## 📊 ReTiNA Column Descriptors

Each data entry in the ReTiNA-1 dataset is comprised of 7 columns. Examples and descriptions are detailed below. Note that all dictionary, list, and tuple representations are encoded as strings in the retina1_dataset.csv file, and will need to be converted via ```ast.literal_eval()``` for processing.

| Column Index | Column Descriptor | Example | Data Type | Description |
|--------------|-------------------|---------|-----------|-------------|
| 0 | compound | CSCCCCCCCN=C=S | ```str``` | The eluted compound in question, represented as a SMILES string. |
| 1 | solvents | {'A': [{'CC#N': 100}, {'C(=O)O': 0.0133}], 'B': [{'O': 100}, {'C(=O)O': 0.0133}]} | ```dict``` | A dictionary representation of solvent fronts A & B, with the solvents and their proportions out of 100 accompanied by pH contributing additives and their molar concentrations in M. |
| 2 | gradient | [(0, 99.9), (2, 91.0), (7, 87.5), (10, 70), (12, 50), (13, 50), (15, 0.1)] | ```list``` | The solvent gradient used, with the 1st element of each internal tuple representing the time in seconds, and the 2nd element of each tuple representing the % of solvent front B. |
| 3 | column | ('RP', 2.1, 50, 1.7) | ```tuple``` | Details on the LC-MS column used, with the 1st element representing the column type - HILIC ('HI') vs. reverse phase ('RP'). The 2nd and 3rd elements represent the internal column diameter and column length, in mm. The third element represents the stationary phase particle width, in µm. |
| 4 | flow_rate | 0.25 | ```float``` | The solvent front flow rate, measured in mL/min. |
| 5 | temp | 40.0 | ```float``` | The column temperature, measured in degrees Celsius. |
| 6 | source | massbank | ```str``` | The source the data was extracted from. |

## 📋 Data Sources Used

[DynaStl](https://doi.org/10.3390/metabo9050085) (dynastl): A Dynamic Retention Time Database for Steroidomics

- 188 compound-environment combinations
- 188 unique compounds
- 1 unique LC-MS setup environment


[EnamineRT](https://chemrxiv.org/engage/chemrxiv/article-details/67c96c1981d2151a024cf390) (enaminert): Machine Learning-Based Retention Time Prediction Tool for Routine LC-MS Data Analysis

- 20,000 compound-environment combinations
- 20,000 unique compounds
- 1 unique LC-MS setup environment

[MassBank](https://github.com/MassBank/MassBank-data/tree/main) (massbank)

- Data sourced from 5 individual MassBank datasets
- 13,597 compound-environment combinations
- 2,177 unique compounds
- 5 unique LC-MS setup environments

[MCMRT](https://doi.org/10.1038/s42004-024-01135-0) (mcmrt): Generic and Accurate Prediction of Retention Times in Liquid Chromatography by Post–Projection Calibration

- 10,073 compound-environment combinations
- 342 unique compounds
- 30 unique LC-MS setup environments

[METLIN](https://doi.org/10.1038/s41467-019-13680-7) (metlin): A Small Molecule Dataset for Machine Learning-Based Retention Time Prediction

- 80,038 compound-environment combinations
- 80,038 unique compounds
- 1 unique LC-MS setup environment

[AgResearchDB](https://doi.org/10.1016/j.chroma.2011.07.105) (nz): Predicting Retention Time in Hydrophilic Interaction Liquid Chromatography Mass Spectrometry and its use for Peak Annotation in Metabolomics

- 116 compound-environment combinations
- 116 unique compounds
- 1 unique LC-MS setup environment

[RepoRT](https://doi.org/10.1038/s41592-023-02143-z) (report): A Comprehensive Repository for Small Molecule Retention Times

- 7,640 compound-environment combinations
- 2,217 unique compounds
- 185 unique LC-MS setup environments

[ReTip](https://pubs.acs.org/doi/10.1021/acs.analchem.9b05765) (retip): Retention Time Prediction for Compound Annotation in Untargeted Metabolomics

- 2,252 compound-environment combinations
- 2,178 unique compounds
- 2 unique LC-MS setup environments