Giter Club home page Giter Club logo

gnn_dti's Introduction

Screenshot

Main Results

Screenshot

https://pubs.acs.org/doi/10.1021/acs.jcim.9b00387

Training command example

python -u train.py --dropout_rate=0.3 --epoch=1000 --ngpu=1 --batch_size=256 --num_workers=0

We added only about data of 1000 samples in the data folder due to the size of the dataset so the performance is much lower than the paper. Each sample is saved in a pickle file and it consists of two rdkit objects of a ligand and protein. The inputs of the neural network are processed in the dataset class on the fly.

gnn_dti's People

Contributors

jaechanglim avatar seonghwanseo avatar

Stargazers

Lilian Denzler avatar  avatar  avatar  avatar Yoshitaka Inoue avatar Marine avatar Yan avatar  avatar Gary Tom avatar Rui P. Ribeiro avatar seeun avatar Shuting xu avatar Cheng Ge avatar Deepanshu Mody avatar Ohmura Takuto avatar  avatar Grace Hu avatar Gökhan Özsari avatar  avatar Haifeng_Liu avatar Qijin Yin avatar Seyong Kim avatar Jihun Jeung avatar  avatar sun avatar mortlock_19 avatar  avatar  avatar Lee Joongwon avatar Xiaoyong Zhu avatar jin avatar Ma Chang avatar Wei Lu (陆威) avatar  avatar  avatar dxyzx avatar Zhang Bing avatar  avatar Geunho Choi avatar Sooheon Kim avatar  avatar  avatar gefrorenemango avatar Hosung Kim avatar Rodrigo avatar Darren An avatar DovahCode avatar  avatar pan xiaolin avatar Shuangjia Zheng avatar Jiezhong Qiu avatar DingShizhe avatar  avatar Roxana Noelia Villafañe avatar Derek Jones avatar Jianing Lu avatar  avatar Cinkate Ren avatar Rocco Meli avatar Hongchao Ji avatar  avatar tbwxmu avatar Xinhao Li avatar Ho Leung Ng avatar

Watchers

 avatar

gnn_dti's Issues

About data folder

I still don't know how to create data folder after reading the previous questions,and i want to know how the files in the folder are generated and what the format is.
could you answer my question?
Many thanks!!!

Can this model be applied to predicting affinity value?

Hi,
Thanks for your solid work!

I noticed that this model can predict whether the input protein-ligand is active or not, which belongs to a binary classification task. However, I wonder GNN-DTI can be used for regression tasks? Can this model predict the binding affinity values well?

Thanks

About Dataset

how to prepare dataset (how to use Smina to deal with DUD-E dataset ) ?
could you show me an example?
many thanks!

안녕하세요. 궁금한 부분이 있어 질문 남깁니다.

안녕하세요.
Deep learning 과 cheminformatics 에 대해 공부하고 있는 학생입니다.
좋은 논문 내주셔서 감사합니다. 잘 읽었습니다.

한편, 궁금한 점이 생겨 질문드립니다.
코드와 논문을 통해 보았을 때, 모델에서는 Fully Connected layer에 들어가기 전에,
filter-wise(feature-wise)로 summation하여 그 값을 기반으로 FC layer를 거쳐 classification하는 것으로 보입니다.
(제가 틀렸다면 죄송합니다. 알려주시면 감사하겠습니다.)

한편, node-wise(atom-wise)하게 summation하여 FC층을 거치는 부분도 고려하셨을 수 있었을 것 같은데,
hidden feature를 기준으로 flatten한 이유가 있으신지 궁금합니다.
(물론 node-wise하게 summation하였을 경우에, pocket과 ligand의 atom의 수가 pair마다 달라
flatten size가 계속 달라지는 issue가 있을 것 같기는 합니다만,, )

혹은 두 부분 모두 실험적인 부분을 통해 이와 같은 모델을 사용하신 것인가요?
저자님의 견해가 궁금해 질문 남겼습니다..!
감사합니다!

안녕하세요. 논문과 코드를 흥미롭게 잘 읽었습니다.

안녕하십니까.

신약개발과 관련해서 열심히 공부중인 사람입니다.

작성하신 논문과 코드를 아주 흥미롭게 잘 읽었습니다. 정말 참신한 아이디어와 모델인듯합니다.

논문과 코드를 읽다보니 조금 궁금한 점이 있어서 이렇게 연락드리게 되었습니다.

data 폴더의 파일들이 확장자가 없는 pkl파일들이던데 이 데이터를 어떻게 만드신지 궁금합니다.

PDB나 Dud-E 데이터를 전처리하여 이 파일을 만드는 코드가 혹시 올려놓으신 파일들에 포함되어 있는지 궁금합니다. 만약 코드가 올려놓으신 데이터에 포함되어 있다면 위치를 좀 알려주실 수 있으신가요? 혹시 없다면, 어떤 데이터로 어떻게 전처리하신지 좀 알려주실 수 있으신지요?

논문 정말 감사히 잘 읽었습니다.

답변 기다립니다.

"data" File Preparation

Hello!
The preprocessing code cannot be disclosed, could you please disclose some sample files (pickle files), maybe I want to standardize the file name through the sample, or even correct my data preparation.
As for the pickle file formation method, is it filtered by SMINA (8 angstrom) and processed separately by RDKit for the target ligand to form a pickle file?
Thanks!

generation of pickle file

Hi,
where is the script that generates the pickle file? It is interesting to know cause it will affect much the size of the graph. I can see that the protein atoms in your files have small number of atoms so what was the cutoff used to get those atoms and are they selected from native pose only or differ according to the docked pose?

model fine tuning

hi ,i am very interesting about your code and paper ,could you share your hyper-parameters ?
thanks!

Fully trained model

Do you share the fully trained model somewhere without having to retrain?

data folder

Is the Data Folder already gone? Can I upload it again?

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.