Update README.md
Browse files
README.md
CHANGED
|
@@ -51,6 +51,10 @@ Deep learning models for unbiased sequence-based PPI prediction plateau at an ac
|
|
| 51 |
## Description
|
| 52 |
This is a balanced binary protein-protein interaction dataset with positives from [HIPPIE](https://cbdm-01.zdv.uni-mainz.de/~mschaefer/hippie/) and paritioned with [KaHIP](https://kahip.github.io/). There are no sequence overlaps in splits, furthermore, they are split based on a maximum CD-HIT of 40% pairwise sequence similarity. Node degree bias was also reduced.
|
| 53 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 54 |
## Example use
|
| 55 |
```python
|
| 56 |
def get_ppi_data():
|
|
|
|
| 51 |
## Description
|
| 52 |
This is a balanced binary protein-protein interaction dataset with positives from [HIPPIE](https://cbdm-01.zdv.uni-mainz.de/~mschaefer/hippie/) and paritioned with [KaHIP](https://kahip.github.io/). There are no sequence overlaps in splits, furthermore, they are split based on a maximum CD-HIT of 40% pairwise sequence similarity. Node degree bias was also reduced.
|
| 53 |
|
| 54 |
+
Note:
|
| 55 |
+
|
| 56 |
+
[Q96PU5](https://www.uniprot.org/uniprotkb/Q96PU5) is not located in the provided SwissProt fasta file but is used in the train split several times. We added this before dataprocessing so no rows were dropped.
|
| 57 |
+
|
| 58 |
## Example use
|
| 59 |
```python
|
| 60 |
def get_ppi_data():
|