6db442762365bcd8859050751dcd4
3
- size 716365788
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5939d8d503f61b365c85891f32598ab8eb38dc7fd1bdcab6c39681a3fc221c57
3
+ size 716402686