I'm not familiar with picard and their reorderSam function, but as far as I know/understand from their documentation they mean this:
The ordering of the contigs while using a reference sequence.
Like this:
Figure 5: Anatomy of
whole-genome assembly. In whole-genome assembly, the BAC fragments
(red line segments) and the reads from five individuals (black line
segments) are combined to produce a contig and a consensus sequence
(green line). The contigs are connected into scaffolds, shown in red,
by pairing end sequences, which are also called mates. If there is a
gap between consecutive contigs, it has a known size. Next, the
scaffolds are mapped to the genome (gray line) using sequence tagged
site (STS) information, represented by blue stars. © 2001 American
Association for the Advancement of Science Venter, C. et al. The
sequence of the human genome. Science 291, 1304–1351 (2001). All
rights reserved. (source)
ReorderSAM (Picard)
So in Picard you have your INPUT (File), the reads in this file are then mapped on the REFERENCE (File). This can also be seen in their code:
// write the reads in contig order
109
for (final SAMSequenceRecord contig : refDict.getSequences() ) {
110
final SAMRecordIterator it = in.query(contig.getSequenceName(), 0, 0, false);
111
writeReads(out, it, newOrder, contig.getSequenceName());
112
}
(code source)
ReorderSam reorders reads in a SAM/BAM file to match the contig
ordering in a provided reference file
Some more background
There are two main approches two obtain a genome sequence:
there are two "main" approches for this:
g. Second-generation
sequencing technologies produce millions of short(a few hundred bp)
strings of nucleotides (reads), which is ideal for resequencing when
reads are mapped to a reference genome (reference-based assembly). De
novo genome assembly based on second-generation sequencing is
challenging due to difficulties with GC- or AT-rich and homonucleotide
DNA stretches, which are under-represented in the sequencing output (source)
The characteristics of these are:
de novo
- no bias towards a reference genome
- no template to adapt to
- the assembly is normally more fragmented
- it normally works better for large-scale/median scale differences (source)
reference mapping
- less contigs
- in most methods the reads that don't map are not used in the final sequence (this is also the case with reorderSAM:
Reads mapped to contigs absent in the new reference are dropped
- you look what is similar to your reference genome
- SNPs and very small veriations are more easily positioned and compared among groups (source)
I would highly recommend to watch this short animation to differentiate between these two and understand what reference genome mapping is.
