Fenglei104 / DeepPROTACs

GNU General Public License v3.0
46 stars 16 forks source link

Accuracy of predicting the case files in the protacs folder #8

Open bwang-ecnu opened 1 year ago

bwang-ecnu commented 1 year ago

Great job! I want to try to use the three cases in the protacs folder for data preparation and prediction. I uses the protas_example.csv file as the 'protacs.csv' file. I set up a new data file using prepare_data.ipynb. But when I used case_study.ipynb and test.pb for predicting, I found that the prediction was the opposite of label([name],[predicted],[label]): ['1_BRD7_VHL'] [0] [1] ['1_BRD9_VHL'] [0] [1] ['2_BRD7_VHL'] [0] [1], Maybe I made a mistake in the process of preparing the data. I would like to ask how accurate is the model to predict these three structures?

Fenglei104 commented 1 year ago

Hi, thank you for your feedback! I checked the results just now and got the predicted output of all "[0]". (I also modified the code slightly, just for convenience). Maybe you can check the running process of openbabel. For convenience, I insert the shell script to the prepare_data.ipynb, which may lead to the invisible running process. You can check it in the command line for verifying. Another which may influence the results is the version of openbabel. The version that I use is 2.3.2.

bwang-ecnu commented 1 year ago

Thank you for your reply. According to your prompt, I have checked the reason: I used openbabel3.1 before, so ‘obabel’ command should be used instead of ‘babel’. But whether I used openbabel3.1 or reinstalled openbabel2.4.0(I didn't find the source package for 2.3.2), the predicted value for all three structures was ["1"]. Maybe I have a problem when I process the data. Could you please provide a complete script for testing these three structures (from preparing the data to predicting the results)?

Fenglei104 commented 1 year ago

The openbabel is installed with the command sudo apt-get install openbabel by default in ubuntu 18.04. And here are the steps:

  1. git clone the project
  2. change the name of protacs_example.csv to protacs.csv
  3. run the prepare_data.ipynb
  4. change the root values in case_study.ipynb and run it.
bwang-ecnu commented 1 year ago

Thank you for your reply, I have reinstalled openbabel2.3.2 according to your suggestion, but I still cannot reproduce your result according to the above process. Could you please provide the intermediate files generated by running this case for comparison?

Fenglei104 commented 1 year ago

You can download from https://cowtransfer.com/s/60d7d34229e14f.

bwang-ecnu commented 1 year ago

Thank you!