Paired End Read Distribution Width for Similar Bacteria

[seb567]

Thanks Seb that's helpful. My peaks do overlap in the sense that the frequency of the paired end reads does not drop to zero between the peaks. This is an assembly of a single bacteria. The median and mean would be somewhere between the peaks and the mode would just select the tallest of the peaks. I have heard that a bimodal distribution is sometimes evidence of problems with the library preparation but can also be evidence of broken paired end reads because of repetitive sequence. Any ideas?

I never saw bimodal distributions with the standard Illumina Nextera.

I did saw bimodal distribution for very large inserts prepared by circularization. In these cases, the second lower peak is simply an expected contamination.

But in your case, it is strange.

[lg36]

Thanks Seb that's helpful. My peaks do overlap in the sense that the frequency of the paired end reads does not drop to zero between the peaks. This is an assembly of a single bacteria. The median and mean would be somewhere between the peaks and the mode would just select the tallest of the peaks. I have heard that a bimodal distribution is sometimes evidence of problems with the library preparation but can also be evidence of broken paired end reads because of repetitive sequence. Any ideas?

[seb567]

Does your peaks overlap ?

Usually assemblers will need an average and a standard deviation.
The mode is usually near the average, but there is usually a tail on
the right of the distribution.

You can use Ray to estimate the paired end length distribution.

Ray is a de novo assembler too.

Then, you can process with these estimates.

Command:

mpiexec -n 4 \
Ray \
-p file1R1.fastq file1R2.fastq \
-p file2R1.fastq file2R2.fastq \
-p file3R1.fastq file3R2.fastq \
-o MyAssembly

Then check MyAssembly/LibraryStatistics.txt and
MyAssembly/Library0.txt MyAssembly/Library1.txt MyAssembly/Library2.txt


This may be helpful.

see http://denovoassembler.sf.net

Is this a single-strain de novo assembly ?

Or a mix of strains ?

Dear All,

The consensus sequence I produce differs significantly based on the paired end distribution width I choose.

My paired end distribution is bi-modal with a smaller wider peak at around 150 and a taller narrower peak around 350.

Any ideas what paired end distribution width I should select for de novo assembly? I've asked the sequencing company for information regarding their insert size and my plan was to use this as a guide for selecting the PE width.

I've heard that the narrower the width (focused around the mode) the more broken paired reads but the more differences may be detected between similar bacteria. While the wider the difference fewer broken reads but less precision. Is there any truth in this??

Apologies for any misconceptions in advance.

Best wishes lg36

[lg36]

Dear All,

The consensus sequence I produce differs significantly based on the paired end distribution width I choose.

My paired end distribution is bi-modal with a smaller wider peak at around 150 and a taller narrower peak around 350.

Any ideas what paired end distribution width I should select for de novo assembly? I've asked the sequencing company for information regarding their insert size and my plan was to use this as a guide for selecting the PE width.

I've heard that the narrower the width (focused around the mode) the more broken paired reads but the more differences may be detected between similar bacteria. While the wider the difference fewer broken reads but less precision. Is there any truth in this??

Apologies for any misconceptions in advance.

Best wishes lg36