ClinProTools

Click here to visit our new ClinProt User Forum.

ClinProTools™ provides a powerful basis for mining potential biomarkers in complex protein profiles. Visualisation tools serve for the interactive inspection and comparison of large data sets originating from samples with different clinical diagnosis.

Additionally, the ClinProTools software supplies highly sophisticated mathematical algorithms for the discovery of complex biomarker pattern models. Besides validation, also class-prediction tools are integrated to allow implementation of the whole biomarker detection and evaluation process.

ClinProTools completes the ClinProt™ workflow, allowing easy access to mass spectrometry data for rapid and comprehensive evaluation to mine potential biomarkers in complex protein profiles.

ClinProTools supports a typical biomarker discovery project workflow:

Analysis of training data sets to generate the best biomarker pattern model
Cross validation and/or validation of models by an independent data set
Classification of patient sets for class prediction

Concept of ClinProTools

Visualization of large numbers of data sets with intuitive tools: Spectra view, virtual gel view, stack plots, contour plot, cross-section view, box-and-whiskers plot
Creation of predictive biomarker pattern models through multivariate bioinformatics tools by employing Genetic Algorithm and Support Vector Machine
Validation of the models by external samples with determination of specificity and sensitivity
Classification of unknown samples

Visualization of Data

Display of single and average spectra to allow inspection of data
A gel view display of all spectra at a time
Intuitive display functions
Visualization of peak info from each sample for best separating peaks

Model generation: Univariate and multivariate statistical tools

ClinProTools 2.0 offers both statistical tests for expected normally distributed data, as well as for data sets that are not normally distributed. As an output, the software creates a table of peaks that can be sorted according to lowest p-value. The p-value describes the significance and probability of single peaks in order to separate the classes. The output has a design which gives the user a maximum level of convenience together with high flexibility. All algorithms used for model generation support multiple classes.

Single peaks statistics with a quick univariate sorting algorithm

For the statistics of single peaks, CPT 2.0 uses the Quick Classifier algorithm (QC) which is a univariate sorting algorithm. Univariate analysis can be very useful for discovering statistically significant biomarker candidates.

Two powerful multivariate analysis tools are part of ClinProTools 2.0

1. Genetic Algorithm

Genetic Algorithms are inspired by the theory of evolution. These algorithms mimic the evolutionary process by finding the fittest solution from multiple models. Here, a model is defined as combination of peaks. These models undergo - in analogy to genetics - processes like chromosomal cross-over, mutation, and a selection of the fittest result. As a result, a new generation of models will be created which will again undergo selection. After multiple generations (the number can be defined in the software) the fitness will remain stable and the algorithm stops.

2. Support Vector Machine

The Support Vector Machine (SVM) is historically a classifier and not a feature selection algorithm. SVM tries to find a hyperplane that separates one or more classes. In the simplest case, the SVM helps to determine an optimal hyperplane separating two clouds of data. The algorithm tries to find this line in a multidimensional space.

Validation and class prediction

Validation of the models can be achieved by using an independent training data set. As a result, the user will get the sensitivity and specificity of the model as percentage of correctly classified data (Fig. 8). If only a limited set of data is available, cross validation (Fig. 8 B) may be used which means that one or more spectra are taken out of the training set and used for clustering.

: click for larger image

: click for larger image

: click for larger image

: click for larger image

Technical Data

Support of all Bruker spectra
Parameters definitions for peak picking
Automatic Normalization and recalibration of data sets
Multiple cluster analysis
sophisticated visualization tools
Univariate and multivariate statistical tools
Validation and class prediction
Peak-list export
Copy and paste for spectra export
Running on Windows 2000 and Windows XP
Support of multiple classes
New QuickClassifier (QC) univariate sorting algorithm
New Support Vector Macine (SVM) multiclass multivariate analysis
Correlation matrix to calculate positive or negative peak correlations

Now available: ClinProTools 2.1

The new version of our state-of-the-art biomarker profiling solution includes additional features for data analysis and visualisation. Especially, the software allows the researcher to merge statistical analysis on profiling spectra with molecular images of tissue distributions of biomarkers from the MALDI Molecular Imager™ system.

New features:

Supervised Neural Network™ algorithm (SNN)
Bruker Daltonics developed the SNN algorithm, an approach based on a widely accepted prototype based classifier with high generalization ability. It is specially suited for high dimensional multi modal data.
Principle Component Analysis (PCA)
The PCA is an unsupervised data analysis approach, which allows an easy visual inspection of the data distribution.
Receiver Operating Characteristics (ROC)
The per-peak ROC curve gives a visual overview of the class separating capability of single peaks acc. specificity and sensitivity.
Quality control functionality
Data quality control is supported by the help of the PCA and the 2D Peak Distribution views allowing to evaluate the data quality by analyzing class internal variations.
Working together with MALDI Molecular Imager™
In combination with the new flexImaging 2.0 software, ClinProTools 2.1 is the first commercially available bioinformatics package, which merges statistical analysis of MALDI-TOF mass spectrometry-based profiling data of samples from different classes with imaging of tissue distribution of peptide and protein biomarkers using the MALDI Molecular Imager™.