/MVCMDA

Primary LanguageJupyter NotebookGNU General Public License v3.0GPL-3.0

MVCMDA

用于miRNA-药物关联预测的多通道多层次通道注意力图表示学习模型

Contribution

  1. construct miRNA-Drug association network
  2. construct miRNA and Drug multi-view properties graph
    • miRNA: sequence and function
    • Drug: sequence and structure
  3. design association network update strategy based neighborhood info
  4. design multi-view multi-channel attention graph neural network

TODO

  • multi-view
  • neural architechure search
  • constractive learning

Requirement

  • python 3.8.16
  • torch-cluster 1.6.1+pt20cu117
  • torch-geometric 2.3.0
  • torch-scatter 2.1.1+pt20cu117
  • torch-sparse 0.6.17+pt20cu117
  • torch-spline-conv 1.2.2+pt20cu117
  • pytorch 2.0.0
  • numpy 1.21.2
  • scipy 1.9.1
  • tensorflow 2.2.0
  • networkx 2.8
  • rdkit 2022.9.5
  • GraphEmbedding
  • Mol2vec
  • gensim 4.1.2
  • biopython 1.79
  • tqdm 4.65.0

File structure

-- ./datasets/ : oringinal data
  -- ./datasets/multiview_dataset_updatamd.pt : multi-view dataset after update associate network
-- ./MatrixFactorization/ : code of MF model
  -- ./MatrixFactorization/NMF.py : code of MF model
  -- ./MatrixFactorization/NMFwithNIF.py : code of MF using neighborhood info model

-- ./MultiView/ : multi-view network construction
  -- ./MultiView/Mol2Vec.ipynb : get drug mol2vec embedding
  -- ./MultiView/rna.ipynb: get miRNA seq simlarity
  -- ./MultiView/multiview_dataset.ipynb : get multi-view dataset in pt format 

-- ./NIMCcode/ :code of GNNs model and training
  -- ./NIMCcode/main_cv_multiview.py : train code of multiview GNNs model
  -- ./NIMCcode/model_attenGNN_multiview.py : model code of multiview GNNs model

-- ./Embedding/ : graph struc embedding method
  -- ./embedding_cv_element.ipynb :code of struc learning, element--wise 
  -- ./embedding_cv_row_col.ipynb :code of struc learning, row--col--wise

-- ./Tools/ : some tools
  -- ./embedding.ipynb :code of struc learning demo
  -- ./datacheck.ipynb :check dataset
  -- ./metric.ipynb :visualization of metric & result
  -- ./metric_union.ipynb :visualization of all prediction

Device

Device 1

  • GPU: Nvidia GTX1080ti * 2
  • CPU: Intel(R) Xeon(R) Gold 6146 CPU @ 3.20GHz
  • RAM: 128G

Device 2

  • GPU: NVIDIA Quadro RTX 5000
  • CPU: 15 vCPU AMD EPYC 7543 32-Core Processor
  • RAM: 30GB

Info

  • Author: Yi ZhengYe From HZAU
  • Date: 2023-03 ~ 2023-06
  • Modified: 2023-05-26