Justin K. Huang

A Network of Conserved Synthetic Lethal Interactions for Exploration of Precision Cancer Therapy

An emerging therapeutic strategy for cancer is to induce selective lethality in a tumor by exploiting interactions between its driving mutations and specific drug targets. Here we use a multi-species approach to develop a resource of synthetic lethal interactions relevant to cancer therapy. First, we screen in yeast ∼169,000 potential interactions among orthologs of human tumor suppressor genes (TSG) and genes encoding drug targets across multiple genotoxic environments. Guided by the strongest signal, we evaluate thousands of TSG-drug combinations in HeLa cells, resulting in networks of conserved synthetic lethal interactions. Analysis of these networks reveals that interaction stability across environments and shared gene function increase the likelihood of observing an interaction in human cancer cells. Using these rules, we prioritize ∼10(5) human TSG-drug combinations for future follow-up. We validate interactions based on cell and/or patient survival, including topoisomerases with RAD17 and checkpoint kinases with BLM.

ERCC1 and TS Expression as Prognostic and Predictive Biomarkers in Metastatic Colon Cancer.

In patients with metastatic colon cancer, response to first line chemotherapy is a strong predictor of overall survival (OS). Currently, oncologists lack diagnostic tests to determine which chemotherapy regimen offers the greatest chance for response in an individual patient. Here we present the results of gene expression analysis for two genes, ERCC1 and TS, measured with the commercially available ResponseDX: Colon assay (Response Genetics, Los Angeles, CA) in 41 patients with de novo metastatic colon cancer diagnosed between July 2008 and August 2013 at the University of California, San Diego. In addition ERCC1 and TS expression levels as determined by RNAseq and survival data for patients in TCGA were downloaded from the TCGA data portal. We found that patients with low expression of ERCC1 (n = 33) had significantly longer median OS (36.0 vs. 10.1 mo, HR 0.29, 95% CI .095 to .84, log-rank p = 9.0x10-6) and median time to treatment to failure (TTF) following first line chemotherapy (14.1 vs. 2.4 mo, HR 0.17, 95% CI 0.048 to 0.58, log-rank p = 5.3x10-4) relative to those with high expression (n = 4). After accounting for the covariates age, sex, tumor grade and ECOG performance status in a Cox proportional hazard model the association of low ERCC1 with longer OS (HR 0.18, 95% CI 0.14 to 0.26, p = 0.0448) and TTF (HR 0.16, 95% CI 0.14 to 0.21, p = 0.0053) remained significant. Patients with low TS expression (n = 29) had significantly longer median OS (36.0 vs. 14.8 mo, HR 0.25, 95% CI 0.074 to 0.82, log-rank p = 0.022) relative to those with high expression (n = 12). The combined low expression of ERCC1/TS was predictive of response in patients treated with FOLFOX (40% vs. 91%, RR 2.3, Fisher’s exact test p = 0.03, n = 27), but not with FOLFIRI (71% vs. 71%, RR 1.0, Fisher’s exact test p = 1, n = 14). Overall, these findings suggest that measurement of ERCC1 and TS expression has potential clinical utility in managing patients with metastatic colorectal cancer.

A global transcriptional network connecting noncoding mutations to changes in tumor gene expression

Although cancer genomes are replete with noncoding mutations, the effects of these mutations remain poorly characterized. Here we perform an integrative analysis of 930 tumor whole genomes and matched transcriptomes, identifying a network of 193 noncoding loci in which mutations disrupt target gene expression. These ‘somatic eQTLs’ (expression quantitative trait loci) are frequently mutated in specific cancer tissues, and the majority can be validated in an independent cohort of 3,382 tumors. Among these, we find that the effects of noncoding mutations on DAAM1, MTG2 and HYI transcription are recapitulated in multiple cancer cell lines and that increasing DAAM1 expression leads to invasive cell migration. Collectively, the noncoding loci converge on a set of core pathways, permitting a classification of tumors into pathway-based subtypes. The somatic eQTL network is disrupted in 88% of tumors, suggesting widespread impact of noncoding mutations in cancer.Analysis of whole-genome sequences and transcription data from tumors identifies noncoding loci in which mutations affect target gene expression. These somatic eQTLs can classify tumors into pathway-based subtypes and are disrupted in 88% of tumors.

Disruption of NSD1 in head and neck cancer promotes favorable chemotherapeutic responses linked to hypomethylation

Human papillomavirus (HPV)–negative head and neck squamous cell carcinoma (HNSCC) represents a distinct classification of cancer with worse expected outcomes. Of the 11 genes recurrently mutated in HNSCC, we identify a singular and substantial survival advantage for mutations in the gene encoding Nuclear Set Domain Containing Protein 1 (NSD1), a histone methyltransferase altered in approximately 10% of patients. This effect, a 55% decrease in risk of death in NSD1-mutated versus non-mutated patients, can be validated in an independent cohort. NSD1 alterations are strongly associated with widespread genome hypomethylation in the same tumors, to a degree not observed for any other mutated gene. To address whether NSD1 plays a causal role in these associations, we use CRISPR-Cas9 to disrupt NSD1 in HNSCC cell lines and find that this leads to substantial CpG hypomethylation and sensitivity to cisplatin, a standard chemotherapy in head and neck cancer, with a 40% to 50% decrease in the IC50 value. Such results are reinforced by a survey of 1,001 cancer cell lines, in which loss-of-function NSD1 mutations have an average 23% decrease in cisplatin IC50 value compared with cell lines with wild-type NSD1. Significance: This study identifies a favorable subtype of HPV–negative HNSCC linked to NSD1 mutation, hypomethylation, and cisplatin sensitivity. Mol Cancer Ther; 17(7); 1585–94. ©2018 AACR.

Systematic Evaluation of Molecular Networks for Discovery of Disease Genes

Gene networks are rapidly growing in size and number, raising the question of which networks are most appropriate for particular applications. Here, we evaluate 21 human genome-wide interaction networks for their ability to recover 446 disease gene sets identified through literature curation, gene expression profiling, or genome-wide association studies. While all networks have some ability to recover disease genes, we observe a wide range of performance with STRING, ConsensusPathDB, and GIANT networks having the best performance overall. A general tendency is that performance scales with network size, suggesting that new interaction discovery currently outweighs the detrimental effects of false positives. Correcting for size, we find that the DIP network provides the highest efficiency (value per interaction). Based on these results, we create a parsimonious composite network with both high efficiency and performance. This work provides a benchmark for selection of molecular networks in human disease research.

Typing tumors using pathways selected by somatic evolution

Many recent efforts to analyze cancer genomes involve aggregation of mutations within reference maps of molecular pathways and protein networks. Here, we find these pathway studies are impeded by molecular interactions that are functionally irrelevant to cancer or the patient’s tumor type, as these interactions diminish the contrast of driver pathways relative to individual frequently mutated genes. This problem can be addressed by creating stringent tumor-specific networks of biophysical protein interactions, identified by signatures of epistatic selection during tumor evolution. Using such an evolutionarily selected pathway (ESP) map, we analyze the major cancer genome atlases to derive a hierarchical classification of tumor subtypes linked to characteristic mutated pathways. These pathways are clinically prognostic and predictive, including the TP53-AXIN-ARHGEF17 combination in liver and CYLC2-STK11-STK11IP in lung cancer, which we validate in independent cohorts. This ESP framework substantially improves the definition of cancer pathways and subtypes from tumor genome data.Informative pathways driving cancer pathogenesis and subtypes can be difficult to identify in the presence of many gene interactions irrelevant to cancer. Here, the authors describe an approach for cancer gene pathway analysis based on key molecular interactions that drive cancer in relevant tissue types, and they assemble a focused map of Evolutionarily Selected Pathways (ESP) with interactions supported by both protein–protein binding and genetic epistasis during somatic tumor evolution.

Genomic Landscape of Appendiceal Neoplasms

Purpose Appendiceal neoplasms are heterogeneous and are often treated with chemotherapy similarly to colorectal cancer (CRC). Genomic profiling was performed on 703 appendiceal cancer specimens to compare the mutation profiles of appendiceal subtypes to CRC and other cancers, with the ultimate aim to identify potential biomarkers and novel therapeutic targets. Methods Tumor specimens were submitted to a Clinical Laboratory Improvement Amendments-certified laboratory (Foundation Medicine, Cambridge, MA) for hybrid-capture-based sequencing of 3,769 exons from 315 cancer-related genes and 47 introns of 28 genes commonly rearranged in cancer. Interactions between genotype, histologic subtype, treatment, and overall survival (OS) were analyzed in a clinically annotated subset of 76 cases. Results There were five major histopathologic subtypes: mucinous adenocarcinomas (46%), adenocarcinomas (30%), goblet cell carcinoids (12%), pseudomyxoma peritonei (7.7%), and signet ring cell carcinomas (5.2%). KRAS (35% to 81%) and GNAS (8% to 72%) were the most frequent alterations in epithelial cancers; APC and TP53 mutations were significantly less frequent in appendiceal cancers relative to CRC. Low-grade and high-grade tumors were enriched for GNAS and TP53 mutations, respectively (both χ2 P < .001). GNAS and TP53 were mutually exclusive (Bonferroni corrected P < .001). Tumor grade and TP53 mutation status independently predicted OS. The mutation status of GNAS and TP53 strongly predicted OS (median, 37.1 months for TP53 mutant v 75.8 GNAS-TP53 wild type v 115.5 GNAS mutant; log-rank P = .0031) and performed as well as grade in risk stratifying patients. Conclusion Epithelial appendiceal cancers and goblet cell carcinoids show differences in KRAS and GNAS mutation frequencies and have mutation profiles distinct from CRC. This study highlights the benefit of performing molecular profiling on rare tumors to identify prognostic and predictive biomarkers and new therapeutic targets.

pyNBS: a Python implementation for network-based stratification of tumor mutations

Summary We present pyNBS: a modularized Python 2.7 implementation of the network-based stratification (NBS) algorithm for stratifying tumor somatic mutation profiles into molecularly and clinically relevant subtypes. In addition to release of the software, we benchmark its key parameters and provide a compact cancer reference network that increases the significance of tumor stratification using the NBS algorithm. The structure of the code exposes key steps of the algorithm to foster further collaborative development. Availability and implementation The package, along with examples and data, can be downloaded and installed from the URL https://github.com/idekerlab/pyNBS.

Abstract AP23: A PLATINUM–RESISTANT SUBTYPE OF HIGH–GRADE SEROUS OVARIAN CANCER IDENTIFIED BY A NETWORK OF SOMATIC MUTATIONS

PURPOSE: Like many solid tumors, high grade serous ovarian carcinoma (HGSOC) is a heterogeneous entity with a widely variable clinical course even among patients of the same stage and histological subtype. Although most patients will achieve remission with platinum-based chemotherapy, approximately 20% will display primary platinum resistance. If it were possible to platinum resistance from the molecular profile of a tumor, patients with platinum resistant tumor could be identified for alternative therapy, however, currently no such biomarker exists. METHODS: We have recently developed an analysis method called Network-based Stratification (NBS), which combines genome-scale somatic mutation profiles with genetic interaction networks. Briefly, somatic mutations for each patient are mapped onto a network, and then the influence of each mutation is propagated over its network neighborhood to give each gene a propagated mutation score (PM score). Unsupervised clustering is then used to assign the propagated networks into subtypes. RESULTS: Using NBS a high risk subtype consisting of approximately 20% of the patients was identified in both TCGA (n=330) and the ICGC (n=92) HGSOC cohorts. The median overall survival (OS) for the high risk (HR) subtype was a year less than standard risk (SR) subtype (36 vs. 48 mo, Log rank p = 1.6x 10-5) in TCGA cohort and similarly shorter (22 mo vs. not-yet-reached, p = 3.6x10-4), see figure 1. These survival differences were independent of age, tumor stage and residual tumor after surgical resection. The PM scores of the HR tumors from TCGA and ICGC cohorts we remarkably correlated (Pearson r2 = 0.94, p We then built a classifier to identify molecularly matched cell line models for in vitro study. The three cell lines with mutation profiles match the HR subtype (Kuramochi,Ovkate, OAW28) were significantly more resistant to cisplatin relative to cell lines characteristic of SR tumors (COV318, TYK-NU, OVCAR4), (IC50 14.4 vs. 3.3 µM, p < 0.0001). There was no significant difference in sensitivity to paclitaxel. CONCLUSIONS: NBS can be used to identify a molecular distinct subtype of HGSOC characterized by poor patient survival and primary platinum resistance. Citation Format: John Paul Shen, Ana Bojorquez-Gomez, Justin Huang, Matan Hofree, Kristin Klepper, Alex Beckett, Cheryl Saenz, Jason Kreisberg, Trey Ideker. A PLATINUM–RESISTANT SUBTYPE OF HIGH–GRADE SEROUS OVARIAN CANCER IDENTIFIED BY A NETWORK OF SOMATIC MUTATIONS [abstract]. In: Proceedings of the 11th Biennial Ovarian Cancer Research Symposium; Sep 12-13, 2016; Seattle, WA. Philadelphia (PA): AACR; Clin Cancer Res 2017;23(11 Suppl):Abstract nr AP23.