Aim
To identify genetic risk factors that associate with development of
bortezomib induced peripheral neu
europathy in multiple myeloma
patients.
We hypothesize that SNPs are important in the pathogenesis BiNP
and may be
be relevant
relevant to identify patients at
at high
high-risk for BiNP
BiNP.
Phase III: PAD vs VAD as induction treatment
HOVON 65 MM / GMMG-HD4 study
MM Stage II or III, Age 1865
Randomization
3 x VAD
3 x Bort-
AD
CAD + GCSF
CAD + GCSF
MEL 200 + PBSCT
MEL 200 + PBSCT
Depending on local
Depending on local
Allogeneic
policy for
for patients PR
policy
py for patients
p
PR
MEL 200 + PBSCT
Tx
MEL 200 + PBSCT
Thalidomide
Bortezomib
50 mg/day for
1.3 mg/m2 / 2 weeks
2 years
for 2 years
maintenance
maintenance
Sonneveld et al. EHA 2009 Abstract 473 (Oral Presentation)
French Phase 3 trial IFM 2005-01 :
B t
or
ib/D
ezom
th
exame asone for ind
t
uc i
tion
prior to HDM (Vel/Dex vs VAD)
A1
A2
B1
B2
VAD x 4
VAD x 4
Induction
Vel/Dex x 4
Vel/Dex x 4
DCEP x 2
Consolidation
DCEP x 2
Melphalan
Melphalan
Melphalan
Melphalan
200mg/m2
200mg/m2
Transplant 1
200mg/m2
200mg/m2
+ ASCT
+A
+ SCT
ASCT
+ ASCT
+ ASCT
Second ASCT or RIC allo if <VGPR
Harousseau et al. ASH 2007 # 450
DCEP: Dexamethasone/Cyclophosphamide/ Etoposide/Cisplatin
VAD: Vincristine/Adriamycin/Dexamethasone
ASCT: Autologous Stem Cell Transplantation
Project Plan
Q: Do SNPs determine BiNP
HOVON65/HD4
IFM 2005-1
phase III Trial
Trial
phase III Trial
Trial
N=415
N=246
BOAC Platform
BOAC Platform
N=196
N=246
Cross-validation of
Pre-processing of data
datasets and methods
Final dataset
PLINK analysis for
Building a predictor for
associations
PNP
Incidence of BiNP
250
200
150
100
50
0
NO BiNP
BiNP grade 1
BiNP grade 2
BiNP grade 3
BiNP grade 4
HOVON65/GMMG HD4
88
55
26
20
7
IFM2005-01
144
27
60
11
4
Stratified Association Analysis
HOVON65
IFM 200501
Bor + AD
Bor + Dex
3x
4x
49 cases
86 controls
72 cases
139 controls
CochranMantelHaenszel test
This test provides a test based on an "average" odds ratio that controls for the
potential confounding due to the cluster variable
SNP associations
associations with BiNP
Table 1. SNPs associated with BiNP (P < .01) using a Cochran-Mantel-Haenszel stratified association analysis.
The genomic inflation factor is 1.0375.
SNP associa
tions
associa
with BiNP
(Pro)inflammatory genes are associated with BiNP
· Three SNPs that lie within or in close proximity of the TNF-
TNF gene (rs2857605,
rs2857598 and rs2228088) are found to be associated with BiNP
Figure 2. SNPs genotyped across
o
the TNF- gene in the region of
rs2857605.
Figures were generated using the web-based tool SNAP (Johnson, A.D.
et al., Bioinformatics 2008).
SNP associations
associations with BiNP
Association with neuropathy susceptibility genes
g
· A number of SNPs in the SERPINB2 gene are significantly associated with
BiNP risk. The main effect can be attributed to three SNPs (rs6102, rs6103,
and rs6104
rs
) that lie within a haplotype, and potentially influence SERPINB2
SERPINB
expression levels through their effect on splicing regulation
Haplotyp
pype block in the SERPINB2
region viewed in Haploview 4.0.
SNPs associated with BiNP are
shown in green.
Score Analysis
Scatter plot of BiNP risk score
A score for each patient is generated using the program PLINK v1.06
(Purcell S.
S et al.,. Am J Hum Genet 2007), and based on an allelic
scoring system involving SNPs significantly associated with BiNP (P
< .05). Score values for controls and BiNP cases are significantly
different (P = 0.00000001)
Conclusions (1)
SNPs associated with BiNP are located in
· (pro)inflammatory genes
· neuropathy susceptibility genes
The observed associated SNPs differ significantly from those
associated with thalidomide induced peripheral neuropathy (Johnson
DC et al., ASH 2008)
S. Corthals abstract # 1794
Developing a classifier for BiNP
Preprocessing
pg of data
·
Removal of SNPs missing in > 25% of the samples (n=155) resulted in
elimination of 59% of the uncalled data while only 1% of the genotyp
gy ed
p
data were lost.
·
Identify non European samples by a comparative PCA based on HapMap
samples from Asian, African and European ancestry
Dots are BOAC samples. Stains are HapMap clustering regions
·
Imputation was performed to generate complete genotypes. For BOAC
SNPs intersecting with HapMap, the CEU panel was used. Remaining SNPs
were imputated on the basis of allelic frequency found on the BOAC chip
itself h
.T eover ll
a estimated error by imputation was 0.21%.
Developing a classifier for BiNP
Preprocessing of data and Model building
R.Kuiper abstract # 1800
Importan
Import t:
an
·
Exclusion of noise (BiNP = 1)
·
Separation of model building on a
learning set and model validation by
prediction on independent test set
Results
Performance Linear SVM
70.
.60
CO
rR
5
de
0.
un
reaA
.4
300.
020
40
60
80
100
Number of selected SNPs
PREDICTED AS:
BiNP
NO BiNP
84%
TRUE
BiNP
894
171
SENSITIVITY
STA
ST T
A U
T S
U :
S
No
34%
BiNP
1271
664
SPECIFICITY
41% PPV
80% NPV
Conclusion (2)
· Robust classifier (AUC:61% +/ 7%).
· Classifier based on BiNP
gr 234
2,3,4, pr
edicts
pr
welll for BiNP
gr 1
· Future analysis will determine whether the classifier shown here can be
improv
impr ed
ov
and will
eva
ev lu
l ate
a its clinical use. For this purpose the dat
da a
t set
described here will be subjected to novel classifying methods including
testing combinations of SNPs.
· Cross validation in 3rd independent dataset is warranted
Acknowledgements
Erasmus MC
IFM
SC
S Corth
th l
a s
HA
H A t
ve -Li
Loiseau
R Kuiper
JL Harousseau
Y de Knegt
F Magrangeas
H Vietor
S Minvielle
MD
M van D i
u n
A Uytterlinden
CS
R Hajek
HOVON
H Lokhorst
BOAC project
S Zweegman
G Morgan
R van der Holt
D Johnsen
Bv
B an
van Ness
Ness
GMMG
BGM Durie
H Goldschmidt
M Katz
D Hose
Jh
Johnson&J
&J h
o nson