454 homopolymer errors or????

[ian Adams]

I have some 454 data produced from cDNA randomly amplified using proofreading taq, sheared and then sequenced. Once the amplification tags were removed, we'ved assembled it using CLC genomic work bench. Looking at the protein translations of two particular genes we have very good homology to genes from related organisms but in two places we appear to have frame shifts in the DNA sequences which introduce stop codons leading to incomplete proteins. Correcting the frameshifts would be CCC to CCCC in one case and CCCC to CCC in the other and completely restore the protein homology. Both regions have over 200 fold coverage reading the frameshift with only one read giving the "corrected" unshifted sequence. I have read that the 454 does have problems with homopolymers. What I am wondering is if this is biological (a mutation leading to an unusual terminated protein--I think unlikley these proteins are essential) a sequencing error (introduced by the 454 / emPCR) or sample prep (PCR error-but I was using proofreading taq). Does anyone have any experience of this sort of thing and any comments on the "introduced by the sequencing / emPCR option"?
We are going to Sanger sequence across these regions and see what we get but that will take time.

thx

Ian

[timread]

Ian -

IMHO, when you have a ratio of 200 read to 1 it is very unlikely that you are seeing homoploymer miscalls. Typically this occurs when the base calling software has difficulty accurately setting the threshold between say 3C and 4C for a particular extension and you will see both reads well represented. I'd go for a PCR problem, or more interestingly, a real frameshift. Verify the result with a Sanger read.

tim

[glacerda]

I have compared a segment of 100K that we have sequenced by sanger (at > 20X) and by 454 ( > 50X)

Even at 50X, 454 made indel errors (all of them in homopolymers greater than 5 bp)

in you case, you have a 4bp homopolymer in which an indel error would be very very rare. I also thin this is a pcr issue or a real frameshift.

[vasvale]

below is a reference for the PCR error, can somebody eplain why homopolymers error are seen with Roche and not with other platforms (as I heard)

Clarke LA, Rebelo CS, Gonçalves J, Boavida MG, Jordan P.
PCR amplification introduces errors into mononucleotide and dinucleotide repeat
sequences.
Mol Pathol. 2001 Oct;54(5):351-3.
PMID: 11577179 [PubMed - indexed for MEDLINE]

[bioinfosm]

As I understand..
454 does not have reverse end-blockers to nucleotides and attach multiple bases at once, using the total intensity to determine how many bases may have been incorporated.

Others use blockers to make sure only one nucleotide goes at a time, and make it a simple yes/no question with less chance for error

[Alex Clop]

454 sequences by Pyrosequencing. As defined by Nature Reviews Genetics Glossary, Pyrosequencing is A DNA sequencing technique that relies on detection of pyrophosphate release on nucleotide incorporation rather than chain termination with dideoxynucleotides.

This chemistry i) incorporates a known nucleotide to the sequencing reaction at a time, ii) then eliminates the remaining nucleotides and iii) then incorporates another known nucleotide again. It does this for the 4 nucleotides sequentially and then the cycle is repeated again. Each time a nucleotide is incorporated at the elongating chain, a luminiscent reaction (luciferase) will occur and detected by a CCD camera. Thus, and will permit the basecalling and the sequence "reading".

Regarding your question, the intensity of the luciferase ("I") activity is proportional to the number of - identical - nucleotides incorporated ("N") in the synthesized chain. However, if "N" is too high (I don't think there is a well defined threshold), "I" will reach saturation which will lead to wrong "quantification" of the number of contiguous A, T, C, or G that have been incorporated.

It is not simple to explain but it may help.

[timread]

The thresholds are calculated for each run by the analysis software based on the signal distribution. If you look at the 454BaseCallerThresholds.txt file in the Analysis (D_) directory you can see some stats for the run. e.g.

distributionPeaks
{
signalPeak = 0, 0.12, Found;
signalPeak = 1, 0.98, Found;
signalPeak = 2, 2.00, Found;
signalPeak = 3, 2.98, Found;
signalPeak = 4, 3.92, Found;
signalPeak = 5, 4.84, Found;
signalPeak = 6, 5.68, Found;
}

thresholdsUsed
{
threshold = 0, 1, 0.68;
threshold = 1, 2, 1.58;
threshold = 2, 3, 2.52;
threshold = 3, 4, 3.48;
threshold = 4, 5, 4.44;
threshold = 5, 6, 5.32;

interpolationAmount = 0.94
}


For a good description of 454 errors see this manuscript ..

http://genomebiology.com/2007/8/7/R143

[Layla]

Dear 454 data analysers...
We have data generated from 454 using nimblegens capture array experiment. I have read that 454 has issues with AT rich genomes and homopolymers. Having looked at the HCdiffs file I was going to focus on variances with Quality scores > 40, but I then read in a paper that stated 64% of errors were assigned a Newbler QS of 64! Does anyone know how to calculate the error rate from 454 data using the files created from gsMapper or if this error rate can be seen in any of the generated files? Ideally I would like to see how many of the variances seen are "real".

New to sequence data analysis!
Thankyou
Layla

[timread]

Dear 454 data analysers...
... but I then read in a paper that stated 64% of errors were assigned a Newbler QS of 64!

Layla

Layla - can you give the reference for this statement? Thanks in advance,

Tim

[Layla]

Hi Tim,

It was a presentation I came across on the web by Stephan Trong on microbial genomes..viewed by downloading as a pdf. I must say this error was found using next gen sequencing on microbial genomes.

Layla