Recent applications of NGS sequencing in
cancer studies
Andrew Gentles
CCSB NGS workshop
September 2012
You’ve slogged through QC, trimming,
alignment, realignment, variant calling
What next ?
• Mutational processes molding the genomes of 21
breast cancers/The life history of 21 breast cancers
– Nik-Zainal et al. (2012) Cell 149(5):994-1007
• Clonal evolution of preleukemic hematopoietic stem
cells precedes human acute myeloid leukemia
– Jan et al. (2012) Sci Trans Med 4, 149ra118
• Transcriptome sequencing across a prostate cancer
cohort identifies PCAT-1, an unannotated lincRNA
implicated in disease progression
– Prensner et al. (2011) Nat Biotech 29: 742-9
Companion papers from Cell May 2012
Whole genome sequencing of 21 Breast cancers
Sample
PD3851
PD3890
PD3904
PD3905
PD3945
PD4005
PD4006
PD4085
PD4086
PD4088
PD4103
PD4107
PD4109
PD4115
PD4116
PD4120*
PD4192
PD4194
PD4198
PD4199
PD4248
Previous
Histo patho
Age at first histopatholo
logical
ER Status
diagnosis
gical
Grade
diagnosis
61
Ductal
III
+
41
Ductal
III
39
Ductal
III
+
34
Ductal
III
59
Ductal
III
+
39
Ductal
III
39
Ductal
III
64
Ductal
III
+
58
Ductal
III
32
Ductal
III
+
46
Ductal
III
+
33
Ductal
III
67
Ductal
III
54
Ductal
III
+
32
Ductal
III
+
60
Ductal
II
+
70
Ductal
III
43
Lobular
III
+
59
Ductal
III
+
59
Ductal
II
48
Ductal
II
-
PR Status
HER2
Status
+
+
+
+
+
+
+
+
-
+
+
+
+
-
>30x coverage tumor and normal (188x for *)
BRCA
mutations
BRCA1
BRCA2
BRCA1
BRCA2
BRCA1
BRCA1
BRCA1
BRCA2
BRCA2
Analysis outline
• WGS sequencing to >30x coverage tumor/normal
– ~100 bp paired-end reads
– BWA alignment
• Compare tumor/normal for variant calling
– CaVEMan, Pindel
• Detection of structural rearrangements
– In-house method
• Inference of copy number changes
– ASCAT
Summary of somatic mutations
• 183916 somatic mutations (SNVs) identified in total
• 1372 missense, 117 nonsense, 2 stop-lost, 37 splice,
521 silent
• Most frequent mutations in known cancer genes
such as TP53, GATA3, PIK3CA, MAP2K4, SMAD4,
MLL2, MLL3, NCOR1
Higher rate in
BRCA1/2
C>A most
common
Mutational
spectrum in
breast cancer
Kataegis: regions of enhanced mutation rate
Kataegis is highly focal
upon zooming in
Kataegis associated with structural rearrangements
A very deep look into mutation frequencies to
reconstruct tumor evolution
PD4120a
• 188x coverage – enables deep look at mutation
frequencies
• 70690 somatic substitutions
– Some in <5% of reads
– Mainly C>* in TpC context
– High rate of validation
Patterns of copy number
alteration in PD4120a
Relatively few CNVs
Some sub-clonal
Mutation frequencies show clusters representing major and minor clones
D
C
B
A
1. 35% of reads -> all tumor cells since tumor is 70% tumor (cluster D)
2. Trisomy 1q early since few mutations with high read fraction – most are subclonal
3. 3 major clusters of sub-clonal mutations (A,B,C)
15600
5% 11% 19%
26762
35%
Founder clone
“most-recent
common ancestor”
D
C
B
A
4. Cluster C ~19% - more than half of tumor cells (since >1/2*35%)
“Pigeonhole principle”: for any 2 mutations, at least one tumor cell must have both –
must be on same part of phylogenetic tree
If one such mutation in greater fraction than another, must have occurred earlier
Cluster C must be on same phylogenetic branch as del13
• If SNVs close enough to SNPs, can
be phased with them
• 2171 on chr13
• 756 can be phased
Phasing of somatic mutations (Supp Fig 4)
Phasing of somatic mutations (Supp Fig 4)
Found 17 mutually exclusive, 76 examples of sub-clonal evolution
Figure 3:
Reconstructed
evolution of tumor
(see paper for details)
Sci Trans Med 2012
Prospective separation of residual HSC from leukemic patients
Residual HSC lack AML FLT3-ITD mutations
Strategy for identifying pre-leukemic mutations in HSC
67-239x exome coverage
Occurrence of AML mutations in residual HSC
~25000x
targeted
coverage
Mutations in HSC or both HSC/LSC
HSC with the pre-leukemic
mutations are capable of
differentiating to produce
functional immune cells
Filtering to identify ncRNAs
Enrichment of histone modification marks around transcripts
H3K4me2
Figure 2
H3K4me3
Novel transcripts are highly
expressed in prostate cancer
PCAT-1 is highly expressed in
metastatic/high-grade
prostate cancer
Figure 4b
Figure 3f
PCAT-1 expression is mutually
exclusive with EZH2
Relationship of PCAT-1 to EZH2/PRC complex
• RNA-seq discovers novel ncRNAs
• PCAT-1 highly expressed in high grade/metastatic
prostate cancer
• PCAT-1 promotes proliferation
• Hypothesized role with EZH2 (c.f. HOTAIR)
Final items
• Please fill out evaluation form!
• Slides:
– Available soon from http://ccsb.stanford.edu
• Sequence answers forum:
– http://seqanswers.com
• Stanford discussion group
• https://mailman.stanford.edu/mailman/listinfo/wgs_club
_stanford