Provided by: bio-rainbow_2.0.4-1_amd64 bug

NAME

       rainbow - manual page for rainbow 2.0.4 -- <ruanjue@gmail.com, chongzechen@gmail.com>

SYNOPSIS

       rainbow <cmd> [options]

DESCRIPTION

       rainbow 2.0.4 -- <ruanjue@gmail.com, chongzechen@gmail.com>

              cluster

       Input     File    Format:    paired    fasta/fastq    file(s)    Output    File    Format:
       <seqid:int>\t<cluster_id:int>\t<read1:string>\t<read2:string>

       -1 <string> Input fasta/fastq file, supports multiple '-1'

       -2 <string> Input fasta/fastq file, supports multiple '-2' [null]

       -l <int>
              Read length, default: 0 variable

       -m <int>
              Maximum mismatches [4]

       -e <int>
              Exactly matching threshold [2000]

       -L     Low level of polymorphism

              div

       Input File  Format:  <seqid:int>\t<cluster_id:int>\t<read1:string>\t<read2:string>  Output
       File                                                                               Format:
       <seqid:int>\t<cluster_id:int>\t<read1:string>\t<read2:string>[\t<pre_cluster_id:int>]

       -i <string> Input file [stdin]

       -o <string> Output file [stdout]

       -k <int>
              K_allele, min variants to create a new group [2]

       -K <int>
              K_allele, divide regardless of frequency when num of  variants  exceed  this  value
              [50]

       -f <float>
              Frequency, min variant frequency to create a new group [0.2]

              merge

       Input                                     File                                     Format:
       <seqid:int>\t<cluster_id:int>\t<read1:string>\t<read2:string>[\t<pre_cluster_id:int>]

       -i <string> Input rbasm output file [stdin]

       -a     output assembly

       -o <string> Output file for merged contigs, one line per cluster [stdout]

       -N <int>
              Maximum number of divided clusters to merge [300]

       -l <int>
              Minimum overlap when assemble two reads (valid only when '-a' is opened) [5]

       -f <float>
              Minimum fraction of similarity when assembly  (valid  only  when  '-a'  is  opened)
              [0.90]

       -r <int>
              Minimum number of reads to assemble (valid only when '-a' is opened) [5]

       -R <int>
              Maximum number of reads to assemble (valid only when '-a' is opened) [300]

       Usage: rainbow <cmd> [options]

              cluster

       Input     File    Format:    paired    fasta/fastq    file(s)    Output    File    Format:
       <seqid:int>\t<cluster_id:int>\t<read1:string>\t<read2:string>

       -1 <string> Input fasta/fastq file, supports multiple '-1'

       -2 <string> Input fasta/fastq file, supports multiple '-2' [null]

       -l <int>
              Read length, default: 0 variable

       -m <int>
              Maximum mismatches [4]

       -e <int>
              Exactly matching threshold [2000]

       -L     Low level of polymorphism

              div

       Input File  Format:  <seqid:int>\t<cluster_id:int>\t<read1:string>\t<read2:string>  Output
       File                                                                               Format:
       <seqid:int>\t<cluster_id:int>\t<read1:string>\t<read2:string>[\t<pre_cluster_id:int>]

       -i <string> Input file [stdin]

       -o <string> Output file [stdout]

       -k <int>
              K_allele, min variants to create a new group [2]

       -K <int>
              K_allele, divide regardless of frequency when num of  variants  exceed  this  value
              [50]

       -f <float>
              Frequency, min variant frequency to create a new group [0.2]

              merge

       Input                                     File                                     Format:
       <seqid:int>\t<cluster_id:int>\t<read1:string>\t<read2:string>[\t<pre_cluster_id:int>]

       -i <string> Input rbasm output file [stdin]

       -a     output assembly

       -o <string> Output file for merged contigs, one line per cluster [stdout]

       -N <int>
              Maximum number of divided clusters to merge [300]

       -l <int>
              Minimum overlap when assemble two reads (valid only when '-a' is opened) [5]

       -f <float>
              Minimum fraction of similarity when assembly  (valid  only  when  '-a'  is  opened)
              [0.90]

       -r <int>
              Minimum number of reads to assemble (valid only when '-a' is opened) [5]

       -R <int>
              Maximum number of reads to assemble (valid only when '-a' is opened) [300]

SEE ALSO

       The  full  documentation  for  rainbow is maintained as a Texinfo manual.  If the info and
       rainbow programs are properly installed at your site, the command

              info rainbow

       should give you access to the complete manual.

rainbow 2.0.4 -- <ruanjue@gmail.com, chongzAugustg2015.com>                            RAINBOW(1)