Glioblastoma Multiforme (GBM) – Subtype Analysis

Download Report

Transcript Glioblastoma Multiforme (GBM) – Subtype Analysis

Glioblastoma Multiforme
(GBM) – Subtype Analysis
Lance Parsons
Introduction
• Clinicians (meat readers) determine
histological categorization:
Astrocytoma, Oligodendrocytoma, Mixed,
or Glioblastoma multiforme (GBM)
• GBM patients have poor prognosis, but
some surive unexpectly long.
• Molecularly and clinically distinct subtypes
of GBMs
Incorporating Biological Knowledge
• “Tiers” of classification
can assist with discovery
of downstream groups
• Glioma Classification
– Histological Level
– Clinical Level (Survival,
Age, etc)
– Transcriptiome (Gene
Expression Level)
• Gene Classification
– GO Hierarchy
– Pathway Databases
– Expression Level
(Microarray Data)
Age
Young
Tumor Type
Medium
Old
Astro Mixed Oligo GBM
Gene Subset
Expression
X
Y
X
Glioma Subtypes
A
B
C
Survival
Short
Long
Age and Survival
•
•
•
Young patients show greater variability in survival
Use this level of the “hierarchy” to assist in downstream analysis.
Very simple method is to use only the Young samples and find the groups
within that set of samples.
Normalization
• Making the numbers comparable
– Log Transform – Equalize variance, lineraize
data
– Median Center Arrays – Correct for
differences between arrays
– Standardize to unit variance?
Noise Filter
• Removing noise from the dataset
– Affymetrix software does some of this with
Present/Absent calls
– Fold-change filter?
– Other methods?
Feature (Gene) Selection
• Find genes highly correlated with patient
survival, within young sample group.
• Cox Proportional Hazards model
– Regression model that accounts for
“censored” data
• Permutation test can improve robustness
• Simple Cox selects 39 genes (permutation
pending)
Exploration of Results
• The genes we select
are statistically
significant (as
dictated by our Cox
testing methodology),
but they may not be
biologically or
clinically significant.
• Initial exploration
through hierarchical
clustering.
Clinical Validation
• Kaplan-Meier curves fit the two groups to a
survival model
Biological Validation
• file:///C:/Documents%20and%20Settings/L
ance/My%20Documents/Research/project
s/HenryFord/HFAnalysis-GBMYoung_annotations.html