Add files using upload-large-folder tool
Browse files
README.md
CHANGED
|
@@ -52,10 +52,10 @@ paper_test = causal.filter(
|
|
| 52 |
|
| 53 |
| Scope | Rows | Held-out `test` rows |
|
| 54 |
|---|---:|---:|
|
| 55 |
-
| Complete release | 186,
|
| 56 |
-
| Corrected paper categories | 167,
|
| 57 |
|
| 58 |
-
Labels are balanced: 93,
|
| 59 |
Every pair shares `pair_id`.
|
| 60 |
|
| 61 |
The row-level `split` uses held-out chromosomes `chr2`, `chr3`, `chr16`, and
|
|
@@ -74,7 +74,7 @@ For each control, the sampler avoids the union of retained annotation
|
|
| 74 |
intervals and previously selected annotation-free controls. The control has
|
| 75 |
the same chromosome and length as its feature. Independent verification
|
| 76 |
repeats this check without using the sampler and finds zero overlaps across
|
| 77 |
-
all 93,
|
| 78 |
context contains only uppercase `A/C/G/T`.
|
| 79 |
|
| 80 |
Exclude `noncoding_regions` for the corrected ten-category paper view.
|
|
@@ -122,6 +122,8 @@ ambiguous feature/control group and every conflicting-label exact-input group.
|
|
| 122 |
Same-label duplicates remaining within a config share `context_group_id`;
|
| 123 |
independent checks find zero exact inputs crossing train/test and zero
|
| 124 |
conflicting-label groups.
|
|
|
|
|
|
|
| 125 |
|
| 126 |
## Processing and citation
|
| 127 |
|
|
|
|
| 52 |
|
| 53 |
| Scope | Rows | Held-out `test` rows |
|
| 54 |
|---|---:|---:|
|
| 55 |
+
| Complete release | 186,370 | 37,240 |
|
| 56 |
+
| Corrected paper categories | 167,234 | 33,450 |
|
| 57 |
|
| 58 |
+
Labels are balanced: 93,185 `feature` and 93,185 `random-noannot` rows.
|
| 59 |
Every pair shares `pair_id`.
|
| 60 |
|
| 61 |
The row-level `split` uses held-out chromosomes `chr2`, `chr3`, `chr16`, and
|
|
|
|
| 74 |
intervals and previously selected annotation-free controls. The control has
|
| 75 |
the same chromosome and length as its feature. Independent verification
|
| 76 |
repeats this check without using the sampler and finds zero overlaps across
|
| 77 |
+
all 93,185 controls. It also confirms every causal and bidirectional control
|
| 78 |
context contains only uppercase `A/C/G/T`.
|
| 79 |
|
| 80 |
Exclude `noncoding_regions` for the corrected ten-category paper view.
|
|
|
|
| 122 |
Same-label duplicates remaining within a config share `context_group_id`;
|
| 123 |
independent checks find zero exact inputs crossing train/test and zero
|
| 124 |
conflicting-label groups.
|
| 125 |
+
The held-out ATG benchmark has priority over an exact duplicate functional
|
| 126 |
+
training input, so the affected chr7 coding pair is removed consistently.
|
| 127 |
|
| 128 |
## Processing and citation
|
| 129 |
|
functional-random-noannot-gamba-bidi.parquet
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1b609aae14350a88c7bbd14e9b578b6dc19c27278fb74c75ab073233f383ee4d
|
| 3 |
+
size 135822251
|
functional-random-noannot-gamba-causal.parquet
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a74978d642d64f8086f204f24e843dd0860e67c87a3312b6b8684a2db3b98649
|
| 3 |
+
size 128879246
|