wzxxxx / mga

MGA

Multi-task Graph Attention (MGA) framework

Multi-task Graph Attention (MGA) is a novel framework to make fully use of available toxicity data, which can simultaneously learn the regression and classification tasks for toxicity prediction. Here is the overview of the Multi-task Graph Attention framework.

requirements：
python 3.6
anaconda
dgl 0.4.3
xgboost
rdkit
pytorch
sklearn

mga's Issues

what yout version of rdkit?

I run drawer = rdMolDraw2D.Molraw2DSVG(280, 280) of the weight_visualization.py and return AttributeError: module 'rdkit.Chem.Draw.rdMolDraw2D' has no attribute 'Molraw2DSVG'.
I also search this .Molraw2DSVG in http://www.rdkit.org/docs/source/rdkit.Chem.Draw.rdMolDraw2D.html, there still didn't have this attribute, i guess maybe you have old version in your code?

Cannot run XGBOOST_xxx.py since data missing.

I can run RGCN.py and Transfer_RGCN.py without problem, but for all files starting with "XGBOOST", I cannot find the training data. Is there something I missed?

Pretraining data used for paper

What is the pretraining data used for the ADMETLab 2.0 paper? Is it the toxicity data that's provided here in this github repo?

Cannot run Toxicity_MGA.py since file missing

I tried to run Toxicity_MGA.py, but failed with missing toxicity_group.csv and toxicity.bin, could you please provide these files?