Update README.md
Browse files
README.md
CHANGED
|
@@ -3,8 +3,7 @@ license: cc-by-4.0
|
|
| 3 |
---
|
| 4 |
# Synthetic mass spectra dataset No. 1
|
| 5 |
This is the first variant of the GC-EI-MS synthetic dataset generated by [NEIMS] and [RASSP] models from 4.8 million structures downloaded from [ZINC 20].
|
| 6 |
-
The overlap with version 2 is approximately 8k structures (negligible), so it's usable as an extension to synth2.
|
| 7 |
-
and synth2, see section Dataset choice.
|
| 8 |
|
| 9 |
|
| 10 |
## Spectra generators
|
|
@@ -41,13 +40,13 @@ neims_custom_gen/test.jsonl 242485
|
|
| 41 |
TOTAL 4849716
|
| 42 |
```
|
| 43 |
|
| 44 |
-
|
| 45 |
## Dataset choice
|
| 46 |
If you're choosing between synth1 and synth2 (and don't need both), choose synth2. The origin of synth2 is better documented in the 3_datapreprocessing.ipynb notebook in our [GitHub repository],
|
| 47 |
even though the creation process is in essence the same for both datasets. There are the same compound in the splits of both RASSP and NEIMS generated datasets but some
|
| 48 |
duplicities emerged, so the sets are not identical (!the scale of the leak is less than 10 compounds in over 2M set, so we call it negligible!)
|
| 49 |
|
| 50 |
-
|
| 51 |
## Data format
|
| 52 |
|
| 53 |
Each line of every file is a `json` comprising three items:
|
|
|
|
| 3 |
---
|
| 4 |
# Synthetic mass spectra dataset No. 1
|
| 5 |
This is the first variant of the GC-EI-MS synthetic dataset generated by [NEIMS] and [RASSP] models from 4.8 million structures downloaded from [ZINC 20].
|
| 6 |
+
The overlap with version 2 is approximately 8k structures (negligible), so it's usable as an extension to synth2.
|
|
|
|
| 7 |
|
| 8 |
|
| 9 |
## Spectra generators
|
|
|
|
| 40 |
TOTAL 4849716
|
| 41 |
```
|
| 42 |
|
| 43 |
+
<!--
|
| 44 |
## Dataset choice
|
| 45 |
If you're choosing between synth1 and synth2 (and don't need both), choose synth2. The origin of synth2 is better documented in the 3_datapreprocessing.ipynb notebook in our [GitHub repository],
|
| 46 |
even though the creation process is in essence the same for both datasets. There are the same compound in the splits of both RASSP and NEIMS generated datasets but some
|
| 47 |
duplicities emerged, so the sets are not identical (!the scale of the leak is less than 10 compounds in over 2M set, so we call it negligible!)
|
| 48 |
|
| 49 |
+
-->
|
| 50 |
## Data format
|
| 51 |
|
| 52 |
Each line of every file is a `json` comprising three items:
|