Deep-Learning-Based Multivariate Curve Resolution (DeepResolution) method has been proposed for automatic resolution of GC-MS data. It has excellent performance in resolving overlapping peaks and is suitable for large-scale data analysis. Compared with the classical multi-curve resolution method, it has the characteristics of fast, accurate, scalable and fully automatic.
Python 3.5.2，available at https://www.python.org.
TensorFlow (version 1.14.0-GPU)，available at https://github.com/tensorflow.
The packages mainly include: numpy,Scipy,Matplotlib,pandas,sklearn,csv and os.
These packages are included in the integration tool Anaconda (https://www.anaconda.com).
Since the model exceeded the limit, we have uploaded all the models and some example data to the google drive.
Run the file 'component_identification.py'.
The corresponding example data have been uploaded to the example data folder（in google drive）named 'data_1.npy' and 'labels_1.npy'.These are augmented data for a component.
Run the file 'DeepResolution.py'.
Example data have been uploaded to the data folder named 'zhi10-5vs1.CDF'. The file named 'component.csv' is the components's name of all our CNN models. Download the model and these example data，DeepCID can be reload and predict easily.
More example data can be gotten form google drive.
Xiaqiong Fan: xiaqiongfan@csu.edu.cn