Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -18,10 +18,12 @@ Converting irregularly spaced mass spectrometry (MS/MS) data into computable rep
|
|
| 18 |
|
| 19 |
## Dataset
|
| 20 |
|
|
|
|
|
|
|
| 21 |
- `inference_data/Massive-KB-v1_test_sorted.h5`: 107,713 held-out spectra.
|
| 22 |
- `inference_data/PXD017407_library_from_fdr_peptides_sorted.h5`: 2,301 additional spectra.
|
| 23 |
|
| 24 |
-
|
| 25 |
|
| 26 |
## Download
|
| 27 |
|
|
|
|
| 18 |
|
| 19 |
## Dataset
|
| 20 |
|
| 21 |
+
- `training_data/`: 30 MassiveKB H5 files, 1,938,843 spectra.
|
| 22 |
+
- `val_data/Massive-KB-v1_val.h5`: 107,713 validation spectra.
|
| 23 |
- `inference_data/Massive-KB-v1_test_sorted.h5`: 107,713 held-out spectra.
|
| 24 |
- `inference_data/PXD017407_library_from_fdr_peptides_sorted.h5`: 2,301 additional spectra.
|
| 25 |
|
| 26 |
+
Each H5 file contains `peaks` (`[N, 200, 2]` m/z and intensity), `sequence`, `charge`, `precursor_mass` (precursor m/z), and `protein`. Peptide sequences can include the lowercase modification codes described in the code repository.
|
| 27 |
|
| 28 |
## Download
|
| 29 |
|