Introduction to the analysis of microarray data by GeneCluster

1
Introduction to the analysis of microarray data
by GeneCluster

• Guide for the first use of the program

2
(No Transcript)
3
GeneCluster Capabilities

• Dataset Preprocessing/Filtering
• Data Analysis
• Marker Selection/Neighborhood Analysis
• Find Classes of Similarly Expressed Genes
• Build Predictor Sets of Genes

4
Dataset Format

• .GCT or .RES files are used for (already
  normalized) gene expression data
• .GCT files are tab delimited files of expression
  values with simple headers
• .RES files are the same, but also include
  Presence calls
• .GCT files can be automatically created in dChip
• .CLS files are used for class data
• Separate .CLS files are required for every
  covariate

5
Example .GCT file
6
Example .CLS files
7
Dataset Preprocessing/Filtering

• Clip values to Min or Max values
• Keep if Max/Min gt ratio threshold
• Keep if Max-Min gt difference threshold
• Normalize to mean 0, variance 1
• Scale by minimum value
• Shift global/row mins
• Log10 transform

8
Marker Selection/Neighborhood Analysis

• For a set of classification markers in the .cls
  file, GeneCluster will look for the genes that
  most highly correlate with the classification.
• Possible metrics include t-tests and signal to
  noise ratios.
• Permutation tests quantify significance of the
  selected genes.

9
Top 5 over- and under-expressed genes correlating
with MMC status
10
Find Classes of Similarly Expressed Genes

• Use k-means clustering or self-organizing map
  (SOM)
• User can specify exactly how many clusters or let
  GeneCluster decide
• Genes with similar expression patterns over the
  treatment conditions will cluster together

11
SOM of filtered Fanconi data
12
Build Predictor Sets of Genes

• GeneCluster can be used for tumor
  subclassification using supervised learning
• (Golub et al 1999,Slonim et al 2000)
• GeneCluster selects predictor genes of clinical
  status using a training data set, and can be
  tested by predicting clinical status for a test
  data set

13
GeneCluster Resources

• http//www-genome.wi.mit.edu/cancer/software/genec
  luster2/gc2.html
• Tutorial and Instruction Manual in MACF, Smith
  1158

14
(No Transcript)