DeSigN: connecting gene expression with therapeutics for drug repurposing and development Bernard lee GIW 2016, Shanghai 8 October 2016 1
Motivation Average cost: USD 1.8 to 2.6 billion ~2% Attrition rate is high To develop discovery pipelines to repurpose FDA-approved drugs (Duration: 10 15 years) Modified from Ware et al. 2016. Trends Biotechno. 2
Background Cancer genome characterization can be used to link genetic features to drug response Melanoma patients with the BRAF V600E mutation à improved overall survival with vemurafenib treatment (Chapman et al. 2011) NSCLC patients with KRAS-mutant showed poorer clinical outcomes when treated with erlotinib and chemotherapy (Eberhard et al. 2005) 3
Background Connectivity Map (CMap) - mine patterns of association between drug sensitivity and gene expression signatures (Lamb et al. 2006) A B C D A B C Drug X D DeSigN Differentially Expressed Gene Signatures INhibitors Freely available at http://design.cancerresearch.my/ (Lamb et al. 2006) Feature CMap DeSigN Genetic signature pre & post treatment GE profile baseline GE profile # Compounds 1,309 140 # Cell lines 4 604 solid tumors 4
Aim To identify novel drugs that have good potential to be repurposed for cancer therapy Objectives To build a user-friendly tool for predicting drug efficacy against cancer cell lines using gene expression patterns To validate drug sensitivity prediction To identify drugs that can repurposed for cancer treatment 5
DeSigN Platform 6
Reference Database Rank 1 for largest value 6. Sort and rank the gene lists according to moderated t- statistics 1. Download microarray data from (GDSC) 604 solid tumors GE 2. Normalized using MAS5 Median Sensitive 5. Differential analysis (Sen vs Res) 140 drugs Resistant 4. Rank cell lines according to IC 50 3. Collapsed Probe sets using GSEA 12,772 unique 7 genes
DeSigN Platform 8
Query Signatures Differentially expressed genes (DEGs) were selected using joint filtering of pvalue and fold change (Xiao et al. 2014) log2 fold change > 1 and pvalue < 0.01 9
DeSigN Platform 10
Pattern-matching algorithm Association of query signatures to rankordered gene expression profile database à Kolmogorov-Smirnov (KS) statistic To correlate drugs in DeSigN database that enrich similar DEG based on the IC 50 drug sensitivity profiles 11
DeSigN Web Interface Maximal efficacy Minimal efficacy 12
Validations of DesigN 2 approaches: (i) Published studies GEO datasets (ii) In vitro testing oral cancer lines 13
NCBI Gene Expression Omnibus (GEO) datasets GEO reference Drug Response Number of sensitive samples Number of resistant samples Platform Reference GSE4342 Gefitinib Sensitive 17 12 GPL96 Coldren et al. 2006 GSE16179 Lapatinib Sensitive 3 3 GPL570 Liu et al. 2009 GSE9633 Dasatinib Sensitive 11 5 GPL571 Wang et al. 2007 GSE35141 Gemcitabine Resistant 6 6 GPL4133 Saiki et al. 2012 Sensitive cell lines: IC 50 < 1 µm; Resistant cell lines: IC 50 > 1 µm 14
DeSigN validation using GEO datasets GEO reference Drug Expected Drug Sensitivity DeSigN Rank Target Connectivity Score p-value GSE4342 Gefitinib Sensitive 1 EGFR 1.00 0.000 GSE16179 Lapatinib Sensitive 6 EGFR, ERBB2 0.87 0.015 GSE9633 Dasatinib Sensitive 6 ABL, SRC, KIT, PDGFR 0.83 0.025 GSE35141 Gemcitabine Resistant 129 DNA replication -0.83 0.025 15
Potentially efficacious Drugs for Oral Cancer Lines Differential analysis of5 oral cancer and 2 normal keratinocytes 69 up- and 86 down-regulatedgenes Evaluatedthe efficacy of bosutinib Have no known effects againsthuman oralcancer Efficacy of bosutinibis not anticipatedwhen used against oral lines 16
Bosutinib induced cytotoxic effect Bosutinib was tested on three oral cancer lines Using cell viability assay (MTT) Demonstrated significant mean IC 50 value compared to sensitive (HSC-4) and resistant control (MCF7) Sensitive Resistant CancerLines Mean IC 50 ± SE -log 10 (p-value) relative to HSC-4 ORL-196 (n = 4) 0.75 ± 0.03 5.8 1.9 ORL-204 (n = 3) 0.90 ± 0.04 3.6 1.9 ORL-48 (n = 5) 1.19 ± 0.05 4.1 1.9 HSC-4 (n = 3) 1.82 ± 0.03 - - -log 10 (p-value) relative to MCF7 MCF7 (n = 3) 12.22 ± 1.32 - - 17
Bosutinib induced apoptotic effect Bosutinib induced cell death in a time-dependentmanner By 72 hrs, significant number of apoptotic cells were detected in all three oral cancer lines (p-values < 0.01) HSC-4 MCF7 ORL-196 ORL-204 ORL-48 18
Bosutinib induced anti-proliferative effect Significant reduction in the number of proliferatingcells By 72 hrs, three oral cancer lines demonstrated growth inhibition of~40 80% at 1 μm Control 19
Summary DeSigN à developed and validated Identified bosutinib as a drug that could be repurposed for oral cancer treatment Future improvements: Integrating higher order data such as pathway related and post-translational modification genes 20
Acknowledgement High Impact Research, Ministry of Higher Education (HIR-MOHE) from University of Malaya (UM.C/625/1/HIR/MOHE/DENT-03) Cancer Research Malaysia Bernard Lee bernard.lee@cancerresearch.my Thank You 21
22
Used permutation approach to simulate the null distribution of Connectivity Score Simulate 1,000 random gene sets, each having the same size as the size of the input gene signature 23