454 quality score, z-score,..

[nii]

Hey!
I read something about 454 sequencing, but there are some things I don't understand. (http://www.nature.com/nature/journal...ture03959.html)

Quality Score: Why is the quality score the probability, that a homopolymer of at least length n caused the signal, not exactly length n? And in the formula, the j-variable of the sum-symbol goes from n to infinity?

Z-Score: If we need a reference sequence to calculate the Z-Score, we can only use it when resequencing a genome, not when we are doing a de novo assembly. In the supplementary methods it says it is also applicalbe when doing the de novo sequencing, but how? Just align the reads? And what is aligned, the peaks of the flowgrams?
Furthermore, if the Z-Score is the difference between the averaged signal values of the different reads and the nearest integer, divided by the standard deviation of the read signals, wouldn't an averaged signal of e.g. 6.5 give rise to a higher Z-Score than an averaged signal of 6.9 when having the same standard deviation?? Or do I misinterpret the meaning of the Z-Score (confidence, that there is one particular base at a particular position)?

Flow order: There appear two PPi Kernels during sequencing. Are they used to measure the amount of enzyme-beads and fragments (to correct the signal strength for each well)?

Thanks,
david

[TSR]

454 quality score

Hi,

as far as I can tell, there must be an error in the quality score definition...

In the link you provide above (http://www.nature.com/nature/journal...ture03959.html), Margulies et al. (2005) states that the quality score is:
Q = -10 log10( P(≥n|s) )
where s is the observed signal and n is the length of the homopolymer that produced the signal.

That would mean that:
Q = 40 <=> P(≥n|s) = 0.0001
Q = 20 <=> P(≥n|s) = 0.01

So the probability that the homopolymer is of length ≥n is higher when the quality score is lower! That must be a mistake.

How should one interpret the quality scores?

Margulies (2005) writes:

"For instance, if two A’s are called for a particular signal, the quality score for the second A is given by the probability that the observed signal came from a homopolymer of length two or greater."

Here is some real 454 data:

Position: 1  2  3  4  5  6  7  8  9  10 11
Residue:  T  G  G  G  A  G  G  G  G  A  T 
Quality:  30 27 30 30 28 14 14 13 13 19 20

So in the data above, what does it mean (in terms of probabilities), that:
a) position 2 has lower quality than the rest of the homopolymer?
b) position 7 has higher quality than the rest of the homopolymer?

Anybody know how the quality score is calculated (eg. sourcecode)?

Cheers,
T

[zhengyuanwang]

454 quality score, no error

Their formula and coding are fine.
Reported is the probability to their ERROR Model.