Update README.md

README.md

@@ -1,21 +1,62 @@
-## MS DIAL (only lipidome atlas) 
-* Source: [Supplementary information](https://www.nature.com/articles/s41587-020-0531-2#Sec17) of Tsugawa, H., Ikeda, K., Takahashi, M. et al. A lipidome atlas in MS-DIAL 4. Nat Biotechnol 38, 1159–1163 (2020). https://doi.org/10.1038/s41587-020-0531-2
-* Intruments: TOF
-* Complete MS DIAL DB (not only lipids): https://zenodo.org/records/10953284
+## Download
+training_set, test_set, test_set2, can be downloaded from: https://huggingface.co/datasets/Bowen999/lipids_ms2/tree/main
 
+## Data Source
+- **Training Set**:  
+  The training set contains 959,031 entries of lipid MS/MS spectra and corresponding structural information.  
+  It is constructed by integrating data from multiple sources, including **MS DIAL Lipidome Atlas**, **GNPS**, **MassBank**, **MoNA**, **PNNL-Lipids**, **HCE**, and **IOBA_NHC**.
 
-## MassSpecGym:
-* Source: https://huggingface.co/datasets/roman-bushuiev/MassSpecGym
-* Instruments: TOF, Orbitrap
+- **Test Set**:  
+  The test set is sourced from **MassSpecGym**, containing 4,631 MS/MS spectra.  
+  Among them, 1,372 spectra are **novel** (not present in the training set).
 
-## GNPS
-* Source: https://external.gnps2.org/gnpslibrary
+- **Test Set2**
+MS2 data of standards from Li Lab and Nova Medical Testing (non-public data)"
 
-## MassBank
-* Source: https://external.gnps2.org/gnpslibrary
+*Preprocessing for **comprehensive MS2 databases** (like MassSpecGym) involved retaining only lipids by filtering the **InChI Key main block** using references from the MS DIAL Lipidome Atlas and Swiss Lipids*.
+
+
+## Dataset Description
+| Column # | Column Name     | Description |
+|:---------|:----------------|:------------|
+| 1 | **name** | Full lipid name including class and detailed acyl-chain notation (e.g. “PC 18:1/16:0”) |
+| 2 | **simple_name** | Simplified lipid name summarizing total carbons and double bonds (e.g. “PC 34:1”) |
+| 3 | **class** | Lipid class abbreviation (e.g. PC, PE, TAG) |
+| 4 | **chain** | Acyl-chain composition string (e.g. “18:1/16:0”) |
+| 5 | **num_c_1** | Number of carbons in chain 1; **Note: `num_c_1` can be equal to 0, but the `num_c` values for all four chains cannot all be 0 at the same time.**|
+| 3 | **class** | Lipid class abbreviation (e.g. PC, PE, TAG) | 
+| 6 | **num_db_1** | Number of C=C in chain 1|
+| 7 | **extra_1** | Extra modifications on chain 1 (e.g. (2OH)), if not present, then None |
+| 8 | **num_c_2** | Number of carbons in chain 2 |
+| 9 | **num_db_2** | Number of C=C in chain 2 |
+| 10 | **extra_2** | Extra modifications on chain 2 |
+| 11 | **num_c_3** | Number of carbons in chain 3 |
+| 12 | **num_db_3** | Number of C=C in chain 3 |
+| 13 | **extra_3** | Extra modifications on chain 3 |
+| 14 | **num_c_4** | Number of carbons in chain 4 |
+| 15 | **num_db_4** | Number of C=C in chain 4 |
+| 16 | **extra_4** | Extra modifications on chain 4 |
+| 17 | **precursor_mz** | Mass-to-charge ratio (m/z) of the precursor ion |
+| 18 | **adduct** | Adduct ion type detected (e.g. [M+H]+, [M+Na]+) |
+| 19 | **ion_mode** | Ionization mode (Positive or Negative) |
+| 20 | **charge** | Observed charge state of the ion |
+| 21 | **smiles** | SMILES string representation of the molecule |
+| 22 | **inchi** | Full IUPAC InChI identifier |
+| 23 | **inchikey** | InChIKey |
+| 24 | **inchikey_main** | Main layer (first 14 characters) of the InChIKey |
+| 25 | **exact_mass** | Theoretical monoisotopic neutral mass |
+| 26 | **formula** | Chemical formula (e.g. C₃₆H₇₀O₈P) |
+| 27 | **synonyms** | Alternative names |
+| 28 | **retention_time** | Liquid chromatographic retention time |
+| 29 | **lib_quality** | Library match quality score or confidence (1 is best) |
+| 30 | **source** | Origin of the entry (database name) |
+| 31 | **spectrum_id** | Unique identifier for the MS/MS spectrum in source database |
+| 32 | **ms_level** | MS level of the spectrum |
+| 33 | **instrument** | Mass spectrometer used |
+| 34 | **energy** | Collision/fragmentation energy applied (e.g. 40 eV) |
+| 35 | **num_peaks** | Number of peaks recorded in the MS2 spectrum |
+| 36 | **MS2** | fortmat: [[mz1, int1], [mz2, int2]...], raw MS2 spectrum, “raw” refers to data taken directly from the source; however, **the MS2 in that source may already have been normalized ** |
+| 37 | **MS2_norm** | Normalized MS2 intensities, Intensities are scaled to 0–100; if there are more than 100 peaks, only the 100 highest-intensity peaks are retained. |
+| 38 | **novel** | Flag indicating novel lipid entry (y or n); **Only test set have this column**|
 
-## PNNL-LIPIDS
-* Source: https://external.gnps2.org/gnpslibrary
 
-## MONA
-* Source: https://external.gnps2.org/gnpslibrary