The original bioRxiv version link: [https://www.biorxiv.org/content/10.1101/2023.03.20.533432v1.abstract]
The accepted version link: [https://academic.oup.com/bioinformaticsadvances/article/3/1/vbad066/7178008]
How to use it:
Our Test Environment (Windows or Linux, the higher environment version beyond this might be also feasible):
- Python 3.6.13
- Pytorch 1.10.2
- CUDA tool kit 11.3.1
- geoopt 0.4.1
Step1. Unfold the file ./DDT_triple_prediction/DDT/data/Cross_validation_split.rar, to obtain the required model input for each independent repeat (based on DTINet).
Step2. Follow the instruction in each following .py file to execute different variants described in the manuscript on each independent repeat:
- Eucli_ECFP6_seqsimilarity.py: for Euclidean based variants using domain knowledge.
- Eucli_noECFP6_noseqsimilarity.py: for Euclidean based variants without using domain knowledge.
- Eucli_visualization.py: visualize the spatial layout of target embeddings under given drug and disease based on Euclidean KGC methods.
- Hypo_ECFP6_noontology_noppi.py: for hyperbolic based variants using ECFP6 + self-contained target embedding look-up table.
- Hypo_ECFP6_ontology_ppi.py: for hyperbolic based variants using ECFP6 + GO/PPI graph.
- Hypo_ECFP6_ontology_ppi_detached1.py: for pre-training hyperbolic target embeddings using GO/PPI graph.
- Hypo_ECFP6_ontology_ppi_detached2.py: mainly for hyperbolic based variants using ECFP6 + pre-trained target embeddings from detached1.py or using ECFP6 + target sequence similarity.
- Hypo_noECFP6_noontology_noppi.py: for hyperbolic based variants using self-contained drug and target embedding look-up tables.
- Hypo_noECFP6_ontology_ppi.py: for hyperbolic based variants using self-contained drug embedding look-up table + GO/PPI graph.
- Hypo_visualization.py: visualize the spatial layout of target embeddings under given drug and disease based on hyperbolic KGC methods.
Check visualization results in the manuscript: files in ./DDT_triple_prediction/DDT/outputs:
- hypo_coordinate_set_original_fold4_515_449.pickle and hypo_coordinate_set_transformed_fold4_515_449.pickle: 2-D coordinates of the hyperbolic spatial layout in the illustration (A) and (B) of the manuscript (based on Hypo_visualization.py).
- eucli_coordinate_set_original_fold4_515_449.pickle and eucli_coordinate_set_transformed_fold4_515_449.pickle: 2-D coordinates of the Euclidean spatial layout in the illustration (C) and (D) of the manuscript (based on Eucli_visualization.py).