I have a environmental sample which has been subject to two diiferent extraction procedures. The resulting proteins have been digested and run through a mass spectrometer giving results (through Scaffold) which give the identity of the protein (sequence, GO number) and the number of times the spectra have been seen . I want to see if there are any differences between the extracts in terms of protein families (annotated from GO terms using RAST system ).
In some results I have many triplicate measurements of proteins between the two data sets , maybe 10 or 20 . in other data sets there may be only 1 or 2 representatives of the familiy leaving the statistics very slim . Are there common ways proteomics researchers use to analyse such data ?
Thanks Laurent
It's not that the spectra measurements are unbalanced , there are just more measurements in some catagories than others . For example if our catagory is calcium ion binding we have 4 different identified proteins that contain calcium in their GO terminology.
i.e.
PROTEIN NAME NO OF SPECTRA a X Y Z NO OF SPECTRA b X Y Z
calcium ion binding proteinA spectra methodA 3 2 1 spectramethodB 6 5 6
calcium ion binding proteinB spectra methodA 2 2 1 spectramethodB 3 5 6
calcium ion binding proteinC spectra methodA 3 2 1 spectramethodB 6 5 6
calcium ion binding proteinD spectra methodA 2 1 1 spectramethodB 6 5 8
calcium ion binding proteinE spectra methodA 1 2 1 spectramethodB 6 7 6
calcium ion binding proteinF spectra methodA 1 2 1 spectramethodB 4 5 6
then for proteins with iron binding we have
iron binding proteinA spectra methodA 2 1 1 spectramethodB 6 5 6
iron binding proteinB spectra methodA 1 2 1 spectramethodB 3 5 6
and then for proteins with NAD we have
NAD proteinA spectra methodA 2 2 1 spectramethodB 1 1 1
NAD proteinB spectra methodA 4 2 1 spectramethodB 3 5 6
NAD proteinC spectra methodA 2 0 1 spectramethodB 2 5 6
NAD proteinD spectra methodA 1 2 1 spectramethodB 6 5 3
NAD proteinE spectra methodA 3 1 1 spectramethodB 1 5 6
NAD proteinF spectra methodA 2 2 1 spectramethodB 8 5 6
NAD proteinG spectra methodA 2 1 1 spectramethodB 6 4 6
NAD proteinH spectra methodA 4 2 1 spectramethodB 9 5 6
NAD protein I spectra methodA 1 2 1 spectramethodB 6 6 6
so there are a variable number of proteins found in each catagory ,
I am trying to test if one method of extraction is fundimentally different from another method. I have graphed the means of the total numbers with means and standard error bars and they look different. However i need solid estimates of significance to back this up.
The specrtra came from an orbi trap. MS/MS data first loaded into scaffold viewer . There are quite a lot of proteins that are not identified in NCBI database because it's an environmental sample.
You might want to try random forest multivariate procedures that will allow you to build an algorithm with predictions of which proteins are associated with disease versus healthy states. Try this link for a good presentation on the topic:
https://stat.ethz.ch/education/semesters/ss2012/ams/slides/v10.2.pdf
Thanks Vicki , that looks very helpful. I will see how this will integrate with my data.
It should help narrow the field quite a bit with good sensitivity when it comes to identifying and parsing out the protein that is most predictive. When I saw the total number of spectra, I immediately thought of "The Trees and Random Forest"
Cheers!
- Badges
- Science topic
- Similar topics
- Bioinformatics
- Bioinformatics and Computational Biology