Extracts and normalizes isobaric labeling information from an LC-MS/MS experiment.
pot. predecessor tools | ![]() ![]() | pot. successor tools |
PeakPickerHiRes | IDMapper | |
FileFilter |
This tool currently supports iTRAQ 4-plex and 8-plex, and TMT 6-plex and 10-plex as labeling methods. It extracts the isobaric reporter ion intensities from centroided MS2 data, performs isotope correction and stores the resulting quantitation in a consensus map, in which each consensus feature represents one relevant MS2 scan (e.g. HCD; see parameters select_activation
and min_precursor_intensity
). The position (RT, m/z) of the consensus centroid is the precursor position; the sub-elements correspond to the channels (with m/z values of 113-121 for iTRAQ and 126-131 for TMT, respectively).
The input MS2 spectra have to be in centroid mode for the tool to work properly. Use e.g. PeakPickerHiRes to perform centroiding of profile data, if necessary.
Isotope correction is done using non-negative least squares (NNLS), i.e.:
Minimize ||Ax - b||, subject to x >= 0, where b is the vector of observed reporter intensities (with "contaminating" isotope species), A is a correction matrix (as supplied by the manufacturer of the labeling kit) and x is the desired vector of corrected (real) reporter intensities. Other software tools solve this problem by using an inverse matrix multiplication, but this can yield entries in x which are negative. In a real sample, this solution cannot possibly be true, so usually negative values (= negative reporter intensities) are set to zero. However, a negative result usually means that noise was not properly accounted for in the calculation. We thus use NNLS to get a non-negative solution, without the need to truncate negative values. In the (usual) case that inverse matrix multiplication yields only positive values, our NNLS will give the exact same optimal solution.
The correction matrices can be found (and changed) in the INI file (parameter correction_matrix
of the corresponding labeling method). However, these matrices for both 4-plex and 8-plex iTRAQ are now stable, and every kit delivered should have the same isotope correction values. Thus, there should be no need to change them, but feel free to compare the values in the INI file with your kit's certificate. For TMT (6-plex and 10-plex) the values have to be adapted for each kit.
After the quantitation, you may want to annotate the consensus features with corresponding peptide identifications, obtained from an identification pipeline. Use IDMapper to perform the annotation, but make sure to set suitably small RT and m/z tolerances for the mapping, since the identifications will come from the very same MS2 scans that are now represented by consensus features. In general it should be possible to achieve a perfect one-to-one matching of every identification to a single consensus feature.
Note that quantification will be solely on peptide level after this stage. In order to obtain protein quantities, you can use TextExporter to obtain a simple text format which you can feed to other software tools (e.g., R), or you can apply ProteinQuantifier.
The command line parameters of this tool are:
INI file documentation of this tool:
OpenMS / TOPP release 2.0.0 | Documentation generated on Wed Jan 27 2016 17:15:21 using doxygen 1.8.10 |