Spectral counting in Scaffold 2.0

[Adam Byron]

Hi

I'm currently exploring the new quantification features of Scaffold
2.0, which look really interesting. I have a few questions about the
spectral counts that are generated. Firstly, how are the values for
"counts" defined, since they don't appear to be numbers of assigned
spectra per se? Secondly, what are the normalized spectral counts
normalized to exactly? Is it total counts in each biological sample?
And how do the normalized counts differ from the percentage of total
spectra feature that was available in 1.7 (which, by my understanding,
was effectively normalized to the total spectra in each sample)?

Any details would be much appreciated.

Thanks in advance
Adam

[Mark Pitman]

Hi Adam,

There is a white paper explaining this very thing, but it's on another computer. So, I'm cc'ing myself, and I'll post it asap.

Cheers,

Mark

--
Mark Pitman
mzpi...@gmail.com

[mpitman]

Finally found the informattion about the spectral counts, and how
they're assigned.

Our protein normalizing entails averaging the spectrum counts for all
of the biosamples and then multiplying the spectrum counts in each
sample by the average divided by the individual sample's sum.

For ex:

Two samples: A and B

Each sample has three proteins: One Two and Three

Protein One has 12 spectra in sample A and 8 spectra in sample B

Protein Two has 6 spectra in sample A and 3 spectra in sample B

Protein Three has 4 spectra in sample A and 3 spectra in sample B

Scaffold would normalize these spectra as follows:

Sample A has 22 spectra
Sample B has 14 spectra

Average is 18

Scaffold would multiply all of the protein spectral counts in Sample A
by (18/22), and all of the spectral counts in Sample B by (18/14).

For ex: Protein One in sample A would be 12*(18/22). Protein One in
sample B would be 8*(18/14).

- We basically tried to keep the normalization as simple as possible,
yet still be effective.

If you have questions please send them our way.

- Mark P.

[Jon H Man Uni]

Thanks Mark,
Can i ask how this is normalisation is better than the simple %
spectra?
Jon H

> > On Feb 6, 2008 4:37 AM, Adam Byron <adam.by...@gmail.com> wrote:
>
> > > Hi
>
> > > I'm currently exploring the new quantification features of Scaffold
> > > 2.0, which look really interesting. I have a few questions about the
> > > spectral counts that are generated. Firstly, how are the values for
> > > "counts" defined, since they don't appear to be numbers of assigned
> > > spectra per se? Secondly, what are the normalized spectral counts
> > > normalized to exactly? Is it total counts in each biological sample?
> > > And how do the normalized counts differ from the percentage of total
> > > spectra feature that was available in 1.7 (which, by my understanding,
> > > was effectively normalized to the total spectra in each sample)?
>
> > > Any details would be much appreciated.
>
> > > Thanks in advance
> > > Adam
>
> > --
> > Mark Pitman

> > mzpit...@gmail.com

[Jason]

Hello Jon,

In Scaffold the normalization counts all possible spectra who could be
assigned to a protein (the "unweighted" spectra), while the straight %
only counts those spectra that protein prophet actually assigns to a
protein. A good example of unweighted spectra would be if two proteins
(protein A and protein B) shared a spectrum (spectrum #5).
-protein A has spectra #1, #3, and #5
-protein B has spectra #2, #4, #5, and #6

Protein prophet will assign spectra #5 to protein B. So based on
spectral assignments protein B has 4 spectra and protein A has 2.
However, the unweighted spectral counts for these two proteins are:
protein A = 3 spectra and protein B =4 spectra. Spectra #5 is counted
twice. Scaffold normalizes off the unweighted spectral counts, while
straight % is off the assigned spectral counts (as mentioned above).
This is the main difference between the two numbers in Scaffold.

Jason


On Mar 25, 9:50 am, Jon H Man Uni <j.d.humphr...@manchester.ac.uk>
wrote: