Function: SGA

SGA

Perform de novo assembly for sequencing reads using String Graph Assembler (SGA). SGA is designed for experiments using a read length of 100 base pairs or more. Minimum recommended coverage is 20-30x.

The following SGA executables must be on path: sga, sga-align, sga-bam2de.pl and sga-astat.py. Also, abyss-fixmate, bwa and samtools need to be on path. Please follow SGA documentation for installation.

SGA parameters may need to be tuned for both performance and assembly quality reasons. The most important parameters are correctionK, minOverlapAssemble and minOverlapMerge. For machines with limited memory, indexBatchSize may need to be set. SGA is parallelized at two levels: assembly of a single genome is multithreaded, and multiple genomes can be assembled using a cluster.

Below are benchmarks for SGA 0.10.12 that provide hints on CPU and memory requirements:

E. coli, 1.7 Gbp sequence: 1.0 GB memory, 7 CPU hours

C. elegans, 6.7 Gbp sequence: 8.7 GB memory, 49 CPU hours

Human, 148 Gbp sequence: 70 GB memory, 1900 CPU hours

Version	1.0
Bundle	sequencing
Categories	Assembly
Authors	Kristian Ovaska (kristian.ovaska@helsinki.fi)
Issue tracker	View/Report issues
Requires	SGA ; BWA ; SAMtools ; ABySS ; pysam ; Ruffus ; Scala
Source files	component.xml function.scala
Usage	Example with default values SGA( reads = SequenceSet, //mates = SequenceSet, correctionK = 41, indexBatchSize = 0, memoryGB = 16, minBranchLength = 150, minContigLength = 200, minOverlap = 45, minOverlapAssemble = 75, minOverlapMerge = 65, pairedEnd = true, quality = "phred33", threads = 8 )

Version

1.0

Bundle

sequencing

Categories

Assembly

Authors

Kristian Ovaska (kristian.ovaska@helsinki.fi)

Issue tracker

View/Report issues

Requires

SGA ; BWA ; SAMtools ; ABySS ; pysam ; Ruffus ; Scala

Source files

component.xml function.scala

Usage

Example with default values

Inputs

Name	Type	Mandatory	Description
reads	Array<SequenceSet>	Mandatory	Sequencing reads. SGA accepts input as FASTA, FASTQ and their gzipped variants. When paired-end sequencing is used, either these files contain the paired reads after one another, or paired reads are in the second input.
mates	Array<SequenceSet>	Optional	Paired sequencing reads. When paired-end sequencing is used and paired reads and not interlaced in the primary input, these files contain the paired reads in the same order as primary reads. The length of the mates array must be equal to the reads array and read-mate files must be in the same order.

Name

Type

Mandatory

Description

reads

Array<SequenceSet>

Mandatory

Sequencing reads. SGA accepts input as FASTA, FASTQ and their gzipped variants. When paired-end sequencing is used, either these files contain the paired reads after one another, or paired reads are in the second input.

mates

Array<SequenceSet>

Optional

Paired sequencing reads. When paired-end sequencing is used and paired reads and not interlaced in the primary input, these files contain the paired reads in the same order as primary reads. The length of the mates array must be equal to the reads array and read-mate files must be in the same order.

Outputs

Name	Type	Description
contigs	FASTA	Assembled contigs, i.e., contiguous sequences constructed from the reads.
scaffolds	FASTA	Assembled scaffolds, i.e., concatenated contigs. In non-paired end mode (pairedEnd=false), this file is empty.

Name

Type

Description

contigs

FASTA

Assembled contigs, i.e., contiguous sequences constructed from the reads.

scaffolds

FASTA

Assembled scaffolds, i.e., concatenated contigs. In non-paired end mode (pairedEnd=false), this file is empty.

Parameters

Name	Type	Default	Description
correctionK	int	41	Error correction k-mer size. Corresponds to the -k argument of "sga correct".
indexBatchSize	int	0	When constructing BWT indexes, this many reads are indexed in one batch, using disk-based BWT construction. If the value is 0, an efficient in-memory "ropebwt" algorithm is used; this is suitable for reads up to 200 bp and can index 1.5 billion reads using 64 GB memory. If the value is negative, an in-memory "sais" algorithm is used (suitable for long reads). For very large data sets, use a value in the range of 2-10 million.
memoryGB	int	16	An estimate on the maximum amount of memory in gigabytes that an individual SGA process will need. This is only used in a cluster environment.
minBranchLength	int	150	In the assembly step, branches shorter than this number of base pairs are removed. Corresponds to the -l argument in "sga assembly". The default is tuned of 100 bp reads; experiments using longer reads should increase this value.
minContigLength	int	200	Minimum contig length in scaffolding. Corresponds to the -m argument in sga-bam2de.pl, sga-astat.py, "sga scaffold" and "sga scaffold2fasta".
minOverlap	int	45	Minimum length of read overlap in the overlap computation step. Corresponds to the -m argument of "sga overlap". The value must be smaller than the read length.
minOverlapAssemble	int	75	Minimum length of read overlap in the assembly step. Corresponds to the -m argument of "sga assemble". The value must be smaller than the read length. Too small values increase memory and CPU usage. SGA recommends a value of 75 for 100 bp reads.
minOverlapMerge	int	65	Minimum length of read overlap in the FM index merge step. Corresponds to the -m argument of "sga fm-merge". The value must be smaller than the read length and minOverlapAssemble. Too small values increase memory and CPU usage. SGA recommends a value of 65 for 100 base pair reads.
pairedEnd	boolean	true	If false, paired-end sequencing is not used. If true, paired reads are either contained in the primary input, or split between the primary and secondary inputs.
quality	string	"phred33"	For FASTQ input files, type of base quality values present. Legal values are phred33 and phred64.
threads	int	8	Maximum number of threads used by one process.

Name

Type

Default

Description

correctionK

int

Error correction k-mer size. Corresponds to the -k argument of "sga correct".

indexBatchSize

int

When constructing BWT indexes, this many reads are indexed in one batch, using disk-based BWT construction. If the value is 0, an efficient in-memory "ropebwt" algorithm is used; this is suitable for reads up to 200 bp and can index 1.5 billion reads using 64 GB memory. If the value is negative, an in-memory "sais" algorithm is used (suitable for long reads). For very large data sets, use a value in the range of 2-10 million.

memoryGB

int

An estimate on the maximum amount of memory in gigabytes that an individual SGA process will need. This is only used in a cluster environment.

minBranchLength

int

150

In the assembly step, branches shorter than this number of base pairs are removed. Corresponds to the -l argument in "sga assembly". The default is tuned of 100 bp reads; experiments using longer reads should increase this value.

minContigLength

int

200

Minimum contig length in scaffolding. Corresponds to the -m argument in sga-bam2de.pl, sga-astat.py, "sga scaffold" and "sga scaffold2fasta".

minOverlap

int

Minimum length of read overlap in the overlap computation step. Corresponds to the -m argument of "sga overlap". The value must be smaller than the read length.

minOverlapAssemble

int

Minimum length of read overlap in the assembly step. Corresponds to the -m argument of "sga assemble". The value must be smaller than the read length. Too small values increase memory and CPU usage. SGA recommends a value of 75 for 100 bp reads.

minOverlapMerge

int

Minimum length of read overlap in the FM index merge step. Corresponds to the -m argument of "sga fm-merge". The value must be smaller than the read length and minOverlapAssemble. Too small values increase memory and CPU usage. SGA recommends a value of 65 for 100 base pair reads.

pairedEnd

boolean

true

If false, paired-end sequencing is not used. If true, paired reads are either contained in the primary input, or split between the primary and secondary inputs.

quality

string

"phred33"

For FASTQ input files, type of base quality values present. Legal values are phred33 and phred64.

threads

int

Maximum number of threads used by one process.

Test cases

Test case	Parameters▼	IN reads	IN mates	OUT contigs	OUT scaffolds
case1	(missing)	reads	(missing)	contigs	scaffolds
case2_split	(missing)	reads	mates	contigs	scaffolds
case3_fq	properties	reads	mates	contigs	scaffolds
	quality = phred64
case4_unpaired	properties	reads	(missing)	contigs	scaffolds
	pairedEnd=false

Test case

Parameters▼

IN
reads

IN
mates

OUT
contigs

OUT
scaffolds

case1

(missing)

reads

(missing)

contigs

scaffolds