Virus Host Predict

This repository contains code and instructions to get the data to predict host taxonomic information from viral genomes.

Data

Data was downloaded from https://www.genome.jp/virushostdb/ on the 25_1_2019.

1. Download the following two files:
   a. virushostdb.genomic.fna.gz -> data/fna
   b. virushostdb.cds.faa.gz -> data/fna

2. Run the unpackfastafiles notebook to unpack the files.

3. Unzip the pfs.zp into the data/pfs folder.

The data directory contains 3 sub direcories.
a. Nucleic acid. Refseq(s) for each virus > data/fna
b. Amino acid sequence for the CDS regions for each refseq > data/faa
c. Pfam domains - extract from the above aa sequences with HMMER for each refseq> data/pfs

Inputs

CSV files containing the information for each dataset run with the corresponding results are in the results folder.
Virushostdb.tsv > inputs/VHDB_25_1_2019.csv

Code

Contains the notebooks used to train and test the different datasets.
1.TrainTest.ipynb notebook contains the best annotations. (select which datasets to run)
Each notebook generates one csv file in the results folder.
2. holdout_bact - trains and test the holdout classifiers for selected bacteria datasets.
3. Kernel_combining notebook - shows an example of how combining the kernels from different genome representations can improve prediction.

class files

1. vhdb.py - load virus host data base table as an object.
2. datasets.py - class containing dataset information.
3. featureset.py - class containing information for each featureset for the dataset object.