added Fast ani component (#130)

* added fast ani component

* removed unused params

* fixed output name

* fixed input name

* fixed image tag

* added log to stderr

* added implementation of multi fasta parsing for fastani

* added channel for parameter fragLen

* added parser for main fasta

* changed docker img tag

* fixed file naming

* fixed pep8 issue

* added join to file name

* added documentation and publish dir

* updated changelog

* fixed file naming

* removed output_type

* fixed changelog

Author: tiagofilipe12

changelog.md

@@ -5,7 +5,11 @@
 ### Components changes
 
 Updated images for components `mash_dist`, `mash_screen` and 
-`mapping_patlas`
+`mapping_patlas`.
+
+### New components
+- Added component `fast_ani`.
+
 
 ## 1.3.0
 

docs/user/available_components.rst

@@ -99,7 +99,13 @@ Distance Estimation
   plasmid database and generates a JSON input file for pATLAS. This component
   searches for containment of a given sequence in read sequencing data.
   However if a different database is provided it can use mash screen for other
-  purporses.
+  purposes.
+
+- :doc:`components/fast_ani`: Performs pairwise comparisons between fastas,
+given a multifasta as input for fastANI. It will split the multifasta into
+single fastas that will then be provided as a matrix. The output will be the
+all pairwise comparisons that pass the minimum of 50 aligned sequences with a
+default length of 200 bp.
 
 - :doc:`components/mash_sketch_fasta`: Performs mash sketch for fasta files.
 

docs/user/components/fast_ani.rst

@@ -0,0 +1,47 @@
+fast_ani
+========
+
+Purpose
+-------
+
+This component performs pairwise comparisons between fastas,
+given a multifasta as input for fastANI. It will split the multifasta into
+single fastas that will then be provided as a matrix. The output will be the
+all pairwise comparisons that pass the minimum of 50 aligned sequences with a
+default length of 200 bp.
+
+Input/Output type
+------------------
+
+- Input type: ``fasta``
+- Output type: ``None``
+
+
+Parameters
+----------
+
+- ``fragLen``: Sets the minimum size of the fragment to be passed to
+`--fragLen` argument of fastANI.
+
+
+Published results
+-----------------
+
+- ``results/fast_ani/``: A text file with the extension `.out`, which has all
+the pairwise comparisons between sequences, reporting ANI.
+
+
+Published reports
+-----------------
+
+None.
+
+
+Default directives
+------------------
+
+- ``fastAniMatrix``:
+    - ``container``: flowcraft/fast_ani
+    - ``version``: 1.1.0-2
+    - ``cpus``: 20
+    - ``memory``: { 30.GB * task.attempt }

flowcraft/generator/components/distance_estimation.py

@@ -200,3 +200,32 @@ def __init__(self, **kwargs):
         self.status_channels = [
             "mashSketchFastq",
         ]
+
+
+class FastAniMatrix(Process):
+
+    def __init__(self, **kwargs):
+
+        super().__init__(**kwargs)
+
+        self.input_type = "fasta"
+
+        self.params = {
+            "fragLen": {
+                "default": 3000,
+                "description": "Set size of fragment. Default: 3000."
+            }
+        }
+
+        self.directives = {
+            "fastAniMatrix": {
+                "container": "flowcraft/fast_ani",
+                "version": "1.1.0-2",
+                "cpus": 20,
+                "memory": "{ 30.GB * task.attempt }"
+            },
+        }
+
+        self.status_channels = [
+            "fastAniMatrix",
+        ]

flowcraft/generator/engine.py

@@ -68,6 +68,7 @@
         "filter_poly": readsqc.FilterPoly,
         "integrity_coverage": readsqc.IntegrityCoverage,
         "fasterq_dump": downloads.FasterqDump,
+        "fast_ani": distest.FastAniMatrix,
         "kraken": meta.Kraken,
         "mafft": alignment.Mafft,
         "mapping_patlas": mapping_patlas.PatlasMapping,

flowcraft/generator/templates/fast_ani.nf

@@ -0,0 +1,30 @@
+IN_fragLen_{{ pid }} = Channel.value(params.fragLen{{ param_id }})
+
+// runs fast ani for multiple comparisons (many to many mode)
+process fastAniMatrix_{{ pid }} {
+
+    {% include "post.txt" ignore missing %}
+
+    tag { sample_id }
+
+     publishDir 'results/fast_ani/fast_ani_{{ pid }}/',
+
+    input:
+    set sample_id, file(fasta) from {{ input_channel }}
+    val fragLenValue from IN_fragLen_{{ pid }}
+
+    output:
+    set sample_id, fasta, file("*.out")
+    {% with task_name="fastAniMatrix", sample_id="sample_id" %}
+    {%- include "compiler_channels.txt" ignore missing -%}
+    {% endwith %}
+
+    """
+    mkdir fasta_store
+    fasta_spliter.py ${fasta}
+    fastANI --ql files_fastani.txt --rl files_fastani.txt \
+    -t ${task.cpus} --fragLen ${fragLenValue} \
+    -o ${sample_id.take(sample_id.lastIndexOf("."))}_fastani.out
+    """
+
+}

flowcraft/templates/fasta_spliter.py

@@ -0,0 +1,53 @@
+#!/usr/bin/env python3
+
+"""
+Purpose
+-------
+
+This module is intended to split all fastas in a multifasta file into different
+fasta files.
+
+Code documentation
+------------------
+
+"""
+
+import os
+import sys
+
+
+def main():
+
+    cwd = os.getcwd()
+    # a var to check if out_handle is started and if so it enables to control
+    # how it should be closed
+    out_handle = False
+    # opens the input file of the process
+    input_file = open(sys.argv[1])
+    # a file with the list of all paths to fasta files that will be used by
+    # fastANI
+    list_files = open("files_fastani.txt", "w")
+    # iterates by each entry in the fasta file
+    for line in input_file:
+        if line.startswith(">"):
+            if out_handle:
+                out_handle.close()
+            # writes the output to fasta store folder inside cwd, respective
+            # workdir
+            path_to_file = os.path.join(cwd, "fasta_store",
+                                        "_".join(line.split("_")[0:3])
+                                        .replace(">", "") + ".fas")
+            # writes to list of files
+            list_files.write(path_to_file + "\n")
+            out_handle = open(path_to_file, "w")
+            out_handle.write(line)
+        else:
+            out_handle.write(line)
+
+    out_handle.close()
+    input_file.close()
+    list_files.close()
+
+
+if __name__ == "__main__":
+    main()