LOWESS/Spline normalization

Program download

Main program is downloadable from here. Normalization program.

Data matrix for the demonstration is downloadable from here (please keep this format!). Sample.


Recently, we separated the normalization part of MS-DIAL and MRMPROBS programs. The powerful mathematics especially for cohort studies is essential to normalize the drifts of MS-signal intensities. The basic concept of this method is to use the intensities over an analytical run of 'quality control (QC)' sample which is a mixture of all biological samples.

Briefly, measurement data of QC samples are smoothed by the LOWESS of the single-degree least-squares. And then, the coefficient values between QC samples are interpolated by the cubic spline. Lastly, the entire datases is aligned to the spline result.

Please cite one of the following papers

  • MS-DIAL: data independent MS/MS deconvolution for comprehensive metabolome analysis. Nature Methods, 12, 523-526, 2015 [PubMed]
  • MRMPROBS Suite for metablomics using large-scale MRM assays. Bioinformatics, 30, 2379-2380, 2014 [PubMed]

Experimental design

The QC samples should be periodically (5-8 samples each) analyzed throughout an analytical run as shown the below figure.

How to use

1. Text (tab-delimited) data table should be prepared. The first, second, and third columns should contain the sample name, QC or Sample by TRUE or FALSE, and injection order, respectively. Then, from the fourth column, each cell should include the intensities of each metabolite.

2. Open 'LowessNormalizationSample.exe'

3. Click 'Browse' and select your data matrix.

4. Click 'Load'. You will find your QC number and the minimum span number for LOWESS smoothing if you can correctly load the data matrix.

5. If you click 'Span opt.', you can see the optimal 'Span' value which is coming from 'seven-fold cross validation' in the QC samples.

6. Click 'Export', then you will find the normalized result in the same directly of your data.

Back to Top