This is the Matlab code related to the research project published in the paper "3DSpectra: A 3-dimensional quantification algorithm for LC–MS labeled profile data": http://www.sciencedirect.com/science/article/pii/S1874391914004163
Mass spectrometry-based proteomics can generate highly informative datasets, as profile three-dimensional (3D) LC–MS data: LC–MS separates peptides in two dimensions (time, m/z) minimizing their overlap, and profile acquisition enhances quantification. To exploit both data features, we developed 3DSpectra, a 3D approach embedding a statistical method for peptide border recognition.
3DSpectra efficiently accesses profile data by means of mzRTree, and makes use of a priori metadata, provided by search engines, to quantify the identified peptides. An isotopic distribution model, shaped by a bivariate Gaussian Mixture Model (GMM), which includes a noise component, is fitted to the peptide peaks using the expectation–maximization (EM) approach. The EM starting parameters, i.e., the centers and shapes of the Gaussians, are retrieved from the metadata. The borders of the peaks are delimited by the GMM iso-density curves, and noisy or outlying data are discarded from subsequent analysis.
The 3DSpectra program was compared to ASAPRatio for a controlled mixture of Isotope-Coded Protein Labels (ICPL) labeled proteins, which were mixed at predefined ratios and acquired in enhanced profile mode, in triplicate. The 3DSpectra software showed significantly higher linearity, quantification accuracy, and precision than did ASAPRatio in this real use case simulation where the true ratios are known, and it also achieved wider peptide coverage and dynamic range.
These instructions are meant to enable the user to reproduce submitted results
1.- preliminary step: download mzXML data files from:
https://maspectras.genome.tugraz.at/maspectras/
login as guest (pwd: guest). On the up left corner click on Menu and unfold the MASPECTRAS 1 Data folder and then the AutoQuantificationProfile sub-folder. Download all data in there and save them in a single folder on your file system.
2.- start a Matlab session and type in the command window: open classpath.txt
add at the end of the classpath.txt file your local file paths to the needed jars, as in the example here below:
/Users/saran/work/3DSpectra/jrap_StAX_v5.jar
/Users/saran/work/3DSpectra/maspectras.jar
/Users/saran/work/3DSpectra/massMatlab.jar
/Users/saran/work/3DSpectra/mzRTree.jar
/Users/saran/work/3DSpectra/igbbiojava.jar
/Users/saran/work/3DSpectra/outliers.jar
3.- mex update on new hardware (hist2data) as follows: mex hist2data.c
4.- check additional needed toolboxes (e.g., Bioinformatics, Statistics) are available typing in the command window: ver
5.- cd to 3DSpectra folder
6.- (not needed for evaluation data) create the needed peptide libraries following the instructions given in the readme.txt provided in the ID_files folder
7.- cd to 3DSpectra folder once more if needed
8.- run the 3DSpectra algorithm by the main.m.
In this main execution mzRTree for each data file will be created running the script mzRTree_utils.m and data analysis by means of 3DSpectra executed.
If you run into problems with 3DSpectra uncomment rethrow(ME) in the catch statement to check out the occurring error.
Finally, result evaluation is executed as well.
9.- analysis can take some hours, depending on the machine (e.g., ~3 hours on a 2.2 GHz Intel Core i7 with 8 GB 1333 MHz DDR3)
10.- run the main_results.m after cd-ing to the results folder to evaluate results and compare them to MASPECTRAS ones (not needed if main.m is run)
Please notice that in this implementation the following data feature was exploited: sample resolution is constant both along retention times (~1sec) and m/z (0.04Da). This allowed to exchange retention time (m/z) indexes and their actual retention time (m/z) values. Other data, without constant sampling rates, would need slight modifications to the current implementation. Next release will add this additional feature.
LC/MS data analysis requires extended amounts of memory from the operating system. If you receive errors related to memory, try the following:
- Increase the virtual memory (swap space) for your operating system (with a recommended initial size of 3,069 and a maximum size of 16,368) as described in Memory Usage.
- Set the 3 GB switch (32-bit Windows XP only) as described in Memory Usage.
- If you receive errors related to Java heap space, increase your Java heap space:
- If you have MATLAB version 7.10 (R2010a) or later, see Java Heap Memory Preferences
- If you have MATLAB version 7.9 (R2009b) or earlier, see http://www.mathworks.com/support/solutions/data/1-18I2C.html