Quantity and quality of material 1030 cases have insufficient tissueRNA Limitations of GEP Tests Offlabel CUP tumours Potential for false positives if tested tumour not in classifier ID: 913370
Download Presentation The PPT/PDF document "Technical challenges -" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
Technical challenges - Quantity and quality of material (10-30% cases have insufficient tissue/RNA)
Limitations of GEP Tests
“Off-label” CUP tumours - Potential for false positives if tested tumour not in classifier
Accuracy is
tumour
type dependent
- Classification of
gastroenteropancreatic
, SCC and
mullerian
tumours
Kerr et al 2012
Clin
Cancer Res
Uneven classification performance
Slide3Pillai et al 2013 Mol Diagn
Uneven classification performance
Slide4Technical challenges - Quantity and quality of material (10-30% cases have insufficient tissue/RNA)
Limitations of GEP Tests
“Off-label” CUP tumours - Potential for false positives if tested tumour not in classifier differential
Accuracy is
tumour
type dependent
- Classification of
gastroenteropancreatic
, SCC and
mullerian
tumours
Misclassifications or no similarity match
- Misclassified latent primary (~25%) or no similarity match (~5% CUP)
Slide5MEREDITH: A multi-platform integration approach
Taskesen
et al Scientific Reports v6, a24949 (2016)PCA Dimensionality reduction and (tSNE
) 2D clustering
TCGA integrated analysis
- 4674 cancers, 19 types, 4 data types
http://
pancancer-map.ewi.tudelft.nl
/
Integration mRNA, Methylation,
miR
, CNV
18 clusters
8 show exclusivity of one cancer type
5 subtype exclusive
3 complex mix (SCC, CRC, Kidney)
Supervised
ToO
Classification
(AUC)
mRNA gene-expression = 0.93
DNA methylation = 0.93
miR
expression =0.93
CNV= 0.72
Combined data = 0.94
Slide6COP -I (KIRP)
COP-II
Cluster outside primary – “COPs”
COP I
– Cluster with other cancer types n=50 (enriched LUAD, LUSC, KIRC)
COP II
- Cluster together and away from other cancers n= 14 (LI, ACC, LUSC,BRCA)
cell cycle, RB-P107, immune, RNA splicing
MEREDITH: A multi-platform integration approach
Zoom from all sample
tSNE
COPs clustered alone
Slide7Talk outline
Brief history of gene-expression profiling for cancer type classification
Current commercially available tests - development and performance Clinical application
Problems and limitations
How DNA sequencing and mutation profiling can potentially help
Slide8CUP Presentation
March 2018 PET
: new right thigh soft tissue metastasis and small L adrenal met. Histopathology (03/2018): Thigh biopsy Moderately differentiated adenocarcinoma of uncertain origin. AE1/AE3 ++, CK19+,
TTF1-
,
CDX2+++, CK7-
,
CK20 -
,
CD56+
,
synaptophysin
+, chromogranin +, GATA3+, ER+, HER2+ (ISH
neg
), GCDFP-, SOX10 -. PAX8 -. PDL1 <1%
Prior History
:
March 2017
T2N2M0 NSCLC (Stage 3A). ‘Low grade adenocarcinoma’
Treatment
: Concurrent
ChemoRT
(Cisplatin/Etoposide),
Outcome
: Partial response on repeat CT after treatment
Patient
: 63
yo
female,
smoker (at least 40 pack
yr
hx
).
NSCLC
??
SUPER case with no similarity match
Linda
Mileshkin
(SUPER Study Lead)
Slide9SUPERDx
Nanostring ToO Assay (CUPGuidev2)Low confidence prediction of tissue of origin
Slide10Patient samples:
Tissue sources: Tumour (
organoid) and blood sample
Assay:
WGS: 38xN/60xT.
Driver mutations
:
KRAS, STK11
, CDKN2A, CTNNB1, STAG2, MGA, TMEM127
Tumour mutation Burden:
7.39
mutations/Mb (TMB -
Intermediate
(5 - 20 mutations/Mb)) 52 coding mutations
Whole genome sequencing
Slide11Test case 1105
COSMIC Mutation signatures
Alexandrov
et al Nature. 2013 Aug 22;500(7463):415-21
C>A
C>G
C>T
T>A
T>C
T>G
6 × 4 × 4 = 96 possible events
Trinucleotide
e.g. C
C
T
Smoking signature 4 supports metastatic lung cancer
Slide12ICGC WGS
machine learning 2267 samples18 cancer typesInput featuresRegional mutation density (RMD) (passenger mutation in heterochromatin state regions in cell of origin precursor cell)
Oncogenic driver mutations (OGM)COSMIC Mutation signatures (MS96) Overall Accuracy= 92%
ToO
classification based on whole genome sequencing
Salvadores
et al 2019.
PLoS
Comput
Biol
Contribution of data type
Slide13Gene-expression profiling can accurately classify primary & metastatic tumours
(~85-90% accuracy known primary)
SummaryUseful in resolving CUP cases with concordance of ~75% in latent primaries and IHC valid
.
Limitations
: - Tissue amount and quality (10-30%)
- Uneven classification performance across
tumour
types
- Misclassification “off-label
tumours
”
Mutation profiling can help for “molecularly” undifferentiated CUP cases
Evidence of clinical impact
- Change in patient management
- Improved survival in cases receiving a site directed therapy
Slide14SUPER Team
Linda MileshkinPenelope SchofieldDavid Bowtell
Krista FisherColin WoodDariush Etemadamoghadam
Alex Murray
Lisa
Guccione
Ellen
Schaef
Tharani
Sivakumaran
SUPER Study Sites
CUPGuide
Keith Byron
Adam
Kowalcyk
Fan Shi
Justin
Bedo
Tothill RADIO Lab
Atara
Posner
Shiva
Balanchander
Andrew Pattison
UMCCR (WGS Platform)
Sean
Grimmond
Oliver
Hofmnn
UMCCR Genomics
AOCS/
Bowtell
Lab
Nadia
Traficante
Sian
Fereday
Joy
Hendley
Bioinformatics
Jason Li
Kaushalya
Amarasinghe
Ken
Doig
Joshy
George
Pathology (Peter Mac)
Stephen Fox
Owen
Prall
Andrew Fellowes
Huiling
Xu
Anna
Tanska
Ain
Roesley
David Choong
CUP Consumer Representatives Patients and Family
Kym Sheehan
John Symons (CUP Foundation Jo’s Friends)
Acknowledgements