CSHL 2015 Computational and Comparative Genomics

Class Project

transfer data from student's laptop to CSHL

Transferring data to class:

On student's laptop where the data exists, verify enough disk space
for this operation:

$ cd        # cd with no argument will go to HOME
$ df -h .   # verify disk space available in this directory == on this filesystem
Filesystem   Size   Used  Avail Capacity  iused   ifree %iused  Mounted on
/dev/disk1  233Gi  226Gi  6.1Gi    98% 59386162 1592652   97%   /

# looks like 6 Gb free    ^^^^^

Go to the directory of data to transfer
$ cd oenothera

Measure the amount of data to package:

$ du -hsc *

2.1G    transcriptomes
2.1G    total

Total data is 2.1 Gb, the tar image compression will help.
Generate compressed tar image of this directory:

$ tar -cvzf $HOME/toCSHL.tgz ./

tar command arguments:
   c - create tar file
   v - verbose, show what is being packaged
   z - compress (gzip) while making tar image
   f - file name of tar image to construct
   ./ - package up everyting in this directory

Take a look at the resulting file:

$ cd      # return to home directory where the result file is
$ ls -l *.tgz
-rw-rw-r--  1 hclawson  staff  770380941 Oct 29 22:26 toCSHL.tgz

It is now only 735 Mb of compressed data:

$ du -hsc *.tgz
735M    toCSHL.tgz

Transfer this file to the workstation at CSHL

$ scp -p toCSHL.tgz hclawson@ecg15.cshl.edu:.

scp option '-p' means preserve date/time stamps on the file so it will
appear identical in the copy.

Magic hand-waving here since there are various pathways through the
networking here from wifi laptop connections to the class workstations.
Talk with Dan for correct connection procedures.

Now, on the desktop machines for the class, in the home directory,
unpack the tar image here:
$ mkdir oenothera
$ cd oenothera
$ tar xvzf ../toCSHL.tgz
$ ls -l
total 80
drwxr-xr-x  1 hclawson  staff   330 Oct 29 21:47 transcriptomes

survey names in sequences

To use the UCSC genome browser to view this work, it is helpful to reduce the very long names in the transcriptome fasta sequence
that were constructed by the assembler. A pattern is seen in the names that suggests a substitution algorithm. They all start with:
>Locus_<sequenceNumber>_otherBusiness
or
>NODE_<sequenceNumber>_otherBusiness
The <sequenceNumber> identifiers appear to be unique within each fasta sequence, thus, the Locus_ or NODE_ can be replaced with
a name related to the transcript, and the _otherBusiness can be discarded.
$ awk -F'_' '{print $1}' all.contig.names.txt | sort | uniq -c
1985474 >Locus
1228519 >NODE
As a test, constructing fasta with those short names:
$ cd ~/oentothera/transcriptomes/assemblies
find . -type f | sed -e 's#^./##;' | grep fasta | while read F
do
  B=`basename ${F}`
  D=`dirname ${F}`
  id=`echo $D | sed -e 's/-.*//;'`
  printf "%s %s\n" "${id}" "${B}" 1>&2
  sed -e "s#^>Locus_#>${id}.#; s#^>NODE_#>${id}.#; s#_.*##;" ${F}
done | gzip -c > $HOME/all.contigs.fa.gz

# verify nothing lost (using kent command line programs from ~/bin/)
# from the original source

$ faSize */*fasta*
1354344256 bases (51622 N's 1354292634 real 1351051824 upper 3240810 lower) in 3213993 sequences in 63 files

# to the short name contigs:

$ cd
$ $ faSize all.contigs.fa.gz
1354344256 bases (51622 N's 1354292634 real 1351051824 upper 3240810 lower) in 3213993 sequences in 1 files

# same numbers, nothing lost

For this procedure, it would be good to record the name to name translation so that the original names could be used in the assembly track and thus be available to reference from the genome browser view back to this set of short named contigs..

construct artifical assemblies of each transcriptome

It is convenient for alignment processing and genome browser display to
concatenate all the contigs from one transcriptome into a single sequence,
inserting 100 bases of 'N' between each contig.  Using the following perl script
on each set of transcriptomes:

=================================================
#!/usr/bin/env perl

use strict;
use warnings;

sub usage {
  printf STDERR "usage: ./fakeAssembly [name] [fastaFile] > name.assembly.fa\n";
}

my $argc = scalar(@ARGV);
if ($argc != 2) {
  usage;
  exit 255;
}

sub outputGap {
  for (my $i = 0; $i < 100; ++$i) {
    printf "N";
  }
  printf "\n";
}

my $name = shift;
my $fastaFile = shift;

my $contigsDone = 0;
open (FH, "<$fastaFile") or die "can not read $fastaFile";
while (my $line = <FH>) {
  chomp $line;
  if ($line =~ m/^>/) {
    if (0 == $contigsDone) {
       printf ">%s\n", $name
    } else {
      outputGap;
    }
    ++$contigsDone;
  } else {
    printf "%s\n", $line;
  }
}
close (FH);

It would actually be best to order the contigs by size, largest ones first. This would look more convenient in the browser display where all these contains are going to appear as if they were all on one single chromosome.

Running that perl script on each directory that has a transcriptome fasta file:

#!/bin/bash

mkdir $HOME/fakeAssemblies

find . -type d | grep -v "^.$" | sed -e 's#^./##;'  | while read D
do
  fasta=`ls $D | grep fasta`
  name=`echo $D | awk -F'-' '{print $1}'`
  species=`echo $D | cut -d_ -f2-`
  faName="${name}_${species}"
  printf "%s %s\n" "${faName}" "$fasta" 1>&2
  ./fakeAssembly.pl "${name}_${species}" ${D}/${fasta} | gzip -c > $HOME/fakeAssemblies/${faName}.fa.gz
done

construct assembly hub files

Convert all the short-named artificial assemblies of each transcriptome into a single 2bit file:

faToTwoBit fakeAssemblies/*.fa.gz oenothera.2bit