MS-feature

A deep learning-based feature detection tool for feature detection in Liquid Chromotography-Mass Spectrometry (LC-MS). Details available in Deep Learning Based MS2 Feature Detection for Data-Independent Shotgun Proteomics. The folder for Faster-RCNN framework is forked from here.

Dependencies

• python==3.8
• torch==1.11.0
• torchvision==0.12.0
• opencv-python==4.6.0.66
• jupyter==1.0.0
• numpy==1.22.4
• tqdm==4.64.0

Run

• Put source .mzml files into the data_prep/src folder
• Run mzml_to_img.ipynb jupyter notebook
• Adjust the height and width parameters in the first two lines generate_windows.ipynb for the sizes of the sliding windows. The default of 240x270 has been applied as stated in the paper.
• Run the first 4 cells of generate_windows.ipynb. You should expect the windows in the folder named img_(height)_(width).
• Execute predict.py, changing line 11 to the folder generated from the last step.