Transcript Discovering conserved DNA

Targeted Cancer Therapy
Xiaole Shirley Liu
STAT115, STAT215, BIO298, BIST520
Limited Number of
Cancer Driver Genes and Pathways
~140 genes
2
Limited Number of Cancer Driver Genes
Half Druggable
~479 genes
3
Cancer Profiles vs Treatment
• “The Difficulty is going to be figuring out how to
use the information to help people rather than to
just catalogue lots and lots of mutations.” – Bert
Voglestein, John Hopkins University
• Chemotherapy vs targeted therapy
– Chemotherapy: non-specific cytotoxic drugs, mostly
affecting dividing cells, mostly intravenous
– Targeted: inhibit a specific target, less toxic to normal
cells, mostly oral
• http://www.foundationone.com video
4
ALK Inhibitors
• ALK normally functions in the
brain
• First rearrangement in lung
cancer discovered 2007 in
Japan
• Upstream of multiple cancer
pathways
• 2010 starting clinical trials on
ALK inhibitor
• 2011 FDA approved crizotinib
5
6
Testing on Patients Takes
Lots of Time and Money
Can we do this faster?
7
Cell Line Drug Screens
• CGP: 138 drugs on 727 cell lines
• CCLE: 24 drugs on 1,036 cell lines
8
Targeting a Cancer Pathway
• Why bother screening if we know the target of a
drug? E.g. doesn’t ALK inhibitor inhibit ALK?
9
Cell Line Drug Screens
• Cell lines:
– Expression
– Mutations
– Drug sensitivity
measure: IC50, half
maximal inhibitory
concentration (IC50)
• How to find expression
or mutation
biomarkers for drug
response? HW6
10
Drug Response BioMarkers
• Mutations
• Expression
AHR expression
high or low on
MEK inhibitor
(PD-0325901)
11
Instead of Drug-Focused, Can
we Test Tumor-Specific
Therapies?
12
Targeted Therapy
• ENO1 and ENO2
parallel pathway
• Glioblastoma
tumors with ENO1
deletion (5%) is
sensitive to ENO2
inhibition
13
Genome-wide Loss of Function Screens
• Get rid of a gene (DNA or
RNA) in a cell
• See how it influences one
specific cancer cell as
compared to other cells
(specificity)
• Can we do this in high
throughput?
14
Profile Cancer Cell Vulnerability
15
Genome-Wide CRISPR/Cas9
Knockout Screens
• Each vector contains a guide sequence (sgRNA)
knock out a gene (influence DNA) instead of
knock down expression (influence RNA)
• Detection through sequencing instead of barcoded arrays
Shalem et al, Science 2014; Wang et al, Science 2014
16
Analyzing Ge-LoF Screen Data
• How to normalize raw data?
• What if one shRNA / sgRNA doesn’t work
• How to identify key genes if we have multiple
shRNAs / sgRNA per gene?
17
Summary
• Use expression and mutations as biomarkers to
predict drug response
• Use high throughput screening to identify specific
targets essential for cancer cells
• Can do this in cell lines (or animals) to save time
and $
• Lots of data, great for big data mining and
machine learning!!
18
Acknolwedgement
•
•
•
•
John Pack
James Lechner
Alex Chenchik
Haiyun Wang
19