Phirbo

A tool to predict prokaryotic hosts for phage (meta)genomic sequences. The tool uses information on sequence similarity between phage and bacteria genomes as well as sequence relatedness among bacteria.

Update

See our latest tool for phage-host prediction: PHIST (https://github.com/refresh-bio/PHIST).

Citation

Zielezinski A, Barylski J, Karlowski WM. "Taxonomy-aware, sequence similarity ranking reliably predicts phage–host relationships" BMC Biology, Volume 19, Issue 1, 2021, 10.1186/s12915-021-01146-6.

Requirements

You'll need Python 3.6 or greater to run phirbo.py.

Python dependencies:

• numpy >= 1.16.4

  Use pip to install: python3 -m pip install numpy

• pandas >= 0.22.1.

  Use pip to install: python3 -m pip install pandas

Installation

Since Phirbo is just a single script, no installation is required. You can simply clone it and run it:

git clone https://github.com/aziele/phirbo.git
phirbo/phirbo.py --help

If you plan on using it often, you can copy it to someplace in your PATH variable for easier access:

cp phirbo/phirbo.py ~/.local/bin
phirbo.py --help

Also available from Bioconda:

conda create -n phirbo -c conda-forge -c bioconda phirbo
conda activate phirbo
phirbo.py --help

Method

Phirbo links phage to host sequences through other intermediate sequences that are potentially homologous to both phage and host sequences.

To link phage (P) to host (H) sequence through intermediate sequences, phage and host sequences need to be used as queries in two separate sequence similarity searches (e.g., BLAST) against the same reference database of prokaryotic genomes (D). One BLAST search is performed for phage query (P) and the other for host query (H). The two lists of BLAST results, P → D and H → D, contain prokaryotic genomes ordered by decreasing score. To avoid a taxonomic bias due to multiple genomes of the same prokaryote species (e.g., Escherichia coli), prokaryotic species can be ranked according to their first appearance in the BLAST list. In this way, both ranked lists represent phage and host profiles consisting of the ranks of top-score prokaryotic species.

Phirbo estimates the phage-host relationship by comparing the content and order between phage and host ranked lists using Rank-Biased Overlap (RBO) measure. Briefly, RBO fosters comparison of ranked lists of different lengths with heavier weights for matching the higher-ranking items. RBO ranges between 0 and 1, where 0 means that the lists are disjoint (have no items in common) and 1 means that the lists are identical in content and order.

Input data

You need to provide ranked lists - separately in two directories - for phage and bacteria genomes. Every genome should have its own ranked list in a text file. The text format lists bateria species separated by a new line (if two or more species are the same in rank they should be comma-separated in one line) (see the example file format: example/virus/NC_000866.txt).

Quick usage

To run Phirbo provide two input directories (for phages and bacteria) containing your ranked lists, and an output file name.

phirbo.py example/virus/ example/host/ example/predictions.csv

This will output two files:

• predictions.csv containing phage-host predictions (i.e., a top score host for each phage). See the example output: example/predictions.csv
• predictions.matrix.csv containing a matrix of scores between every phage and every host (phages in columns, bacteria in rows). See the example output matrix: example/predictions.matrix.csv

Full usage

usage: phirbo.py [-h] [--p P] [--k K] [--t NUM_THREADS]
                 virus_dir host_dir output_file

Phirbo (v1.0) predicts hosts from phage (meta)genomic data

positional arguments:
  virus_dir        Input directory w/ ranked lists for viruses
  host_dir         Input directory w/ ranked lists for hosts
  output_file      Output file name

optional arguments:
  -h, --help       show this help message and exit
  --p P            RBO parameter in range (0, 1) determines the degree of top-
                   weightedness of RBO measure. High p implies strong emphasis
                   on top ranked items [default = 0.75]
  --k K            Truncate all ranked lists to the first `k` rankings to
                   calculate RBO. To disable the truncation use --k 0 [default
                   = 30]
  --t NUM_THREADS  Number of threads (CPUs) [default = 32]

Python module

You can use phirbo.py as a Python module.

import phirbo

phage_list = phirbo.read_list('example/virus/NC_000866.txt')
host_list = phirbo.read_list('example/host/NC_010473.txt')
score = phirbo.rbo(phage_list, host_list)

phage_list = [{'E.coli'}, {'S.boydii'}, {'S.flexneri'}, {'Y.rohdei', 'Y.ruckeri'}]
host_list = [{'E.coli'}, {'S.flexneri'}, {'S.boydii'}, {'S.dysenteriae'}, {'E.toletana'}]
score = phirbo.rbo(phage_list, host_list)
score = phirbo.rbo(phage_list, host_list, p=0.98)

Further analyses

The output files can be further analyzed with R, Python or Excel spreadsheet.

Top n hosts for each phage

R:

data <- read.csv("predictions.matrix.csv", row.names=1)
top_n_hosts <- 3

for (col in colnames(data)) {
    print(data[order(data[col], decreasing = T)[1:top_n_hosts],][col])
}

Python:

import pandas as pd

data = pd.read_csv("predictions.csv.matrix.csv", index_col=0)
top_n_hosts = 3

print(data.unstack()
          .groupby(level=0, group_keys=False)
          .nlargest(top_n_hosts)
          .reset_index()
          .to_string(header=None, index=False))

Filter phage-host pairs by a score

For example, show phage-host pairs with score ≥ 0.8.

R:

data <- read.csv("predictions.csv.matrix.csv", row.names=1)
min_score <- 0.5

for(col in 1:ncol(data)){
  d <- data[data[col] >= min_score,]
  d <- d[order(d[col]),][col]
  print(d)
}

Python:

import pandas as pd

data = pd.read_csv("predictions.csv.matrix.csv", index_col=0)
min_score = 0.8

s = data.unstack()
print(s[s >= min_score]
     .reset_index()
     .sort_values(["level_0", 0], ascending=[True, False])
     .to_string(header=None, index=False))

Distribution of scores for a given phage

R:

phage_id <- "NC_000866"
data <- read.csv("predictions.csv.matrix.csv", row.names=1)

scores <- data[[phage_id]]
hist(scores)

print(paste('Min    :', min(scores)))
print(paste('Q1     :', quantile(scores, 0.25)))
print(paste('Median :', median(scores)))
print(paste('Q3     :', quantile(scores, 0.75)))
print(paste('Max    :', max(scores)))

Python:

import pandas as pd

phage_id = "NC_000866"
data = pd.read_csv("predictions.csv.matrix.csv", index_col=0)

scores = data[phage_id]
hist = scores.hist()                    # Needs matplotlib.
hist.figure.savefig('figure.pdf')

print(f'Min    : {scores.min()}')
print(f'Q1     : {scores.quantile(0.25)}')
print(f'Median : {scores.median()}')
print(f'Q3     : {scores.quantile(0.75)}')
print(f'Max    : {scores.max()}')

Distribution of scores for all phages

R:

data <- read.csv("predictions.csv.matrix.csv", row.names=1)
scores <- unlist(data,use.names = FALSE)
hist(scores)

Python:

import pandas as pd

data = pd.read_csv("predictions.csv.matrix.csv", index_col=0)
scores = data.stack()
hist = scores.hist(grid=False)
hist.figure.savefig('figure.pdf')

License

GNU General Public License, version 3