ifyoungnet / admetlab Goto Github PK

A platform for systematic ADME evaluation of drug molecules, thereby accelerating the drug discovery process.

License: GNU Affero General Public License v3.0

Python 100.00%

admetlab's Introduction

ADMETlab

A platform for systematic ADME evaluation of drug molecules, thereby accelerating the drug discovery process.
The platform is avaliable at: http://admet.scbdd.com (Note: the pgp-substrate model have been fixed, also the correct model has been updated in webserver)

Features

Well optimized SAR models with better performance than state-of-the-art models.
User-friendly interface and easy to use.
Support batch computation.
Systematically evaluate molecular druggability.
Provide constructive suggestions for molecular optimization.

About this repository.

This repository provides models of ADMETlab in binary files for expert users. The ordinary users are suggested to use the ADMETlab server to accomplish all related prediction or analysis tasks because of the convenience and no programming requirements.

The expert users could download these models to carry out further research. To use the model, download all the zipped files and unzip, then calculate the descriptors according to the "Documentation" section, then load the model to predict. The example folder contains an example. In the command line, input:

python run.py

and you could get the results.

About the algorithms and explanation.

The modeling process including descriptors, methods, feature selection and model performance has been detailedly described in the "Documentation" section of the website.

Contact.

If you have questions or suggestions, please contact: [email protected] and [email protected]

Please see the file LICENSE for details about the "New BSD" license which covers this software and its associated data and documents.

admetlab's People

Contributors

Stargazers

Watchers

admetlab's Issues

Unable to load PPB_Label.pkl

Hi,
I am unable to load the pickle file named PPB_Label.pkl so please can you share the contents of the file. I just need the list of names of the descriptors used to predict the PPB value.

Known compounds on which this tool worked on

Hi,
Actually I want to understand how this ADMETSar works? I have gone through the papers but didnt understand properly. What are the known compounds or datasets you have worked? Can you give me some links or data regarding the inputs? How can we use those inputs for our research? Please let me know the whole scenario.
Thank you

pgp_substrate model

Hi, would like to know if you would append correct pgp-substrate model in github too. It would be very helpful for our research.
Rest of the models are perfect.

Smiles -> features

Hi, could you enable feature generation from smiles strings? This would allow predictions on new molecules and would be quite helpful. Thank you!

Question about the model.

Thanks for your good job.
I have run the example you offered and it works well.
And now, I noticed you offer 35 model files containing classfication and regression models.
I unzip several files and find they are same, no matter in the classfication or regression models.
So, could you tell whether it's right I use random one model in these files to predict admet.
Or I need to run all the files to do predictions and summary the results ?

KeyError: 'max_depth' when use model

Hi, A error occurred when I use the model downloaded from here. I test the example just like python run.py in python2 environment, and raise the KeyError: 'max_depth'! And I have no ideas about this. Thank you!

Do your team have API or python library for many molecules?

I have read your article here. I am interested with Table 3. Comparison of the main features of ADMETlab (version 1.0–2.0) with other web-based tools. I want to use your tool to my project. In your website, Home/ Services/ADMET Evaluation, it is restricted to apply only a molecule, but i have a dataset of many molecules in *.csv file. Could you make API for everyone access or python library?

Which Python 2 version, skikit-learn, pickle version?

Can someone post all python package versions? I am having issues with unpickling the data, I suspect new/older version incompatibility. It would be nice if these could be pickled into python 3 format, python 2 starts to loose support eventually.

Reproducibility Issues, documentation issues

Hello,

I am going through and attempting to reproduce everything. Here are just some things I noticed.

caco2 model has length of input 27 not length 30 as descriptors docs say
What is Getov? Geometric topological index by Narumi is this the same as Geto? and if they are the same why is logD being trained on two desciptors that are the same?
for T1/2, the docs say there are 40 descriptors but their is actually 50.
LD50 has no descriptors labels on the docs website (but it is in the LABEL folder on github).
This one is just a slight annoyance, but some of the topological descriptors are not in PyBioMed but instead in chemopy (most of them are in PyBioMed however).