Evan O. Paull

Discovering causal pathways linking genomic events to transcriptional states using Tied Diffusion Through Interacting Events (TieDIE)

MOTIVATION Identifying the cellular wiring that connects genomic perturbations to transcriptional changes in cancer is essential to gain a mechanistic understanding of disease initiation, progression and ultimately to predict drug response. We have developed a method called Tied Diffusion Through Interacting Events (TieDIE) that uses a network diffusion approach to connect genomic perturbations to gene expression changes characteristic of cancer subtypes. The method computes a subnetwork of protein-protein interactions, predicted transcription factor-to-target connections and curated interactions from literature that connects genomic and transcriptomic perturbations. RESULTS Application of TieDIE to The Cancer Genome Atlas and a breast cancer cell line dataset identified key signaling pathways, with examples impinging on MYC activity. Interlinking genes are predicted to correspond to essential components of cancer signaling and may provide a mechanistic explanation of tumor character and suggest subtype-specific drug targets. AVAILABILITY Software is available from the Stuart labs wiki: https://sysbiowiki.soe.ucsc.edu/tiedie. CONTACT jstuart@ucsc.edu. SUPPLEMENTARY INFORMATION Supplementary data are available at Bioinformatics online.

Inferring causal molecular networks: empirical assessment through a community-based effort

It remains unclear whether causal, rather than merely correlational, relationships in molecular networks can be inferred in complex biological settings. Here we describe the HPN-DREAM network inference challenge, which focused on learning causal influences in signaling networks. We used phosphoprotein data from cancer cell lines as well as in silico data from a nonlinear dynamical model. Using the phosphoprotein data, we scored more than 2,000 networks submitted by challenge participants. The networks spanned 32 biological contexts and were scored in terms of causal validity with respect to unseen interventional data. A number of approaches were effective, and incorporating known biology was generally advantageous. Additional sub-challenges considered time-course prediction and visualization. Our results suggest that learning causal relationships may be feasible in complex settings such as disease states. Furthermore, our scoring approach provides a practical way to empirically assess inferred molecular networks in a causal sense.

Phosphoproteome Integration Reveals Patient-Specific Networks in Prostate Cancer

We used clinical tissue from lethal metastatic castration-resistant prostate cancer (CRPC) patients obtained at rapid autopsy to evaluate diverse genomic, transcriptomic, and phosphoproteomic datasets for pathway analysis. Using Tied Diffusion through Interacting Events (TieDIE), we integrated differentially expressed master transcriptional regulators, functionally mutated genes, and differentially activated kinases in CRPC tissues to synthesize a robust signaling network consisting of druggable kinase pathways. Using MSigDB hallmark gene sets, six major signaling pathways with phosphorylation of several key residues were significantly enriched in CRPC tumors after incorporation of phosphoproteomic data. Individual autopsy profiles developed using these hallmarks revealed clinically relevant pathway information potentially suitable for patient stratification and targeted therapies in late stage prostate cancer. Here, we describe phosphorylation-based cancer hallmarks using integrated personalized signatures (pCHIPS) that shed light on the diversity of activated signaling pathways in metastatic CRPC while providing an integrative, pathway-based reference for drug prioritization in individual patients.

Pathway-Based Genomics Prediction using Generalized Elastic Net

We present a novel regularization scheme called The Generalized Elastic Net (GELnet) that incorporates gene pathway information into feature selection. The proposed formulation is applicable to a wide variety of problems in which the interpretation of predictive features using known molecular interactions is desired. The method naturally steers solutions toward sets of mechanistically interlinked genes. Using experiments on synthetic data, we demonstrate that pathway-guided results maintain, and often improve, the accuracy of predictors even in cases where the full gene network is unknown. We apply the method to predict the drug response of breast cancer cell lines. GELnet is able to reveal genetic determinants of sensitivity and resistance for several compounds. In particular, for an EGFR/HER2 inhibitor, it finds a possible trans-differentiation resistance mechanism missed by the corresponding pathway agnostic approach.

Molecular Pathways: Extracting Medical Knowledge from High-Throughput Genomic Data

High-throughput genomic data that measures RNA expression, DNA copy number, mutation status, and protein levels provide us with insights into the molecular pathway structure of cancer. Genomic lesions (amplifications, deletions, mutations) and epigenetic modifications disrupt biochemical cellular pathways. Although the number of possible lesions is vast, different genomic alterations may result in concordant expression and pathway activities, producing common tumor subtypes that share similar phenotypic outcomes. How can these data be translated into medical knowledge that provides prognostic and predictive information? First-generation mRNA expression signatures such as Genomic Healths Oncotype DX already provide prognostic information, but do not provide therapeutic guidance beyond the current standard of care, which is often inadequate in high-risk patients. Rather than building molecular signatures based on gene expression levels, evidence is growing that signatures based on higher-level quantities such as from genetic pathways may provide important prognostic and diagnostic cues. We provide examples of how activities for molecular entities can be predicted from pathway analysis and how the composite of all such activities, referred to here as the “activitome,” helps connect genomic events to clinical factors to predict the drivers of poor outcome. Clin Cancer Res; 19(12); 3114–20. ©2013 AACR.

Common and cell-type specific responses to anti-cancer drugs revealed by high throughput transcript profiling

More effective use of targeted anti-cancer drugs depends on elucidating the connection between the molecular states induced by drug treatment and the cellular phenotypes controlled by these states, such as cytostasis and death. This is particularly true when mutation of a single gene is inadequate as a predictor of drug response. The current paper describes a data set of ~600 drug cell line pairs collected as part of the NIH LINCS Program (http://www.lincsproject.org/) in which molecular data (reduced dimensionality transcript L1000 profiles) were recorded across dose and time in parallel with phenotypic data on cellular cytostasis and cytotoxicity. We report that transcriptional and phenotypic responses correlate with each other in general, but whereas inhibitors of chaperones and cell cycle kinases induce similar transcriptional changes across cell lines, changes induced by drugs that inhibit intra-cellular signaling kinases are cell-type specific. In some drug/cell line pairs significant changes in transcription are observed without a change in cell growth or survival; analysis of such pairs identifies drug equivalence classes and, in one case, synergistic drug interactions. In this case, synergy involves cell-type specific suppression of an adaptive drug response.Understanding why some tumor cells respond to therapy and others do not is essential for advancing precision cancer care. Here, the authors perform large-scale transcriptomic profiling of breast cancer cell lines treated with anti-cancer drugs and find that certain drug classes induce cell line specific responses.

ONE-CLASS DETECTION OF CELL STATES IN TUMOR SUBTYPES

The cellular composition of a tumor greatly influences the growth, spread, immune activity, drug response, and other aspects of the disease. Tumor cells are usually comprised of a heterogeneous mixture of subclones, each of which could contain their own distinct character. The presence of minor subclones poses a serious health risk for patients as any one of them could harbor a fitness advantage with respect to the current treatment regimen, fueling resistance. It is therefore vital to accurately assess the make-up of cell states within a tumor biopsy. Transcriptome-wide assays from RNA sequencing provide key data from which cell state signatures can be detected. However, the challenge is to find them within samples containing mixtures of cell types of unknown proportions. We propose a novel one-class method based on logistic regression and show that its performance is competitive to two established SVM-based methods for this detection task. We demonstrate that one-class models are able to identify specific cell types in heterogeneous cell populations better than their binary predictor counterparts. We derive one-class predictors for the major breast and bladder subtypes and reaffirm the connection between these two tissues. In addition, we use a one-class predictor to quantitatively associate an embryonic stem cell signature with an aggressive breast cancer subtype that reveals shared stemness pathways potentially important for treatment.

The number of titrated microRNA species dictates ceRNA regulation

Abstract microRNAs (miRNAs) play key roles in cancer, but their propensity to couple their targets as competing endogenous RNAs (ceRNAs) has only recently emerged. Multiple models have studied ceRNA regulation, but these models did not account for the effects of co-regulation by miRNAs with many targets. We modeled ceRNA and simulated its effects using established parameters for miRNA/mRNA interaction kinetics while accounting for co-regulation by multiple miRNAs with many targets. Our simulations suggested that co-regulation by many miRNA species is more likely to produce physiologically relevant context-independent couplings. To test this, we studied the overlap of inferred ceRNA networks from four tumor contexts—our proposed pan-cancer ceRNA interactome (PCI). PCI was composed of interactions between genes that were co-regulated by nearly three-times as many miRNAs as other inferred ceRNA interactions. Evidence from expression-profiling datasets suggested that PCI interactions are predictive of gene expression in 12 independent tumor- and non-tumor contexts. Biochemical assays confirmed ceRNA couplings for two PCI subnetworks, including oncogenes CCND1, HIF1A and HMGA2, and tumor suppressors PTEN, RB1 and TP53. Our results suggest that PCI is enriched for context-independent interactions that are coupled by many miRNA species and are more likely to be context independent.

Revealing cancer subtypes with higher-order correlations applied to imaging and omics data

BackgroundPatient stratification to identify subtypes with different disease manifestations, severity, and expected survival time is a critical task in cancer diagnosis and treatment. While stratification approaches using various biomarkers (including high-throughput gene expression measurements) for patient-to-patient comparisons have been successful in elucidating previously unseen subtypes, there remains an untapped potential of incorporating various genotypic and phenotypic data to discover novel or improved groupings.MethodsHere, we present HOCUS, a unified analytical framework for patient stratification that uses a community detection technique to extract subtypes out of sparse patient measurements. HOCUS constructs a patient-to-patient network from similarities in the data and iteratively groups and reconstructs the network into higher order clusters. We investigate the merits of using higher-order correlations to cluster samples of cancer patients in terms of their associations with survival outcomes.ResultsIn an initial test of the method, the approach identifies cancer subtypes in mutation data of glioblastoma, ovarian, breast, prostate, and bladder cancers. In several cases, HOCUS provides an improvement over using the molecular features directly to compare samples. Application of HOCUS to glioblastoma images reveals a size and location classification of tumors that improves over human expert-based stratification.ConclusionsSubtypes based on higher order features can reveal comparable or distinct groupings. The distinct solutions can provide biologically- and treatment-relevant solutions that are just as significant as solutions based on the original data.

Abstract 788: Transcriptional landscape of drug response guides the design of potent and synergistic drug combinations

Transcriptional profiling of drug-treated cells yields high dimensional response signatures that allow drugs to be compared with each other. For example, the Connectivity Map collects signatures that are aggregated across multiple cell types. However, most therapeutic drugs are effective only against a subset of disease genotypes, particularly in the case of anti-cancer drugs. Here we ask how transcriptional signatures vary across cell lines and dose and correlate these signatures to the phenotypic response (growth inhibition). Using these cell line specific signatures, we inferred which signaling pathways are perturbed by specific kinase inhibitors and identified synergistic drug combinations. We treated 6 breast cancer cell lines with more than 100 targeted inhibitors at 6 doses and measured their transcriptional response at 2 time points. We focused on inhibitors targeting key the PI3K and MAPK signaling pathways, as well as receptor tyrosine kinases (RTKs) and cyclin-dependent kinases (CDKs); many of them are currently studied in clinical trials. We identified that ∼40% of the perturbations induce a significant difference in their gene expression profile. Clustering revealed the signatures are time point specific. Some clusters contain perturbations from multiple cell lines, like CDK inhibitors that down regulate genes related to the cell cycle in all six lines. In contrast, clusters comprising inhibitors of the PI3K and MAPK pathways are specific to each cell line and pathway. The perturbations induced by RTK and non-RTK inhibitors cluster with either the PI3K or the MAPK inhibitors depending on the cell line. Thus, the transcriptional response allows us to identify differences in pathway usage between cell lines, in particular to which pathway RTKs signal predominantly. We found that the significance of the transcriptional signature is not necessarily related to growth inhibition. In particular, some inhibitors have little effect on growth, yet induce a significant transcriptional signature. The most striking case is the inhibition of MEK and EGFR in BT20 that induces strong transcriptional and biochemical responses but inhibits growth by only ∼20%. Based on the transcriptional signature we inferred and validated experimentally that FoxO, which is generally regulated by the PI3K pathway, is partially activated by MEK or EGFR inhibition. This suggests that EGFR and PI3K inhibitors act synergistically in BT20, which we validated experimentally both at the level of FoxO activation and growth inhibition. We validated the most promising drug pair by treating xenografts. We have shown how we can use measurements of expression signatures and cellular phenotypes following single drug perturbations to identify drug combinations that are synergistic in individual cell lines. This approach is a step toward the rational design of co-drugging strategies with differential effect and larger therapeutic windows. Citation Format: Marc Hafner, Mario Niepel, Qiaonan Duan, Evan Paull, Josh Stuart, Aravind Subramanian, Avi Ma’ayan, Peter K. Sorger. Transcriptional landscape of drug response guides the design of potent and synergistic drug combinations. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 788.