mars-wei/mssl

MSSL2drug

We peopose multi-task joint strategies of self-supervised representation learning on biomedical networks for drug discovery, named MSSL2drug. We design six basic SSL tasks that are inspired by various modality features including structures, semantics, and attributes in biomedical heterogeneous networks. In addition, fifteen combinations of multiple tasks are evaluated by a graph attention-based adversarial multi-task learning framework in two drug discovery scenarios. The results suggest two important findings. (1) Combinations of multimodal tasks achieve the best performance compared to other multi-task joint strategies. (2) The joint training of local and global SSL tasks yields higher performance than random task combinations.

Data description

1. BioHNsdata: biomedical heterogeneous networks

• drug.txt: list of drug.
• protein.txt: list of protein.
• disease.txt: list of disease.
• drug_sim.txt: drug similarity matrix
• protein_sim.txt: protein similarity matrix
• disease_sim.txt: protein similarity matrix
• BioHNs.txt: the adjacency matrix of the biomedical heterogeneous networks.
• BioHNs_code.txt: the initialization features of nodes in the biomedical heterogeneous network.
• BioHNsdata_with order: the corresponding order of node in drug_sim.txt, protein_sim.txt, disease_sim.txt, and BioHNs.txt BioHNs.txt, and BioHNs_code.txt

2. DownStreamdata: drug discovery data

• DDINet.txt: Drug-Drug interaction matrix.
• DTINet.txt: Drug-Protein interaction matrix.

3. SSLdata: Self-supervised representation pretext

• ClusterPre.txt, PairDistance.txt, EdgeMask.txt, PathClass.txt, SimReg.txt and SimCon.txt: a example dataset for six self-supervised task.
• SSL_data_description.txt : The detailed description of PairDistance.txt, EdgeMask.txt, PathClass.txt, SimReg.txt and SimCon.txt

Single task-driven self-supervised representation learning

Graph attention networks (GATs) are used to train the single task-driven self-supervised representation learning. The code of GAT can be downloaded from https://github.com/Diego999/pyGAT.

Multi-task self-supervised representation learning for drug discovey

models/ contains the different implementations of graph attention-based adversarial multi-task learning models for three different training paradigm, including classification task, regression task and similarity constrast task. In MSSL2drug, we develop fifteen combinations of multiple tasks. However, there are same implementations among some multi-task combinations. Therefore, we take PairDistance_EdgeMask_SimCon (PES) as an examples:

1. Warm-start scenarios: bash PES_warm.sh

2. Cold-start scenarios: bash PES_cold.sh
--offset: downstream network
--scenario: warm-start or cold-start prediction
--dataclean: whether to remove the test data from SSL

Requirements

MSSL2drug is tested to work under:

• Python 3.7.7
• numpy 1.16.1
• torch 1.6.0

Contacts

If you have any questions or comments, please feel free to email: [email protected] or [email protected].