Setting the parameter “map_len” in SOAPdenovo assembly software

[Godevil]

Hi,

I'm now using SOAPdenovo to assemble a big genome(Illumina GAIIx data；avg insert 200bp；pair-end reads 150bp x 2 ).

From the software manual, I found there is a parameter called "map_len".

In the example from the software manual, the default map_len=32, when maximal read length = 50 and average insert size = 200). In my case, my PE read length = 150, so, should I increase the map_len value?

Does anyone know how to set up a suitable map_len value?

PS: An example.config from the software manual:
"
#maximal read length
max_rd_len=50
[LIB]
#average insert size
avg_ins=200
...
...
#minimum aligned length to contigs for a reliable read location (default 32)
map_len=32
...

[LIB]
avg_ins=2000
...
...
#minimum aligned length to contigs for a reliable read location
#(default 35 for large insert size)
map_len=35
... "

[tonybolger]

In the example from the software manual, the default map_len=32, when maximal read length = 50 and average insert size = 200). In my case, my PE read length = 150, so, should I increase the map_len value?

For good quality PE libraries, you can probably go a bit bigger - it should help find uniqueness without losing too many reads due to error. The 'map' step will tell you how many reads aligned, so you can play about to maximise that. Then again, link counts are usually not a problem in PE libraries.

For MP libraries, you probably don't want to go too big, because you lose reads once you pass the splicing point, which is more or less randomly located.

BTW, do you really have paired 150bp reads of a 200bp fragment? If so, you're likely to have a lot of adapter in there. You might also want to consider 'pre-flattening' the read pairs into a single longer read, and assembling as SE reads.

[Godevil]

BTW, do you really have paired 150bp reads of a 200bp fragment? If so, you're likely to have a lot of adapter in there.

Thank you very much！ I really found about 10% of my reads contain adapter sequences at their 3' ends. I try to use cutadapt software to trim those adapter sequences.

But, I don't know how to detect and trim adapter/primer-dimer in my reads. Could you give me some advice about this? Can cutadapt do this work?

[Godevil]

The right way to set up the “map_len” value in SOAPdenovo software！

There is an official answer came from a technician in Beijing Genomics Institute (BGI).

I want to share it with everyone here.

"Just leave the option "map_len" alone when you are doing initial assembly.
After the success of initial assembly, try increase "map_len" to gain a better scaffold result (or, sometime, worse) provided that, 1. The "map_len" option will not effect on libraries that reads are long than 100, 2. It's not wise to set over 50.
3. Increase 1 by 1, optimal results usually gain when increase by 2 or 3, don't increase to much, especially for genomes with higher heterozygosity.
"

[tonybolger]

Thank you very much！ I really found about 10% of my reads contain adapter sequences at their 3' ends. I try to use cutadapt software to trim those adapter sequences.

But, I don't know how to detect and trim adapter/primer-dimer in my reads. Could you give me some advice about this? Can cutadapt do this work?

In my experience, you get two problems with adapters. Either the start of the read is adapter followed by junk (i guess this is caused by two adapters sticking together), or you get a short correct fragment but the end of the read is the 'other' adapter reverse-complemented. Given your read length and fragment size, i'd expect the latter to be common - 10% isn't actually bad.

I've developed my own tool to 'pre-process' reads, trimming adapters, and using various criteria to filter by quality, and handles reads becoming 'unpaired'. It's probably not release-ready yet, but if you're brave, i can send you a copy and instructions. I developed it because i was getting tired of the overhead of running 3-4 different trimming tools over 200GB datasets.

On the other hand, i'm sure cutadapt will do the job.