chen1028's picture
Update README.md
536a0fe verified
---
language:
- zxx
license: cc-by-4.0
tags:
- chemistry
- mass-spectrometry
- isotopic-patterns
- tabular
- chlorine
pretty_name: Cl-Containing Compound (MS1 Features)
size_categories:
- 100K<n<1M
source_datasets:
- pubchem
task_categories:
- tabular-classification
configs:
- config_name: default
data_files:
- split: train
path: 80%_618272_train_binary.rds
- split: test
path: 20%_154568_test_binary.rds
- config_name: csv
data_files:
- split: train
path: train.csv
- split: test
path: test.csv
dataset_info:
features:
- name: mz0
dtype: float32
- name: int2_o_int0
dtype: float32
- name: int1_o_int0
dtype: float32
- name: RI2_RI1
dtype: float32
- name: mz_2_0
dtype: float32
- name: mz_1_0
dtype: float32
- name: has_cl
dtype: int8
config_name: default
---
# Dataset Summary
Binary classification of chlorine presence using simulated MS1 isotopic patterns. Inputs are six engineered features; label has_cl indicates Cl-containing (1) vs non-Cl (0).
## Data Sources and Generation
- Simulated MS1 peaks (M, M+1, M+2, M+3, M+4) for PubChem molecular formulas.
- Counts: 968,442 non-Cl; 386,420 Cl (1Cl: 185,303; 2Cl: 117,566; 3–5Cl: 83,551).
- Features: mz0, int2_o_int0, int1_o_int0, RI2_RI1, mz_2_0, mz_1_0.
- Class balancing: downsampled non-Cl to 386,420 (total 772,840).
- Splits: 80% train (618,272), 20% test (154,568).
- Files: 968442_non_cl_filter_S10.rds, 386420_cl_data.rds, 80%_618272_train_binary.rds, 20%_154568_test_binary.rds.
## How train.csv and test.csv were created
- Source splits come from the RDS files above: 80%_618272_train_binary.rds (train) and 20%_154568_test_binary.rds (test).
- Conversion was done locally using Python with pyreadr and pandas (see `data/converter.ipynb`).
- Steps:
1. Read each .rds table using pyreadr.read_r(...)
2. Optionally cast numeric columns to float32/int32 for compact CSVs
3. Save to CSV with index=False as `train.csv` and `test.csv`
Example code used:
```python
import pyreadr, pandas as pd
train_df = next(iter(pyreadr.read_r("80%_618272_train_binary.rds").values()))
test_df = next(iter(pyreadr.read_r("20%_154568_test_binary.rds").values()))
train_df.to_csv("train.csv", index=False)
test_df.to_csv("test.csv", index=False)
```
## Features
- mz0: m/z of M
- int2_o_int0: intensity ratio M+2/M
- int1_o_int0: intensity ratio M+1/M
- RI2_RI1: (M+2/M) − (M+1/M)
- mz_2_0: m/z(M+2) − m/z(M)
- mz_1_0: m/z(M+1) − m/z(M)
- has_cl: target label (0/1)
## Usage
```python
from datasets import load_dataset
# Load CSVs from the Hub using the CSV builder
ds = load_dataset(
"csv",
data_files={
"train": "hf://datasets/chen1028/Cl-Containing-Compound/train.csv",
"test": "hf://datasets/chen1028/Cl-Containing-Compound/test.csv",
}
)
```
## Citation
```bibtex
@article{doi:10.1021/acs.analchem.3c05124,
author = {Zhao, Tingting and Wawryk, Nicholas J. P. and Xing, Shipei and Low, Brian and Li, Gigi and Yu, Huaxu and Wang, Yukai and Shen, Qiming and Li, Xing-Fang and Huan, Tao},
title = {ChloroDBPFinder: Machine Learning-Guided Recognition of Chlorinated Disinfection Byproducts from Nontargeted LC-HRMS Analysis},
journal = {Analytical Chemistry},
volume = {96},
number = {6},
pages = {2590-2598},
year = {2024},
doi = {10.1021/acs.analchem.3c05124},
note = {PMID: 38294426},
url = {https://doi.org/10.1021/acs.analchem.3c05124},
```