PHISGAE

Code and Datasets for "Host specificity-guided graph autoencoder for predicting phage-host interaction"

Developers

Zhen Xiao ([email protected]) and Xingpeng Jiang ([email protected]) from School of Computer Science, Central China Normal University.

Datasets

• data/PHI_all.csv is the dataset with known phage-host pairs between bacteriophages and hosts from different taxa.
• data/phage_ncbi.txt is used to download the gene sequence and protein sequence information of all bacteriophages (fasta file and gb file).
• data/DNA_f is a set of files with features encoded by DNA sequences corresponding to all phages.
• data/Protein_f is a set of files with features encoded by protein sequences corresponding to all phages.
• data/dnapro_f is a set of files with features encoded by DNA sequences and protein sequences corresponding to all phages.
• data/phage_host.csv is a dataset representing the interaction matrix between bacteriophages and hosts.
• data/phage_sim.csv is a dataset representing the refined phage-phage connection network.
• data/host_I.csv is is a dataset representing the host-host connection network.

Environment Requirement

The code has been tested running under Python 3.7 and R 4.2.0. The required packages are as follows:

• numpy == 1.21.5
• pandas == 1.3.5
• tensorflow == 1.15.0
• keras == 2.3.1
• torch == 1.4.0+cpu
• ExPASy

Usage

Users can use their own data to train prediction models. If you want to train PHISGAE with your own dataset, you need to go through the following two processes: 1 Prepare the dataset and 2 Predicting phage-host interactions. In the meantime, we have provided the data that has been processed and you can implement PHISGAE directly using procedure: 2 Predicting phage-host interactions.

1 Prepare the dataset

git clone https://github.com/JennyX212/PHISGAE
cd PHISGAE/code
python dna_f.py   
python pro_f.py   ####compute features derived from DNA and protein sequences

1. Download all phage DNA and protein sequences from NCBI batchentrez, named phage.fasta and phage.gb into the data folder.
2. Run python dna_f.py and python pro_f.py to compute the features derived from DNA and protein sequences. Note If your protein sequences in one .gb file, you need to run python protein_processing.py to split them into different .txt files. For phages missing protein sequence information, you can use the ExPASy tool to translate their gene sequence into protein sequence.
3. Use python sim.py and module ## Other similarity calculation method ## in data_processing.R to calculate the different similarity matrix between phages with data/dna_pro_f/phage_dna_pro.csv.
4. Use module KNN graph in data_processing.R to obtain refined phage-phage connection network.
5. Use python hostI.py to generate a represention of host-host connection network.
6. Convert the known phage-host pairs into different interaction matrices by running python RAWA.py

2 Predicting phage-host interactions

• tensorflow == 1.15.0

git clone https://github.com/JennyX212/PHISGAE
cd PHISGAE/code
python main.py  --input = "input_path" --output = "output_path"	 #### train model and get the prediction result

You just need to run python main.py and type input =../data/train/species/ to train PHISGAE and get the result of PHISGAE for predicting phage-host interactions at species level. You can save the result by typing output ="output_path", where "output_path" is the file path of the outcome. Besides, if you want the prediction results at different taxonomic level, you can directly modify the input path when typing input = and change "species" to the other taxonomic level.

If you want to run PHISGAE with your own dataset, simply enter the path to the data set in the input prompt.

Contact

Please feel free to contact us if you need raw data or any other help.