Signal to noise ratio estimation in MS2 spectra
閲覧: 54 回
最初の未読メッセージにスキップ
Marc Vaudel
2017/02/16 4:05:542017/02/16
To: ProteomicsQA
Dear all,
I have been looking for quite some time for good methods to estimate the noise level in MS2 spectra obtained from peptides. Most search engines have their own internal intensity thresholds, modification localization scores also do this to some extend by optimizing their depth parameter on mass windows, but I am not aware of studies on this specifically. Would you know a piece of literature on the noise characterization and noise level detection in MS2 spectra derived from peptides with different instrumentation? What are your favorite methods?
Many thanks!
Marc
I have been looking for quite some time for good methods to estimate the noise level in MS2 spectra obtained from peptides. Most search engines have their own internal intensity thresholds, modification localization scores also do this to some extend by optimizing their depth parameter on mass windows, but I am not aware of studies on this specifically. Would you know a piece of literature on the noise characterization and noise level detection in MS2 spectra derived from peptides with different instrumentation? What are your favorite methods?
Many thanks!
Marc
Vladimir Gorshkov
2017/02/16 4:32:282017/02/16
To: ProteomicsQA
Hi Marc,
Unfortunately, I have no solid piece of literature on the noise cancellation to offer, but I can share my personal experience.
First, regarding the Thermo instruments that I mostly (> 99%) work with, the basic noise suppression is performed by the instrument itself. It is possible to get the noise levels for high-resolution instruments from the raw file, however, this information is discarded by all (to my knowledge) converters.
When I have to work with TOF-TOF data from ABSciex I tested out a couple of noise suppressions methods on it, like absolute or relative threshold, TopN out of 100 units, and fitting of two normal distributions to the intensity histogram. The later worked best for me, perhaps, since the there were at least several hundred of peaks in each spectrum.
Best regards,
Vladimir
Unfortunately, I have no solid piece of literature on the noise cancellation to offer, but I can share my personal experience.
First, regarding the Thermo instruments that I mostly (> 99%) work with, the basic noise suppression is performed by the instrument itself. It is possible to get the noise levels for high-resolution instruments from the raw file, however, this information is discarded by all (to my knowledge) converters.
When I have to work with TOF-TOF data from ABSciex I tested out a couple of noise suppressions methods on it, like absolute or relative threshold, TopN out of 100 units, and fitting of two normal distributions to the intensity histogram. The later worked best for me, perhaps, since the there were at least several hundred of peaks in each spectrum.
Best regards,
Vladimir
Marc Vaudel
2017/02/16 4:46:482017/02/16
To: ProteomicsQA
Hi again,
Thanks, that confirms my personal experience. The thing is that there is noise in terms of signal, all the small peaks kind of random, this gets easily sorted out by an intensity threshold, and chemical/fragmentation/co-elution noise that is more of an interference, and this one is hard to get rid of because not evenly distributed. Would be curious to know how people are dealing with these sort of things.
And apologies if my terminology is not adequate/rigorous :)
Marc
Thanks, that confirms my personal experience. The thing is that there is noise in terms of signal, all the small peaks kind of random, this gets easily sorted out by an intensity threshold, and chemical/fragmentation/co-elution noise that is more of an interference, and this one is hard to get rid of because not evenly distributed. Would be curious to know how people are dealing with these sort of things.
And apologies if my terminology is not adequate/rigorous :)
Marc
Vladimir Gorshkov
2017/02/16 5:16:542017/02/16
To: ProteomicsQA
Hi Marc,
Excuse the self-advertisement of the complementary ions matching approach I put some time into.
https://www.ncbi.nlm.nih.gov/pubmed/23725413
https://www.ncbi.nlm.nih.gov/pubmed/25978296
https://www.ncbi.nlm.nih.gov/pubmed/27329701
We mostly presented this as a way to go deeper in the proteome, but it, certainly, can be used as a quality filter as well.
Additionally, I can think of something similar to DIAUmpire https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4399776/, that can also be used to identify fragments behaving similarly and, thus, fishing out relevant informaton.
Best regards,
Vladimir
Excuse the self-advertisement of the complementary ions matching approach I put some time into.
https://www.ncbi.nlm.nih.gov/pubmed/23725413
https://www.ncbi.nlm.nih.gov/pubmed/25978296
https://www.ncbi.nlm.nih.gov/pubmed/27329701
We mostly presented this as a way to go deeper in the proteome, but it, certainly, can be used as a quality filter as well.
Additionally, I can think of something similar to DIAUmpire https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4399776/, that can also be used to identify fragments behaving similarly and, thus, fishing out relevant informaton.
Best regards,
Vladimir
Marc Vaudel
2017/02/16 5:18:342017/02/16
To: ProteomicsQA
That is very informative, many thanks!
Marc
sander....@ugent.be
2017/02/24 3:23:202017/02/24
To: ProteomicsQA
We use Threshold Inspector of Waters to set optimal noise thresholds:
While probably not completely applicable to your problem, a variant of the principle might be useful or at least give some inspiration: Randomly select a small amount of spectra, create results with various noise thresholds, extrapolate and apply optimal threshold to all spectra.
Op donderdag 16 februari 2017 11:18:34 UTC+1 schreef Marc Vaudel:
Op donderdag 16 februari 2017 11:18:34 UTC+1 schreef Marc Vaudel:
全員に返信
投稿者に返信
転送
新着メール 0 件