GeneLinker Feature List

GeneLinker Gold 4.6

Data Import

• Flexible, extensible and intuitive data import supporting averaging of within-chip replicates and Affymetrix P-values
• Support for all types of text files, 2 color data and a variety of native chip formats
• Currently there are 20 supported formats including: Affymetrix 4.0 and 5.0, GenePix, Genomic Solutions, CodeLink XML, Quantarray and ScanArray.
• Binary Excel-format (.xls) tabular data.
• Quantarray merge replicates and Quantarray Unicode support.
• NEW! Improved Proteomics Import Scripts for use with the new NIH/NCI Center for Cancer Research Ovarian dataset 8-7-02 (see Protein Biomarker Discovery with GeneLinker Platinum for more information).
• NEW! “Import P-Value Spectrum” script creates a dataset from the p-values and false-discovery rate. This allows exported p-values from ANOVA experiments to be re-imported, enabling novel visualization of channels that best distinguish between different classes (see Protein Biomarker Discovery with GeneLinker Platinum for more information).
• NEW! Import scripts to handle Koadarray ratio and two-color data (including confidence values)
• Additional formats can be added at any time

Pre-processing

• Value removal by reliability measures and using simple expressions
• Missing Value Estimation by central tendency, arbitrary value or nearest neighbors
• Create variables easily within the application. Random variables can be created to assist with statistical analysis
• Sample Filtering (based on variables)

Normalization Methods

• Log (base 2, e and 10), Divide by Maximum, Min-Max Normalization, Standardization
• Central tendency and Linear Regression
• Lowess for 2-color data
• Using Positive and Negative Control Genes

Filtering Methods

• Gene lists
• Maximum culling and Range culling
• N-fold with N
• N-fold with specified number of genes
• Spotted array n-fold culling

Statistics

• Summary statistics (histograms)
• F-Test ANOVA for normally distributed data
• Kruskal-Wallis non-parametric ANOVA
• False Discovery Rate calculations
• Bonferroni corrected p-values
• Replicate merging based on an imported variable (allows multiple categorizations of samples)

Analysis

• Partitional clustering of genes and/or samples using Self-Organizing Maps (SOMs), K-Means and Mutual Nearest Neighbors (Jarvis-Patrick) methods
• Hierarchical clustering of genes and/or samples using single, average, or complete linkage
• Principal Component Analysis (PCA) with 8 different visualizations
• A variety of distance metrics including Euclidean, Euclidean Squared, Manhattan, Pearson Correlation, Pearson Squared and Spearman Rank Correlation
• Profile matching using selected gene, averages of genes, or custom genes

Visualization

• Color matrix plot (array intensity) visualization and dendrogram with selectable nodes, user-definable data range and color schemes.
• Two-way cluster plots enabling simultaneous viewing of gene and sample clustering using partitional and hierarchical cluster experiments (or both)
• Scatter plots, centroid plots, cluster plots
• PNG, PDF and SVG graphics export
• Linked visualizations to highlight selected items across plots
• Coloring by variables and gene lists
• Default visualizations for different experiment types

Database

• User-definable external links to data sources such as GenBank, Unigene, and NetAffx
• MySQL data repository with no size limits
• Includes configuration applications for using IBM DB2 or Oracle 9i
• Support for multiple repositories to make it easier to share copies of GeneLinker among researchers or manage multiple large projects

Scripting

• Powerful scripting and meta-scripting capabilities that allow easy repetition and sharing of workflows

Other

• Gene lists for support of pathways, functional classifications and ontologies
• Variables for replicate identification, or identification of sample classes
• Accurate progress meters, experiment canceling
• Navigator pane that provides a hierarchical history of all analysis procedures and captures all parameter settings
• HTML Experiment and Workflow reports.
• Gene, sample and experiment annotations
• Find functionality for genes allowing fast selection of desired gene
• Ability to export data sets and PCA loadings
• Export to Spotfire's DecisionSite

GeneLinker Platinum 4.6

• Support for external scripting (XScripts) to allow users to integrate their own proprietary analysis techniques into GeneLinker.
• IBIS(tm) (Integrated Bayesian Inference System) is a powerful tool for identifying individual genes and pairs of genes that differ significantly between classes, taking into account the actual distribution of values within your data. It can identify non-linear and combinatorial patterns of gene expression that characterize different toxicity responses, disease states, or treatment outcomes. Furthermore, it can be used to build classifiers that can identify these patterns in new samples.
• SLAM(tm) (Sub-Linear Association Mining) is a patented algorithm licensed by Predictive Patterns that is used to find correlations between discretized variables or to predict the outcome of a categorical variable. As an aid to supervised learning, SLAM? is used to find associations in gene expression data so that a list of interesting genes (features) can be created.
• Committee of Neural Networks allow you to build classifiers that can predict toxicity responses, disease states, or treatment outcomes. The committee architecture is much more robust than a single Neural Network. You also have control over the following parameters:
  • The number of learners in the committee
  • The learner votes required for classification
  • The number of hidden units
  • The type of Conjugate Gradient used
  • Whether to use the Pocket Algorithm or not
  • The Neural Network Stopping Criteria
• Committee of Support Vector Machines, which complements our Committee of Neural Networks with a powerful non-linear classification algorithm which is particularly well-suited to datasets with relatively small sample sizes. You have control over the following parameters:
  • The number of learners in the committee
  • The learner votes required for classification
  • The kernel type (linear, polynomial or radial basis function)
  • The kernel function parameters (degree, gamma, coefficient)
  • The cost

These features are most valuable to users with some knowledge of machine learning but they are accessible to anyone. For example, the Committee of Neural Networks sets sensible defaults for your data, and often yields good performance without adjusting these default values.

Minimum System Requirements

If you have any other questions, please contact: