mutation hotspots are associated with gene expression ...mutation hotspots are associated with gene...

Mutation hotspots are associated with gene expression, signaling pathways, protein domains and drug response

Theo Knijnenburg

Brady Bernard

Ilya Shmulevich

William Poole

The location of mutations in a gene matters

GDSC

Mutations in BRAF

Finding mutation hotspots

Finding mutation hotspots EGFR in PANCAN11

Amino Acid Position

Mutation count

Finding mutation hotspots EGFR in PANCAN11

Mutation count

Smoothed at multiple

bandwidths

Finding mutation hotspots EGFR in PANCAN11

Mutation count

Multiscale clusters

Smoothed at multiple

bandwidths

Finding mutation hotspots EGFR in PANCAN11

Pkinase_Tyr Recep_L_domain Furin-like

Amino Acid Position

Mutation count

Multiscale clusters

Smoothed at multiple

bandwidths

• 75% of the mutations in hotspots are inside protein domains

Mutation hotspots vary substantially in size ~15,000 hotspots in ~2500 genes in PANCAN11

Mutation hotspots vary substantially in size ~15,000 hotspots in ~2500 genes in PANCAN11

Mutation hotspots vary substantially in size ~15,000 hotspots in ~2500 genes in PANCAN11

• 84% of the OncoDrive clusters overlap with the Multiscale clusters

Are these mutation hotspots functional?

Are these mutation hotspots functional? Identify statistical associations with: • Protein domains • Gene expression • Signaling pathways • Drug response

Statistical association between hotspots and gene expression

Binary mutation

calls

Mutation hotspots

Sam

ples

Continuous gene expression profiles

Genes

Sam

ples

Statistical association between hotspots and gene expression

Binary mutation

calls

Mutation hotspots

Sam

ples

Continuous gene expression profiles

Genes

Sam

ples

Pairwise

Statistical Tests Pipeline

Pairwise Correlations and P-values between hotspots and genes

Mut

atio

n ho

tspo

ts

Genes

Statistical association between hotspots and gene expression in UCEC

P=4·10-8

P=1·10-7

Statistical association between hotspots and gene expression in UCEC

P=8·10-9 P=2·10-3

Statistical association between hotspots and gene expression in UCEC

Type # Significant associations

Only in hotspot (hotspot P<10-4, all mutations P>10-2)

60

Only in gene (all mutations P<10-4, hotspot P>10-2)

1350

Both in gene and hotspot (all mutations P<10-4, hotspot P<10-4)

1360

Statistical association between hotspots and signaling pathways

Pairwise Correlations and P-values between hotspots and genes

Mut

atio

n ho

tspo

ts

Genes

Pathway membership in NCI/PID

Genes

Pat

hway

s

Statistical association between hotspots and signaling pathways

Combine P-values of genes in a pathway using Brown’s

method

Pairwise Correlations and P-values between hotspots and genes

Mut

atio

n ho

tspo

ts

Genes

Pathway membership in NCI/PID

Genes

Pat

hway

s P-values between hotspots

and pathways M

utat

ion

hots

pots

Pathways

Statistical association between PTEN hotspots and signaling pathways in GBM

Mutations in PTEN

n=27 n=73 n=7 n=10 n=7 n=5 n=6 n=8

Statistical association between PTEN hotspots and signaling pathways in GBM

Mutations in PTEN

n=27 n=73 n=7 n=10 n=7 n=5 n=6 n=8

Statistical association between PTEN hotspots and signaling pathways in GBM

Mutations in PTEN

n=27 n=73 n=7 n=10 n=7 n=5 n=6 n=8

Statistical association between PTEN hotspots and signaling pathways in GBM

Mutations in PTEN

n=27 n=73 n=7 n=10 n=7 n=5 n=6 n=8

P Lo

op

cα2

cα2

WPD

Lo

op

TI Loop

TI L

oop

Lee et al. Crystal Structure of the PTEN Tumor Suppressor: Implications for Its Phosphoinositide Phosphatase Activity and

Membrane Association. Cell, 1999.

Statistical association between PTEN hotspots and signaling pathways in GBM

WPD Loop

P Loop

Phosphatase domain C2 domain

Drug response depends on mutation hotspots

GDSC Mutations in PIK3CA

EGFR/ERBB2 inhibitor

Drug response depends on mutation hotspots

GDSC Mutations in PIK3CA

PIK3CA inhibitor

Summary

• We have developed a novel multiscale clustering algorithm to robustly identify mutation hotspots in genes

• We uncovered statistical associations between many of these mutation hotspots and gene expression, signaling pathways and drug response

Future work

• From PANCAN11 to the PANCAN Atlas • Integrate the mutation hotspots into Regulome

Explorer (RE) • Make code and hotspots available

Acknowledgements

• William Poole • Brady Bernard • Ilya Shmulevich • Sheila Reynolds • Vesteinn Thorsson • TCGA Genome Data Analysis Center (GDAC)

for Systems Analysis of the Cancer Regulome (CSACR)

Poster: 57

Statistical association between hotspots and gene expression in UCEC

Statistical association between hotspots and gene expression in UCEC

Statistical association between hotspots and gene expression in UCEC

Statistical association between hotspots and gene expression in UCEC

Statistical association between hotspots and gene expression in UCEC

Statistical association between hotspots and gene expression in UCEC