Brain Tumor Update

1
Brain Tumor Update

• William A. Freije, M.D.

2
How to summarize 140,140,533 individual
measurements in twenty five slides
William A. Freije, M.D.
3
Math

• 22,283 probesets in 133A chip
• 22,645 probesets in 133B chip
• 44,928 probesets total
• 20 pairs for each probeset
• 151 samples
• 140,140,533 data points total

4
Tumor types examined to date

• Glial tumors
• Astrocytoma
• Pilocytic/infantile 2/2
• Grade II 2
• Grade III 9
• Oligodendroglioma
• Grade II 8
• Grade III 11
• Mixed oligo/astrocytoma (Grade III) 9
• Glioblastoma multiforme 63
• Gliosarcoma 3
• Ependymoma 3
• Poorly differentiated embryonal tumors
• Medulloblastoma 5
• Menigothelial tumors
• Meningioma 17
• Nerve sheath tumors
• Schwannoma 3
• Metastatic tumors 4 TOTAL 141

5
Summary of completed arrays

• 143 samples from 129 patients
• 141 tumor samples
• 2 normal brain specimens from GBM pts
• 8 normal brain controls
• From autopsy patients
• 151 samples completed on 133A and 133B Affymetrix
  oligonucleotide arrays

6
Patient samples- replicates, multiple sampling,
progression, and autopsy

• Replicates same tumor, same biopsy
• Mixed oligo/astro (2,1) proceeding to GBM (1)
• Oligo III (2) proceeding to GBM (1)
• Multiple sampling with progression analysis
• GBM tumor (2) 2nd resection of GBM (2)
• Progression
• Oligo III (1) 2 pts proceeding to GBM (1) 2 pts
• GBM (1) 2 pts proceeding to 2nd cranio GBM (1) 2
  pts
• Autopsy
• GBM (1) 2 pts with normal contralateral brain (1)
  2 pts

7
Objectives

• Global microarray classifier to discriminate the
  major types of primary brain tumors
  medulloblastoma, glioma, and meningothelial.
• Complement analysis with normal tissue data base.
• Dissect gene expression profiles of glial tumors
• Replace histologic classification scheme with
  expressed gene profile classification
• Prognosis- identify high and low risk disease
• Identify tumors with susceptibility to
  chemotherapeutic agents
• Tumor subclass identification
• Define molecular changes associated with disease
  progression

8
Objectives

• Explore the different mathematical models and
  software used to summarize microarray data.
• Dchip
• Genespring
• Statistical analysis
• t test
• Multi-dimensional scaling
• Random forest prediction
• Fisher linear discriminant analysis
• Log rank and Wilcoxon nonparametric tests
• Kaplan-Meier survival plots

9
Global microarray classifier
10
Global microarray classifier

• Gene expression profiles of the three major brain
  tumor types- medulloblastoma, meningioma, and
  glial- were examined with
• Multidimensional scaling
• Random forest prediction- all gene analysis
• Fisher linear discriminant analysis- univariate

11
MDS plot with 127 samples(normal not included)
12021 genes
12
Top 30 genes separating medulloblastoma,
meningioma, and gliomas Random Forest
13
Boxplots of probesets identified by Random Forest
Prediction which separate medulloblastoma,
meningioma, and glioma
14
Probeset 213033_s_at medulloblastoma specific
gene
Normal tissue project
complementary
Brain tumor data
15
Top 30 genes separating medulloblastoma,
meningioma, and gliomas Fisher Linear
Discriminant Assay
16
GBM analysis
17
GBM samples de novo vs progressive samples

• 63 samples total from 57 patients
• 48 de novo samples from 46 patients
• 15 progressive samples from 11 patients
• 3 de novo samples are autopsy specimens
• 1 progressive sample is from autopsy

18
GBM samples de novo vs progressive samples

• De novo
• Treated
• Untreated
• Total 48
• Progressive
• Treated
• Untreated
• Total 15

19
Clinical parameters

• TTP- time to progression
• Time in days from diagnosis or intervention (at
  UCLA) to the first evidence of disease
  progression
• TTS- time to survival
• Time in days from diagnosis or intervention (at
  UCLA) to the current day if alive or to the day
  of death
• Survival time
• Time in days from initial diagnosis to the
  current day if alive or to the day of death

20
TTP and TTS are related to time of treatment
2nd Rx
1st Rx
TTS 2nd TTP 2nd
DEATH
TTS 1st TTP 1st
Unknown
Birth
Clinical identification
Disease
Progression
Survival
21
GBM clinical parameters- TTP

• TTP (Days)
• Range 7-1031
• Mean 188
• Median 115
• SD 226
• TTP Alive Dead
• Median 219 79

22
GBM clinical parameters- TTS

• TTS (Days)
• Range 7-1247
• Mean 367
• Median 236
• SD 328
• TTS Alive Dead
• Median 927 187

23
GBM clinical parameters- Survival

• Survival (Days)
• Range 56-8330
• Mean 850
• Median 541
• SD 1191
• Survival Alive Dead
• Median 936 520

24
GBM clinical parameters- outliers

• Long term survivors all patients with TTSgt 723
  days. This group of patients all have TTP that
  are gt 612 days (TTP mean 2 SD). 10 patients in
  this group
• Short term survivors all patients with TTP lt 60
  days. This group of patients all had a TTS lt 150
  days. 13 patients in this group.
• All short term survivors are being reviewed for
  neurodecline secondary to other causes.

25
GBM living patients

• 13/57 GBM patients are alive
• 4/13 alive GBM patients have a TTPgt300 (range
  659-1311). All of these patients were identified
  in the outlier group.

26
GBM survival analysisstatistical methods employed

• R square of LDA to determine number of groups
• Best split measures
• Wilcoxon test statistics
• Logrank test statistics
• Kaplan-Meier survival plots

27
Determine Number of Clusters by Separating the 63
samples with gene expression ( with top 2000
brain tissue genes by LDA)
28
Association between gene expression and survival
time in GBM samples top 36 genes
29
Kaplan-Meier survival curves for specific
probesets
30
Future

• Turn this information into clinical difference
• Develop a model through insight developed from
  the microarray screen
• Confirm RT-PCR
• Develop specific assays for genes thought to be
  important in GBM survival (Ab)
• Cell line models
• Invasion assay
• Mouse models
• Example IGF2 highly expressed in poor survivors
• Global epigenetic imprint altered in GBM poor
  survival?

31
IGF2 expression in GBM long and short survivors
Graph courtesy of E. Castro