add to README
Browse files
README.md
CHANGED
@@ -7,6 +7,15 @@ tags:
|
|
7 |
|
8 |
## How to use the data sets
|
9 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
10 |
### Use the already preprocessed data
|
11 |
|
12 |
Load a test/train split using
|
@@ -17,6 +26,12 @@ train = load_dataset("jglaser/binding_affinity",split='train[:90%]')
|
|
17 |
validation = load_dataset("jglaser/binding_affinity",split='train[90%:]')
|
18 |
```
|
19 |
|
|
|
|
|
|
|
|
|
|
|
|
|
20 |
**Loading the data manually**
|
21 |
|
22 |
The file `data/all.parquet` contains the preprocessed data. To extract it,
|
|
|
7 |
|
8 |
## How to use the data sets
|
9 |
|
10 |
+
This dataset contains 1.9M unique pairs of protein sequences and ligand SMILES with experimentally determined
|
11 |
+
binding affinities. It can be used for fine-tuning a language model.
|
12 |
+
|
13 |
+
The data comes from the following sources:
|
14 |
+
- BindingDB
|
15 |
+
- PDBbind-cn
|
16 |
+
- BioLIP
|
17 |
+
- BindingMOAD
|
18 |
+
|
19 |
### Use the already preprocessed data
|
20 |
|
21 |
Load a test/train split using
|
|
|
26 |
validation = load_dataset("jglaser/binding_affinity",split='train[90%:]')
|
27 |
```
|
28 |
|
29 |
+
Optionally, datasets with certain protein sequences removed are available.
|
30 |
+
These can be used to test the predictive power for specific proteins even when
|
31 |
+
these are not part of the training data.
|
32 |
+
|
33 |
+
- `train_no_kras` (no KRAS proteins)
|
34 |
+
|
35 |
**Loading the data manually**
|
36 |
|
37 |
The file `data/all.parquet` contains the preprocessed data. To extract it,
|