swhitfield commited on
Commit
c8ef048
·
verified ·
1 Parent(s): 5f7aa04

Add sourced InterPro annotation datasets

Browse files
README.md CHANGED
@@ -12,7 +12,7 @@ recovery of curated InterPro classifications, not an experimental assay result.
12
 
13
  - InterPro metadata: `https://ftp.ebi.ac.uk/pub/databases/interpro/current_release/entry.list`
14
  - UniProt REST API: `https://rest.uniprot.org/uniprotkb/stream`
15
- - Card generated (UTC): `2026-10-05`
16
  - Organism Taxonomy ID: `9606` (`all` = all organisms).
17
  - Review status: `reviewed` (`reviewed` = Swiss-Prot, `unreviewed` = TrEMBL, `all` = both).
18
  - UniProt query: `(organism_id:9606) AND (reviewed:true)`
@@ -21,39 +21,47 @@ recovery of curated InterPro classifications, not an experimental assay result.
21
  - Consider entries observed in at least 5 proteins.
22
  - After splitting, retain entries with at least 20
23
  positives in training, 3 in validation,
24
- and 3 in test.
 
25
  - `targets` is a multi-hot vector ordered by InterPro accession. Accession,
26
  display name, and type are recorded in `label_vocabulary.json`.
27
- - Proteins without a retained entry before splitting are dropped; after
28
- support filtering, other split variants may contain all-zero targets.
 
29
 
30
  ## Splits
31
 
32
- Whole MMseqs2 `easy-linclust` clusters are assigned to splits targeting
33
- `{'train': 0.8, 'validation': 0.1, 'test': 0.1}`, with minimum identity
34
- `0.3`, minimum coverage
35
- `0.8`, and
36
- `1` thread(s). Seed: `1957723`.
37
- When `create_split_subsets` is enabled, pooled random, stratified, and
38
- hold-cluster-out subsets are also included. Hold-cluster-out subsets require
39
- enough MMseqs clusters to populate all three roles.
 
 
 
 
 
 
40
  The optional maximum sequence length is `None` and is
41
  applied before vocabulary construction and MMseqs2 clustering.
42
 
43
  ## Split sizes
44
 
45
- - `test`: 73 rows
46
- - `test_cluster`: 73 rows
47
- - `test_random`: 73 rows
48
- - `test_stratified`: 73 rows
49
- - `train`: 581 rows
50
- - `train_cluster`: 581 rows
51
- - `train_random`: 581 rows
52
- - `train_stratified`: 581 rows
53
- - `validation`: 73 rows
54
- - `validation_cluster`: 73 rows
55
- - `validation_random`: 73 rows
56
- - `validation_stratified`: 73 rows
57
 
58
  Vocabulary size: 2.
59
 
 
12
 
13
  - InterPro metadata: `https://ftp.ebi.ac.uk/pub/databases/interpro/current_release/entry.list`
14
  - UniProt REST API: `https://rest.uniprot.org/uniprotkb/stream`
15
+ - Card generated (UTC): `2026-10-07`
16
  - Organism Taxonomy ID: `9606` (`all` = all organisms).
17
  - Review status: `reviewed` (`reviewed` = Swiss-Prot, `unreviewed` = TrEMBL, `all` = both).
18
  - UniProt query: `(organism_id:9606) AND (reviewed:true)`
 
21
  - Consider entries observed in at least 5 proteins.
22
  - After splitting, retain entries with at least 20
23
  positives in training, 3 in validation,
24
+ and 3 in test. Because validation and test
25
+ support select the entries, scores are conditional on that label selection.
26
  - `targets` is a multi-hot vector ordered by InterPro accession. Accession,
27
  display name, and type are recorded in `label_vocabulary.json`.
28
+ - Proteins without a retained entry are dropped, both before splitting and
29
+ again after split-support filtering, so no row has all-zero targets.
30
+ - `id` is the UniProtKB accession of the first entry carrying each unique sequence.
31
 
32
  ## Splits
33
 
34
+ Whole MMseqs2 clusters are assigned to splits targeting
35
+ `{'train': 0.8, 'validation': 0.1, 'test': 0.1}`. Seed: `1957723`. Clustering settings:
36
+
37
+ - MMseqs workflow: `easy-cluster`
38
+ - Minimum sequence identity (`--min-seq-id`): `0.3`
39
+ - Minimum coverage (`-c`): `0.8`
40
+ - Coverage mode (`--cov-mode`): `1`
41
+ - Threads: `1`
42
+ - Other MMseqs2 options use MMseqs2 defaults.
43
+
44
+ When `create_split_subsets` is enabled, pooled random, stratified (by each
45
+ protein's globally rarest positive entry), and hold-cluster-out subsets are
46
+ also included. Hold-cluster-out subsets require enough MMseqs clusters to
47
+ populate all three roles.
48
  The optional maximum sequence length is `None` and is
49
  applied before vocabulary construction and MMseqs2 clustering.
50
 
51
  ## Split sizes
52
 
53
+ - `test`: 58 rows
54
+ - `test_cluster`: 58 rows
55
+ - `test_random`: 58 rows
56
+ - `test_stratified`: 58 rows
57
+ - `train`: 436 rows
58
+ - `train_cluster`: 436 rows
59
+ - `train_random`: 436 rows
60
+ - `train_stratified`: 436 rows
61
+ - `validation`: 61 rows
62
+ - `validation_cluster`: 61 rows
63
+ - `validation_random`: 61 rows
64
+ - `validation_stratified`: 61 rows
65
 
66
  Vocabulary size: 2.
67
 
data/test-00000-of-00001.parquet CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:f0303fb0199c5f1ae7345a1c78ad9c667bd667ff63485f98023c537bcc721570
3
- size 25201
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c07138574301d92514241d212f80f9529a3c022a8e35958575e0ddcdf73adf9d
3
+ size 20681
data/test_cluster-00000-of-00001.parquet CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:fd8ed4a261c642c9d2f654355b3a9cbdae8d8da35455f70607037c8aba4cdd82
3
- size 24478
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:33807d2d10bbb52b35948ec7606e90d6be5dcad6b071d591a1d1470da14aa273
3
+ size 21534
data/test_random-00000-of-00001.parquet CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:ebfbc51082cb37a4db811bc5867e836dfc3d557b80a5a4b04ba7329284ba00a6
3
- size 31865
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9e0d8e42b2b7b640bed3620136a227fe51d3a56c7dd0667edea799d6a1cc0534
3
+ size 28126
data/test_stratified-00000-of-00001.parquet CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:09883bd19f0cc2ca6af766758bee39ea7595630b6ba87e0fa55f603dab0bfb67
3
- size 31847
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:64845c917ca5747d611347d0a01e986f036b8ac49027b8b69c077bbb1f8dfa3e
3
+ size 24294
data/train-00000-of-00001.parquet CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:71ba838af69aec06ca01c68c82c9126df1e37e778220a7c579b2be8a0181a511
3
- size 212735
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:973aea9a833f9367dc69ad01f2a9a0012913da7d2367a3b895eff759d1084d0f
3
+ size 159787
data/train_cluster-00000-of-00001.parquet CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:1f9e0d02f2f1d11e51cef0dc6743c1e49c3a52eea2ad5129423a07b5307386d9
3
- size 217059
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9a0bbcf60aba257548782311c604c0a6b0b7708ce60e1601fc42d44ebc1fa84a
3
+ size 154495
data/train_random-00000-of-00001.parquet CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:3f93ffed4e87abe0bf20ea448fba35d484cee0a1d45ed899084d93164c871fae
3
- size 217968
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0485d4d50c9bb986a56d4cffbcef37b195d15678a3dc176bf9b0a49b6a56b456
3
+ size 161911
data/train_stratified-00000-of-00001.parquet CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:b16b4528130bd9e92df97ca9c76981d8dac24efb7c00ad81c7d559b827b6f11e
3
- size 213912
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f4cf8f3da94be3e765a1c710b0bdbda88729b4382ad89917beecd556dbc294db
3
+ size 163001
data/validation-00000-of-00001.parquet CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:8ef0a4b68c878bec57aea5128bfaa977f18f899b6926c3cf9b9281b4b3e5daf8
3
- size 27082
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ac502ddafe0cf2878910114a3ca5d8e88ce9ac8a0d30937431ebfb41f1aad216
3
+ size 24037
data/validation_cluster-00000-of-00001.parquet CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:1d9969d324f377fec8ff818b7c2c33dd508c43303ee799074a6431d740a55749
3
- size 24868
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:32718d3d2381150e5f51da06a8260578346dbd254c910626b91782ef72d78f34
3
+ size 28133
data/validation_random-00000-of-00001.parquet CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:7f5750a6c59ccae25532e564745bd997457724deef42b9c7f478d039acccc1de
3
- size 29949
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9f9009a0827be7d4a9ab3155d3161ea0b751d84a1d0f08cdd46e5cd9a8759abc
3
+ size 24193
data/validation_stratified-00000-of-00001.parquet CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:5327bf6902ff8448e8edecb8ce681bcf1fd44742757c80f4a5ead3deec04c9c1
3
- size 30870
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5d975f145a1bf48d49250a5297656360b7b95fe4c63a289a28309834ae6f9ebf
3
+ size 26214
stats.json CHANGED
@@ -1,18 +1,19 @@
1
  {
2
- "rows_total": 2908,
 
3
  "rows_by_split": {
4
- "train": 581,
5
- "validation": 73,
6
- "test": 73,
7
- "train_random": 581,
8
- "validation_random": 73,
9
- "test_random": 73,
10
- "train_stratified": 581,
11
- "validation_stratified": 73,
12
- "test_stratified": 73,
13
- "train_cluster": 581,
14
- "validation_cluster": 73,
15
- "test_cluster": 73
16
  },
17
  "columns": [
18
  "id",
@@ -24,184 +25,184 @@
24
  {
25
  "file": "test.fasta.gz",
26
  "type": "Protein",
27
- "num_seqs": 73,
28
- "sum_len": 47697,
29
  "min_len": 79,
30
- "avg_len": 653.4,
31
- "max_len": 2054,
32
- "Q1": 415.0,
33
- "Q2": 562.0,
34
- "Q3": 917.0,
35
  "sum_gap": 0,
36
- "sum_n": 1757
37
  },
38
  {
39
  "file": "test_cluster.fasta.gz",
40
  "type": "Protein",
41
- "num_seqs": 73,
42
- "sum_len": 46364,
43
- "min_len": 142,
44
- "avg_len": 635.1,
45
- "max_len": 2058,
46
- "Q1": 444.0,
47
- "Q2": 496.0,
48
- "Q3": 740.0,
49
  "sum_gap": 0,
50
- "sum_n": 1674
51
  },
52
  {
53
  "file": "test_random.fasta.gz",
54
  "type": "Protein",
55
- "num_seqs": 73,
56
- "sum_len": 53412,
57
- "min_len": 101,
58
- "avg_len": 731.7,
59
- "max_len": 4024,
60
- "Q1": 315.0,
61
- "Q2": 553.0,
62
- "Q3": 898.0,
63
  "sum_gap": 0,
64
- "sum_n": 2053
65
  },
66
  {
67
  "file": "test_stratified.fasta.gz",
68
  "type": "Protein",
69
- "num_seqs": 73,
70
- "sum_len": 54761,
71
- "min_len": 92,
72
- "avg_len": 750.2,
73
- "max_len": 2548,
74
- "Q1": 392.0,
75
- "Q2": 596.0,
76
- "Q3": 832.0,
77
  "sum_gap": 0,
78
- "sum_n": 2016
79
  },
80
  {
81
  "file": "train.fasta.gz",
82
  "type": "Protein",
83
- "num_seqs": 581,
84
- "sum_len": 434567,
85
- "min_len": 61,
86
- "avg_len": 748.0,
87
  "max_len": 7570,
88
- "Q1": 384.0,
89
- "Q2": 567.0,
90
- "Q3": 864.0,
91
  "sum_gap": 0,
92
- "sum_n": 15777
93
  },
94
  {
95
  "file": "train_cluster.fasta.gz",
96
  "type": "Protein",
97
- "num_seqs": 581,
98
- "sum_len": 441072,
99
- "min_len": 61,
100
- "avg_len": 759.2,
101
- "max_len": 7570,
102
- "Q1": 393.0,
103
- "Q2": 578.0,
104
- "Q3": 898.0,
105
  "sum_gap": 0,
106
- "sum_n": 16205
107
  },
108
  {
109
  "file": "train_random.fasta.gz",
110
  "type": "Protein",
111
- "num_seqs": 581,
112
- "sum_len": 425708,
113
- "min_len": 61,
114
- "avg_len": 732.7,
115
  "max_len": 7968,
116
- "Q1": 374.0,
117
- "Q2": 543.0,
118
- "Q3": 847.0,
119
  "sum_gap": 0,
120
- "sum_n": 15343
121
  },
122
  {
123
  "file": "train_stratified.fasta.gz",
124
  "type": "Protein",
125
- "num_seqs": 581,
126
- "sum_len": 424462,
127
- "min_len": 61,
128
- "avg_len": 730.6,
129
  "max_len": 7968,
130
- "Q1": 379.0,
131
- "Q2": 551.0,
132
- "Q3": 842.0,
133
  "sum_gap": 0,
134
- "sum_n": 15365
135
  },
136
  {
137
  "file": "validation.fasta.gz",
138
  "type": "Protein",
139
- "num_seqs": 73,
140
- "sum_len": 49414,
141
- "min_len": 61,
142
- "avg_len": 676.9,
143
  "max_len": 7968,
144
- "Q1": 334.0,
145
- "Q2": 481.0,
146
- "Q3": 666.0,
147
  "sum_gap": 0,
148
- "sum_n": 1748
149
  },
150
  {
151
  "file": "validation_cluster.fasta.gz",
152
  "type": "Protein",
153
- "num_seqs": 73,
154
- "sum_len": 44242,
155
- "min_len": 61,
156
- "avg_len": 606.1,
157
- "max_len": 7968,
158
- "Q1": 109.0,
159
- "Q2": 392.0,
160
- "Q3": 796.0,
161
  "sum_gap": 0,
162
- "sum_n": 1403
163
  },
164
  {
165
  "file": "validation_random.fasta.gz",
166
  "type": "Protein",
167
- "num_seqs": 73,
168
- "sum_len": 52558,
169
- "min_len": 61,
170
- "avg_len": 720.0,
171
- "max_len": 2850,
172
- "Q1": 444.0,
173
- "Q2": 627.0,
174
- "Q3": 839.0,
175
  "sum_gap": 0,
176
- "sum_n": 1886
177
  },
178
  {
179
  "file": "validation_stratified.fasta.gz",
180
  "type": "Protein",
181
- "num_seqs": 73,
182
- "sum_len": 52455,
183
- "min_len": 93,
184
- "avg_len": 718.6,
185
- "max_len": 3530,
186
- "Q1": 386.0,
187
  "Q2": 544.0,
188
- "Q3": 896.0,
189
  "sum_gap": 0,
190
- "sum_n": 1901
191
  },
192
  {
193
  "file": "overall",
194
  "type": "Protein",
195
- "num_seqs": 2908,
196
- "sum_len": 2126712,
197
- "min_len": 61,
198
- "avg_len": 731.3,
199
  "max_len": 7968,
200
- "Q1": 382.0,
201
- "Q2": 552.0,
202
- "Q3": 857.0,
203
  "sum_gap": 0,
204
- "sum_n": 77128
205
  }
206
  ]
207
  }
 
1
  {
2
+ "rows_total": 2220,
3
+ "canonical_rows": 555,
4
  "rows_by_split": {
5
+ "train": 436,
6
+ "validation": 61,
7
+ "test": 58,
8
+ "train_random": 436,
9
+ "validation_random": 61,
10
+ "test_random": 58,
11
+ "train_stratified": 436,
12
+ "validation_stratified": 61,
13
+ "test_stratified": 58,
14
+ "train_cluster": 436,
15
+ "validation_cluster": 61,
16
+ "test_cluster": 58
17
  },
18
  "columns": [
19
  "id",
 
25
  {
26
  "file": "test.fasta.gz",
27
  "type": "Protein",
28
+ "num_seqs": 58,
29
+ "sum_len": 38964,
30
  "min_len": 79,
31
+ "avg_len": 671.8,
32
+ "max_len": 2391,
33
+ "Q1": 351.0,
34
+ "Q2": 466.5,
35
+ "Q3": 1009.0,
36
  "sum_gap": 0,
37
+ "sum_n": 1377
38
  },
39
  {
40
  "file": "test_cluster.fasta.gz",
41
  "type": "Protein",
42
+ "num_seqs": 58,
43
+ "sum_len": 42710,
44
+ "min_len": 170,
45
+ "avg_len": 736.4,
46
+ "max_len": 4061,
47
+ "Q1": 351.0,
48
+ "Q2": 644.5,
49
+ "Q3": 869.0,
50
  "sum_gap": 0,
51
+ "sum_n": 1309
52
  },
53
  {
54
  "file": "test_random.fasta.gz",
55
  "type": "Protein",
56
+ "num_seqs": 58,
57
+ "sum_len": 47486,
58
+ "min_len": 90,
59
+ "avg_len": 818.7,
60
+ "max_len": 2527,
61
+ "Q1": 476.0,
62
+ "Q2": 657.0,
63
+ "Q3": 1037.0,
64
  "sum_gap": 0,
65
+ "sum_n": 1806
66
  },
67
  {
68
  "file": "test_stratified.fasta.gz",
69
  "type": "Protein",
70
+ "num_seqs": 58,
71
+ "sum_len": 41055,
72
+ "min_len": 79,
73
+ "avg_len": 707.8,
74
+ "max_len": 1620,
75
+ "Q1": 470.0,
76
+ "Q2": 659.5,
77
+ "Q3": 864.0,
78
  "sum_gap": 0,
79
+ "sum_n": 1490
80
  },
81
  {
82
  "file": "train.fasta.gz",
83
  "type": "Protein",
84
+ "num_seqs": 436,
85
+ "sum_len": 321786,
86
+ "min_len": 90,
87
+ "avg_len": 738.0,
88
  "max_len": 7570,
89
+ "Q1": 399.5,
90
+ "Q2": 594.0,
91
+ "Q3": 822.0,
92
  "sum_gap": 0,
93
+ "sum_n": 11449
94
  },
95
  {
96
  "file": "train_cluster.fasta.gz",
97
  "type": "Protein",
98
+ "num_seqs": 436,
99
+ "sum_len": 308430,
100
+ "min_len": 79,
101
+ "avg_len": 707.4,
102
+ "max_len": 7968,
103
+ "Q1": 368.5,
104
+ "Q2": 548.0,
105
+ "Q3": 803.0,
106
  "sum_gap": 0,
107
+ "sum_n": 11012
108
  },
109
  {
110
  "file": "train_random.fasta.gz",
111
  "type": "Protein",
112
+ "num_seqs": 436,
113
+ "sum_len": 319098,
114
+ "min_len": 79,
115
+ "avg_len": 731.9,
116
  "max_len": 7968,
117
+ "Q1": 367.0,
118
+ "Q2": 572.5,
119
+ "Q3": 855.5,
120
  "sum_gap": 0,
121
+ "sum_n": 11127
122
  },
123
  {
124
  "file": "train_stratified.fasta.gz",
125
  "type": "Protein",
126
+ "num_seqs": 436,
127
+ "sum_len": 322325,
128
+ "min_len": 90,
129
+ "avg_len": 739.3,
130
  "max_len": 7968,
131
+ "Q1": 370.0,
132
+ "Q2": 575.5,
133
+ "Q3": 840.0,
134
  "sum_gap": 0,
135
+ "sum_n": 11319
136
  },
137
  {
138
  "file": "validation.fasta.gz",
139
  "type": "Protein",
140
+ "num_seqs": 61,
141
+ "sum_len": 46971,
142
+ "min_len": 190,
143
+ "avg_len": 770.0,
144
  "max_len": 7968,
145
+ "Q1": 399.0,
146
+ "Q2": 596.0,
147
+ "Q3": 976.0,
148
  "sum_gap": 0,
149
+ "sum_n": 1597
150
  },
151
  {
152
  "file": "validation_cluster.fasta.gz",
153
  "type": "Protein",
154
+ "num_seqs": 61,
155
+ "sum_len": 56581,
156
+ "min_len": 214,
157
+ "avg_len": 927.6,
158
+ "max_len": 1721,
159
+ "Q1": 616.0,
160
+ "Q2": 976.0,
161
+ "Q3": 1239.0,
162
  "sum_gap": 0,
163
+ "sum_n": 2102
164
  },
165
  {
166
  "file": "validation_random.fasta.gz",
167
  "type": "Protein",
168
+ "num_seqs": 61,
169
+ "sum_len": 41137,
170
+ "min_len": 101,
171
+ "avg_len": 674.4,
172
+ "max_len": 2986,
173
+ "Q1": 403.0,
174
+ "Q2": 596.0,
175
+ "Q3": 784.0,
176
  "sum_gap": 0,
177
+ "sum_n": 1490
178
  },
179
  {
180
  "file": "validation_stratified.fasta.gz",
181
  "type": "Protein",
182
+ "num_seqs": 61,
183
+ "sum_len": 44341,
184
+ "min_len": 109,
185
+ "avg_len": 726.9,
186
+ "max_len": 1693,
187
+ "Q1": 416.0,
188
  "Q2": 544.0,
189
+ "Q3": 1036.0,
190
  "sum_gap": 0,
191
+ "sum_n": 1614
192
  },
193
  {
194
  "file": "overall",
195
  "type": "Protein",
196
+ "num_seqs": 555,
197
+ "sum_len": 407721,
198
+ "min_len": 79,
199
+ "avg_len": 734.6,
200
  "max_len": 7968,
201
+ "Q1": 386.5,
202
+ "Q2": 588.0,
203
+ "Q3": 860.5,
204
  "sum_gap": 0,
205
+ "sum_n": 14423
206
  }
207
  ]
208
  }