Page 15 - Read Online

Page 270 Cancer Drug Resist 2018;1:266-302 I http://dx.doi.org/10.20517/cdr.2018.18

cancer (NSCLC) containing activating EGFR mutations. NSCLC tumours driven by EGFR variants are
initially sensitive to tyrosine kinase inhibitors (TKIs), but some patients show renewed disease progression
through expansion of tumour cell clones harbouring additional EGFR mutations. Models of drug resis-
tance are therefore necessary to increase our understanding of resistance mechanisms, to predict variants
that will arise in patients and to improve drug development strategies. CRISPR/Cas9 allows endogenous
modification of genes and facilitates introduction of all possible mutations at a specific locus, whilst main-
taining endogenous gene architecture, chromosomal context and epigenetic landscape. Here we describe a
CRISPR-based methodology developed to identify “on target” mutations in key oncogenes that drive phar-
macological resistance. Using EGFR as a target, we show how directed evolution of some residues in the
EGFR kinase domain can be predictive of the resistance observed in the clinic in response to known TKIs,
such as gefitinib and osimertinib. In fact, in the region analysed, we can recapitulate the variants observed
clinically. We also identify amino acids chemically and structurally able to confer resistance to a TKI in
structurally conserved tyrosine kinases such as BCR-Abl and Kit. The strategy also allows direct genera-
tion of disease relevant cellular models that recapitulate clinical drug resistance.

9. Molecular data integration for response prediction

Lodewyk Wessels

Netherlands Cancer Institute, Amsterdam, Netherlands

The exact mechanisms involved in tumor development and therapy response are still largely unclear. Here
we report on two computational approaches to systematically unravel these mechanisms and show how
these can be employed to predict response to anti-cancer agents. Clinical response to anti-cancer drugs
varies between patients and is modulated by molecular features. Classic approaches to integrate these
data for therapy response prediction almost exclusively employ gene expression data. Such predictors are
difficult to interpret. We developed TANDEM, a two-stage approach in which the first stage explains re-
sponse using upstream features (mutation, copy number and methylation) and the second stage explains
the remainder using downstream features (gene expression). Applying TANDEM to 934 cell lines profiled
across 265 drugs we show that the resulting models are more interpretable and equally predictive as classic
approches. Second, following a more mechanistic integration approach, we constructed Bayesian models
encompassing several of the important driver pathways and resistance mechanisms, and tested how well
these models describe drug response data derived from a panel of breast cancer cell lines. The models
provide estimates of the relative contribution of each of the drivers and resistance mechanisms and allow
estimation of latent variables such as “pathway activation”. We identify 4EBP1 protein expression as an im-
portant modulator of mTOR inhibitor response.

10. Mechanisms of therapy resistance in prostate cancer

Charlotte Bevan

Imperial College London, UK

Prostate tumours grow in response to androgens, which act via the androgen receptor (AR). Therapies
for inoperable, disseminated disease, therefore, are largely designed to inhibit androgen signalling - via
chemical castration (gonadal downregulation), inhibition of steroidogenesis and/or antiandrogens, which
directly bind to and inhibit the action of the AR. These are initially successful in the vast majority of pa-