52
A. Meyerdierks and F.O. Glöckner
sequencing and/or short read sequencing reads, analysis is particularly difficult.
Metagenomic data analysis is still in its infancy when it comes to meeting this sort
of challenge. To address these problems, this description of the bioinformatic analysis of metagenomic data will be split into the analysis of assembled metagenomes
and the analysis of short read metagenomics resulting from high throughput/high
diversity approaches.
2.4.1 Fragment Assembly and Binning
To obtain fragments that can be assembled into scaffolds and contigs of several
kilobases in length it is best to sequence large insert constructs (BACs or fosmids).
These constructs are generally sequenced using a shotgun approach as described,
for example, by Sambrook and Russel 2001. About 400 sequencing reactions are
required to sequence a fosmid sized fragment (∼ 40 kb) to eight-fold coverage.
Because all the resulting reads belong to the same large insert construct, assembly
is then easily performed with standard programs such as Phrap (www.phrap.com),
JAZZ (Aparicio et al. 2002), Arachne 2 (Jaffe et al. 2003) or the Celera Assembler
(http://sourceforge.net/projects/wgs-assembler). If sequencing capacities are limited
the crucial step in this approach is the selection of the appropriate BACs or fosmids for sequencing. Standard approaches include screening for phylogenetic or
metabolic marker genes or random insert-end sequencing.
If a whole metagenome shotgun approach is chosen, the situation becomes immediately more complicated. The success of such a strategy strongly depends on the
phylogenetic complexity of the sample. In low diversity environments, containing
only a small number of dominant species (5–10), it is possible to obtain a good
set of assembled contigs and scaffolds of up to several megabases in length with
reasonable sequencing efforts (Tyson et al. 2004, Martin et al. 2006, Woyke et al.
2006). Nevertheless, problems can arise from biology due to high genome complexity even within closely related species (Johnson and Slatkin 2006) and the
presence of ubiquitous genomic elements like transposons, viruses and inserted
phages (Salzberg and Yorke 2005). Technical problems are even more pronounced.
In a recent study Mavromatis et al. (Mavromatis et al. 2007) showed an increased
chance for chimeric contigs particularly for sequence fragments below 8 kb. In general this is in line with the fact that none of the currently used assembly programs
has been specifically built for metagenomes. In particular, more progressive aligners such as Phrap or JAZZ try to incorporate as many of the reads as possible,
resulting in a large number of contigs. This is the primary goal in single genome
assembly programs. For metagenomes a more conservative approach such as that
implemented by the Arachne assembler is preferable, as this significantly decreases
the chance of misassemblies.
To maintain an optimal balance between the number of contigs and the probability of obtaining chimeras, iterative cycles of “binning” and assembly can be
performed.
A. Meyerdierks and F.O. Glöckner
sequencing and/or short read sequencing reads, analysis is particularly difficult.
Metagenomic data analysis is still in its infancy when it comes to meeting this sort
of challenge. To address these problems, this description of the bioinformatic analysis of metagenomic data will be split into the analysis of assembled metagenomes
and the analysis of short read metagenomics resulting from high throughput/high
diversity approaches.
2.4.1 Fragment Assembly and Binning
To obtain fragments that can be assembled into scaffolds and contigs of several
kilobases in length it is best to sequence large insert constructs (BACs or fosmids).
These constructs are generally sequenced using a shotgun approach as described,
for example, by Sambrook and Russel 2001. About 400 sequencing reactions are
required to sequence a fosmid sized fragment (∼ 40 kb) to eight-fold coverage.
Because all the resulting reads belong to the same large insert construct, assembly
is then easily performed with standard programs such as Phrap (www.phrap.com),
JAZZ (Aparicio et al. 2002), Arachne 2 (Jaffe et al. 2003) or the Celera Assembler
(http://sourceforge.net/projects/wgs-assembler). If sequencing capacities are limited
the crucial step in this approach is the selection of the appropriate BACs or fosmids for sequencing. Standard approaches include screening for phylogenetic or
metabolic marker genes or random insert-end sequencing.
If a whole metagenome shotgun approach is chosen, the situation becomes immediately more complicated. The success of such a strategy strongly depends on the
phylogenetic complexity of the sample. In low diversity environments, containing
only a small number of dominant species (5–10), it is possible to obtain a good
set of assembled contigs and scaffolds of up to several megabases in length with
reasonable sequencing efforts (Tyson et al. 2004, Martin et al. 2006, Woyke et al.
2006). Nevertheless, problems can arise from biology due to high genome complexity even within closely related species (Johnson and Slatkin 2006) and the
presence of ubiquitous genomic elements like transposons, viruses and inserted
phages (Salzberg and Yorke 2005). Technical problems are even more pronounced.
In a recent study Mavromatis et al. (Mavromatis et al. 2007) showed an increased
chance for chimeric contigs particularly for sequence fragments below 8 kb. In general this is in line with the fact that none of the currently used assembly programs
has been specifically built for metagenomes. In particular, more progressive aligners such as Phrap or JAZZ try to incorporate as many of the reads as possible,
resulting in a large number of contigs. This is the primary goal in single genome
assembly programs. For metagenomes a more conservative approach such as that
implemented by the Arachne assembler is preferable, as this significantly decreases
the chance of misassemblies.
To maintain an optimal balance between the number of contigs and the probability of obtaining chimeras, iterative cycles of “binning” and assembly can be
performed.