Decomposing%20Complex%20Clinical%20Phenotypes%20by%20Biologically%20Structured%20Microarray%20Analysis

1
Decomposing Complex Clinical Phenotypes by
Biologically Structured Microarray Analysis

• Claudio Lottaz and Rainer Spang

Berlin Center for Genome Based Bioinformatics,
Berlin (Germany)
Computational Diagnostics, Max Planck Institute
for Molecular Genetics, Berlin (Germany)
2
Overview

• Introduction
• Using functional annotation forsemi-supervised
  classification
• Heterogeneity vs. performance
• Evaluation on cancer related data
• Concluasions

3
Tumor Classification

• Setting
• Data gene expression profiles
• Goal prediction/classification of
  outcome/sub-type
• More formally
• Many expression levels measured
• Samples labelled as disease and control
• Train classifier

4
State-of-the-Art

• Various powerful methods
• Support vector machines
• Shrunken centroids...
• Regularization to fight overfitting
• Feature selection
• Large margins...
• Common hypothesisGenerate a single molecular
  signature

5
Complex Phenotypes

• A single clinical phenotype may be caused by
  different molecular mechanisms
• Our approach discover several sub-classes in
  disease group
• Each sub-class has a homogeneous molecular
  signature

6
Molecular Symptoms

• Classical signatures are globally optimal
• They have no biological focus
• Genes are corregulated thus correlated? in a
  global signature genes can be replaced with
  little loss
• Molecular Symptom
• A functionally focused signature to identify a
  disease sub-class
• High specificity sub-optimal sensitivity

7
Molecular Patient Stratification

• Patterns of molecular symptoms define a
  molecular patient stratification

8
Using Functionl AnnotationsA Priori vs. A
Posteriori

• Common procedure

• Our suggestion

9
Gene Ontology

• Biological terms ina directed graph
• Genes annotatedto terms
• Levelsrepresentspecificityof terms

10
Structured Analysis of Microarrays

• Classification in leaf nodes
• Regularized multivariate classifier
• Local signatures
• Diagnosis propagation
• Combine child diagnoses in inner nodes
• Generate more general diagnoses
• Regularization
• Shrink the classifier graph
• Remove uninformative branches

11
Leaf Node Classification

• Shrunken centroid classification(Tibshirani et
  al. 2002)
• Classificatino according to distance to centroids
• Regularization via gene shrinkage
• Determine probability-like values as
  classification results

12
Propagation of Classification

• Weighted averages
• Weight according to child performance
• Weights are normalized per inner node

Pa
w1
w3
w2
C1
C3
C2
13
Graph Shrinkage

• Weights of nodes are shrunken by a constant
• Negative weights are set to zero? uninformative
  branches vanish
• Best shrinkage level chosen in cross-validation

14
Biased Classifier Evaluation
Calibration of Sensitivity and Specificity
Shrinkage Parameter
Worst Performance in Leaf Node
?Cj DCi ( ?j Dj )-1
15
Classifier Heterogeneity

• Difference between two classifiersmeasures
  inconsistency of classifications
• Nodes redundancy
• Graphs redundancy(K? nodes of the shrunken
  graph)

16
Calibration

• Sensitivity vs. Specificity??
• Best classifiers set to control prevalence
• More molecular symptoms set ? higher than
  control prevalence
• Heterogeneity vs. Performance ?
• Molecular symptoms are heterogeneous
• Thus high ? eliminates them

17
Leukemia Data Set

• Data set by Yeoh et al. 2002
• Acute lymphocytic leukemia
• 327 patients of 7 clinical sub-types
• Expression profiles by HG-U95Av2
• Task for illustration
• Detect MLL sub-type
• 20 MLL samples
• 109 test set / 218 training set

18
Functional Annotations

• Focus on GOs Biological Process branch(8173
  terms)
• 12625 probesets on the chip
• 8679 genes (68.7 of probesets)
• In 1359 leaf nodes
• 845 inner nodes (total 2204 nodes)

19
MLL Classifier

• 2796 genes accessible through 32 nodes

20
MLL Stratification
21
Conclusions

• Semi-supervised classification
• Datect sub-classes
• In labelled disease groups
• Functional annotation
• Use in an a priori fashion
• To find biologically focused signatures??
  molecular symptoms
• Resolve complex clinical phenotypes
  (stratification through molecular symptoms)