Magnifico

1
Magnifico!
2
Will Molecular Diagnostics become part of routine
practice?

• Professor Robert Leonard
• Human Cancer Studies Group Imperial College London

3
Routine?

• Meaning replace pathology?
• Definitely NO!
• Refining pathology/adding value
• IHC-Definitely YES!-already happening
• Meaning having to use Fresh material?
• Definitely NO!

4
Molecular signatures

• Have given us further insight into heterogeneity
  of breast cancer
• eg luminal A and B vs basal
• (but can use ICC to separate these subtypes)
• Therefore may prove valuable for future
  treatment design

5
Molecular signatures

• The question is will molecular signatures become
  part of routine diagnostics?
• Answer- not if they depend on reproducible
  management/processing of fresh tissue

6
Pitfalls in using fresh tissue

• The difficulty in getting fresh tissue
• Attitudes of surgeons OR logistics
• pathology department logistics
• Processing of fresh tissue
• Quality of tissue studied
• Amount of cancer
• Curation/storage effects

7
(No Transcript)
8
samples analysed
81-100
61-80
41-60
tumour epithelium
21-40
Breast N296 (178)
Colorectal N78 (46)
0-20
Renal N38 (18)
H and N N19 (18)
Prostate N104 (33)
9
All samples are not equal... GOOD!
18S and 28S peaks
Gel view
Degraded RNA
RNA graded using Agilent Bioanalyser - based on
presence and AUC of two ribosomal RNA peaks
10
Not so good
Even with strict adherence to SOPs, RNA can be
degraded in some samples but not in others - most
likely due to factors during operation
11
Effect of RNA quality on gene expression analysis
12
Some tissues are more prone to degradation than
others
13
Effect of method of storage on serum proteomics
Villanueva et al., 2005 J Proteome Res 4
1060-1072
14
Quality Assurance paramount at all stages

• Path QA - must be what it says on the tin -
  frozen section assessment needed before RNA
  extraction

• Molecular biology assays sensitive to amount of
  material put in - insufficient tumour epithelium
  means false negative result

• Data processing needs QA too

15
ie All array platforms are not equal.
Detection of differentially regulated genes vary
across platforms
Tan et al., 2003 Nucl Acid Res 31 5676
16
Which patients will benefit and can we afford
this?

• Caveats
• Signatures have been derived from archived
  tissue may not reflect the reality of the
  disease today. e.g. breast cancers smaller and
  now routinely use endocrine therapy if ERve
• Signatures frequently compared with less than
  gold standard pathology improve SOPs for
  pathology?

17
70 gene signature Mammaprint (vant Veer et
al., 2002)
18
Buyse et al., 2006 JNCI 98 1183-92
19
ER/Grade data are not given in Buyse Paper

• 100 of G1 are ER pos
• 76 of G2 are ER pos
• 59 of G3 are ER pos

20
Who can we exclude from testing?

• Less than 10 of G3 are gene low risk
• About 5 of ER neg are gene low risk
• Could exclude these from molecular diagnostics as
  little added value

21
Validation paper patients who can be excluded
from testing?
ER patients included in this study would
receive anti-estrogen therapy if seen in clinic
today their Risk Reduction would therefore be
considered to be at least 33 lower than the
validation set.
22
(No Transcript)
23
(No Transcript)
24
Exclude all G1 100 ER positive and can be
treated by Tam maximum possible risk 9
25
26/307 patients are likely to benefit from this
test is it worth it?
26
Oncotype Dx
27
Oncotype Dx

• Developed from an ER pos tam trial
• Validated in several chemo trials
• Great.but
• Why have they not compared it vs the dominant
  protein components?
• ie HER2/ER/Ki67 /- BCL2 /- invasion stained and
  then see what the gain is-if any?

28
Whats the alternative?
29
Non-molecular classifiers what do they include
(NB node negative cases only)?
30
ERve and ER-ve combined (n97)
Eden et al., 2004 EJC 40 1837-1841
31
ER Positive only (n68)
Eden et al., 2004 EJC 40 1837-1841
32
Neither gene expression classifier nor pathology
based ANN classifier successfully predicted
outcome in ER negative cases
Possible explanation - small numbers (29 ER- vs
68 ER), but both classifiers worked in random
selection of 29 ER cases
Failure likely to be result of difference in
disease - is ER-ve more heterogeneous? (NB
original population skewed for ER )
Eden et al., 2004 EJC 40 1837-1841
33
Gene signatures

• Not predictive of treatment effect (yet)
• Are prognostic signatures and as such need to
  prove themselves better than other tests because
• 1 They are expensive
• 2 They are difficult to do and may require fresh
  tissue
• Use gene signatures if they help to better
  understand pathophysiology

34
Genetic profiles should add to conventional
pathology not replace it
Morphology is the integration of gene expression
- should concentrate on subtle, less obvious
changes using classifiers to understand biology
and develop predictive (rather than prognostic)
profiles
Finding out where molecular biology can really
add value requires access to large numbers of
pathologically reviewed high quality samples -
need to get pathologists on board!
35

• Its the protein,stupid!