3.6 Genome
Assembly
Genome assembly can be performed in multiple ways depending on
the read sets available for input, the abundance of data available for
assembly (in the form of read depth), and the availability of a very
closely related reference genome for the organism of interest. There
are two broad categories of genome assembly strategies available for
prokaryotes: de novo genome assembly and reference-based assembly. Briefly, de novo assembly is a broad term for numerous algorithms that attempt to “stitch” together sequencing reads based on
matching regions into progressively longer strings of contiguous
sequences (called contigs) without a priori knowledge of what the
sequence should be. While this method is an unbiased method of
assembling a genome, results are highly dependent on the sequencing technology used and typically require much deeper coverage
than alternative reference-based assembly strategies. Illumina only
de novo assemblies are almost always fragmented but contain
highly accurate basecalls. On the other hand, Nanopore and other
long read-only assemblies are often “complete” but almost always
contain a substantial number of erroneous bases. Reference-based
strategies, as the name implies, use an existing genome as a reference to guide where reads should be placed and then call differences
between the reads and the reference genome. Here, substantially
less coverage (often an order of magnitude less than de novo
assembly) is needed to make accurate base calls, but results can be
severely biased by the reference genome used, especially if care is
not taken to ensure that the reference genome is as closely related as
possible to the sample being sequenced. In either case, the result is
a set of DNA sequences representing a consensus genome sequence
stored in a FASTA file (file extension “.fasta” or “.fa”). FASTA files
are similar to FASTQ files but do not contain quality scores or an
extra comment line per line of DNA sequence.
3.6.1 Short Read-Only
De Novo Assembly
Illumina-only de novo assemblies are some of the most ubiquitous
prokaryotic genomes available on GenBank, using Illumina paired
end sequencing between depths of 50–150Â to assemble genomes
into 100 contigs. A variety of de novo assembly programs
employing a variety of algorithms have been developed within this
space, including but not limited to Velvet [41], ABySS [42], Tadpole [43], SPAdes [44], and MIRA [45]. All have different
strengths and weaknesses, but SPAdes is one of the most common
prokaryotic de novo assemblers as it often returns the highest N50
value as compared to alternative de novo assemblers. However, a
weakness of the SPAdes assembler is its tendency to produce an
abundance of “chaff” contigs or contigs approximately 200 bp or
shorter that are unusable for downstream analyses. These chaff
contigs should be removed from the FASTA file using a tool such
as seqtk [46] or seqkit [47] before proceeding with downstream
steps. A recent tool called Unicycler [48] makes some improvements on the SPAdes algorithm by including several postassembly
steps that increase the completeness of a SPAdes assembly by
Mycobacterial ID and Characterization Using WGS
415
Assembly
Genome assembly can be performed in multiple ways depending on
the read sets available for input, the abundance of data available for
assembly (in the form of read depth), and the availability of a very
closely related reference genome for the organism of interest. There
are two broad categories of genome assembly strategies available for
prokaryotes: de novo genome assembly and reference-based assembly. Briefly, de novo assembly is a broad term for numerous algorithms that attempt to “stitch” together sequencing reads based on
matching regions into progressively longer strings of contiguous
sequences (called contigs) without a priori knowledge of what the
sequence should be. While this method is an unbiased method of
assembling a genome, results are highly dependent on the sequencing technology used and typically require much deeper coverage
than alternative reference-based assembly strategies. Illumina only
de novo assemblies are almost always fragmented but contain
highly accurate basecalls. On the other hand, Nanopore and other
long read-only assemblies are often “complete” but almost always
contain a substantial number of erroneous bases. Reference-based
strategies, as the name implies, use an existing genome as a reference to guide where reads should be placed and then call differences
between the reads and the reference genome. Here, substantially
less coverage (often an order of magnitude less than de novo
assembly) is needed to make accurate base calls, but results can be
severely biased by the reference genome used, especially if care is
not taken to ensure that the reference genome is as closely related as
possible to the sample being sequenced. In either case, the result is
a set of DNA sequences representing a consensus genome sequence
stored in a FASTA file (file extension “.fasta” or “.fa”). FASTA files
are similar to FASTQ files but do not contain quality scores or an
extra comment line per line of DNA sequence.
3.6.1 Short Read-Only
De Novo Assembly
Illumina-only de novo assemblies are some of the most ubiquitous
prokaryotic genomes available on GenBank, using Illumina paired
end sequencing between depths of 50–150Â to assemble genomes
into 100 contigs. A variety of de novo assembly programs
employing a variety of algorithms have been developed within this
space, including but not limited to Velvet [41], ABySS [42], Tadpole [43], SPAdes [44], and MIRA [45]. All have different
strengths and weaknesses, but SPAdes is one of the most common
prokaryotic de novo assemblers as it often returns the highest N50
value as compared to alternative de novo assemblers. However, a
weakness of the SPAdes assembler is its tendency to produce an
abundance of “chaff” contigs or contigs approximately 200 bp or
shorter that are unusable for downstream analyses. These chaff
contigs should be removed from the FASTA file using a tool such
as seqtk [46] or seqkit [47] before proceeding with downstream
steps. A recent tool called Unicycler [48] makes some improvements on the SPAdes algorithm by including several postassembly
steps that increase the completeness of a SPAdes assembly by
Mycobacterial ID and Characterization Using WGS
415