GeenaR: R based Spectra Analysis

Geena

Build June 18, 2019.

Home - Upload - Help

This system is under active development,
You are welcome to send us your comments.

GeenaR: A Web Tool for Reproducible MALDI-TOF Analysis

Welcome to GeenaR, a tool for MALDI/ToF MS spectra analysis.

Using GeenaR should be straightforward.
Nevertheless, should you have any problem, see the help page and the page on information on format
See also the help page on how to run the test on the example dataset.

NB! Move your mouse over the icon This is the contextual help icon. Move the cursor on it for a short hint on the field. for a contextual help.
Fields marked with a ^* are mandatory!

YOU MUST UPLOAD YOUR SPECTRA BEFORE RUNNING THE ANALYSIS. GO TO THE UPLOAD PAGE NOW IF YOU DID NOT DO IT YET.

Job information

Reference name for your analysis, you may change it. It is included within the name of result files. ^*Job name:

Data set name specified when uploading the spectra. May only include letters (both upper- and lowercases), numbers and the following characters: dash, underscore, dot. Avoid spaces! ^*Data set name:

Provide a valid email address, results will be sent by email. This must correspond to the email address associated with the data set when uploading. ^*Email:

Your country, for statistics only Country:

Input data

You must select a properly formatted file from your computer. This file will be uploaded to the server but removed as soon as it is used. ^*Target file:

Mass spectra can be trimmed both at the beginning and at the end of the m/z range. Here you can specify the lower and higher m/z values for the desired analysis range. Insert 0 in both fields if you do not want any trimming. Trimming: From: m/z

To: m/z

Mass spectra can be checked for the presence of outliers (atipicality score) and their formal correctness. When required by selecting this checkbox, quality control is carried out twice, before and after trimming. Quality control:           If you select this checkbox, a report, including all results, as well as the list of methods and parameters selected, will be made available at the end of the analysis. The report is created in HTML, with and without R code. Reporting: With code:

You must select a properly formatted file from your computer. This file will be uploaded to the server but removed as soon as it is used. Attributes file:



Steps, methods and parameters

Step (select to execute)
Select/unselect all steps Method
Parameters

Pre-processing of spectra

This step overcomes the dependency between variance and mean by transforming intensities through a given function. Variance stabilization Select one among the following methods as transformation function: square root, logarithm to base e, logarithm to base 2, logarithm to base 10). SQRT LOG LOG2 LOG10

This step smooths the mass spectra by using a filtering methods. Smoothing Select one among available methods: Savitzky-Golay (Savitzky, Golay. Analytical chemistry 1964) or moving average. Savitsky-Golay Moving average Specify the size of the half window (hw). It should be smaller when using the moving average. The full size of the window will be (2*hw + 1). Half window size:

This step makes an estimation of the noise and tries to reduce it. The estimation needs the number of iterations for the methods. Baseline removal Select one among available methods: Statistics-sensitive Non-linear Iterative Peak-clipping (SNIP) (Ryan et al. Nuclear Instruments and Methods in Physics Research Section B 1988), fast algorithm on kernels (TopHat) (van Herk. Pattern Recognition Letters 1992), convex hull (Andrew. Processing Letters 1979), and median. SNIP TopHat Convex-Hull Median Specify the number of iterations for the SNIP method. Number of iterations:    Specify the desired half-window size (applies to TopHat only). Half window size:

This step calibrates the intensity of the mass spectra, in order to equalize possible small batch effects. The choice on the right is about the following methods: Total Ion Current (TIC), Probabilistic Quotient Normalization (PQN), and median. Normalization Select one among the available methods: Total Ion Current (TIC), Probabilistic Quotient Normalization (PQN) (Dieterle et al Analytical Chemistry 2006), and median. TIC PQN Median

This step can be selected in presence of different technical replicates for each sample, and provides one averaged mass spectrum per sample. Averaging Select one among the following averaging methods for each m/z value: mean, median or sum of intensities. Mean Median Sum

This step has two different goals: estimating the noise and correcting the phase. A method must be defined for both goals. Alignment Select one amomg available methods for noise estimation: Median Absolute Deviation (MAD), or Friedman’s Super Smoother (Friedman. Stanford University Technical Report No. 5. 1984). Noise estimation: MAD Super Smoother Specify the desired Signal-to-Noise Ratio (SNR) and half-window size for noise estimation. Half window size:   SNR:

Select one of the following methods for phase correction: Local Weight Scatterplot Smoothing (LOWESS) (Cleveland. Journal of the American Statistical Association 1979), linear, quadratic or cubic warping methods. Phase correction: Lowess Linear Quadratic Cubic Specify the tolerance, i.e. the maximum relative deviation of a peak position to be considered as identical to another. Tolerance:

Peak identification, extraction and selection

This step allows to identifiy all peaks, i.e. local maxima above a user defined threshold, and to remove infrequent or sporadic ones. Detection Select one among the available methods for binning: strict, that creates bins never containing two or more peaks of the same spectrum, or relaxed, allowing multiple peaks of the same sample in one bin. Binning: Strict Relaxed

For the selection of peaks to be analysed by following steps, a percentage must be defined. Selection Specify the coverage (minimum frequency) of peaks for their selection (specify 0 for selecting all peakes, 0.5 for selectings peaks present in at least half spectra, etc...). Coverage:

Clustering and visualization

This step clusters peaks selected by the previous step. A similarity matrix is created using the cosine correlation method on the peaks matrix (feature matrix). Clustering Select one among the available methods for the link function: average, complete, ward, and median. Link function: Average Complete Ward Median

K estimation: Gap Statistic Silhouette Manually given Specify the number of clusters. Only used if the 'manually given' method is selected. K value:

This step create a heatmap for the peaks matrix. Heatmap Specify if the heatmap must be reordered along its dimensions by similarity: no, by row (samples), by column (signals), both by row and column. Clustering: None Samples Signals Both

This step carries out a Principal Components Analysis (PCA) of the peak matrix. Three PCA plots are generated, comparing the first three principal components pairwise. Groups are highlighted if defined in the target file. Principal Component Analysis (PCA)



For information, get in touch with:
Paolo Romano, Bioinformatics, IRCCS Ospedale Policlinico San Martino,
Email to Paolo.Romano@HSanMartino.it
Email to gabriela.coronel@HSanMartino.it