Accurate cancer phenotype prediction with AKLIMATE, a stacked kernel learner integrating multimodal genomic data and pathway knowledge.

Advancements in sequencing have led to the proliferation of multi-omic profiles of human cells under different conditions and perturbations. In addition, many databases have amassed information about pathways and gene "signatures"-patterns of gene expression associated with specific cellular and phenotypic contexts. An important current challenge in systems biology is to leverage such knowledge about gene coordination to maximize the predictive power and generalization of models applied to high-throughput datasets. However, few such integrative approaches exist that also provide interpretable results quantifying the importance of individual genes and pathways to model accuracy. We introduce AKLIMATE, a first kernel-based stacked learner that seamlessly incorporates multi-omics feature data with prior information in the form of pathways for either regression or classification tasks. AKLIMATE uses a novel multiple-kernel learning framework where individual kernels capture the prediction propensities recorded in random forests, each built from a specific pathway gene set that integrates all omics data for its member genes. AKLIMATE has comparable or improved performance relative to state-of-the-art methods on diverse phenotype learning tasks, including predicting microsatellite instability in endometrial and colorectal cancer, survival in breast cancer, and cell line response to gene knockdowns. We show how AKLIMATE is able to connect feature data across data platforms through their common pathways to identify examples of several known and novel contributors of cancer and synthetic lethality.

MEK-ERK signaling is a therapeutic target in metastatic castration resistant prostate cancer.

BACKGROUND: Metastatic castration resistant prostate cancer (mCRPC) is incurable and progression after drugs that target the androgen receptor-signaling axis is inevitable. Thus, there is an urgent need to develop more effective treatments beyond hormonal manipulation. We sought to identify activated kinases in mCRPC as therapeutic targets for existing, approved agents, with the goal of identifying candidate drugs for rapid translation into proof of concept Phase II trials in mCRPC. METHODS: To identify evidence of activation of druggable kinases in these patients, we compared mRNA expression from metastatic biopsies of patients with mCRPC (n = 101) to mRNA expression in localized prostate from TCGA and used this analysis to infer differential kinase activity. In addition, we assessed the differential phosphorylation levels for key MAPK pathway kinases between mCRPC and localized prostate cancers. RESULTS: Transcriptomic profiling of 101 patients with mCRPC as compared to patients with localized prostate cancer identified evidence of hyperactive ERK1, and whole genome sequencing revealed frequent amplifications of members of the MAPK pathway in 32% of this cohort. Next, we confirmed elevated levels of phosphorylated ERK1/2 in castration resistant prostate cancer as compared to untreated primary prostate cancer. We observed that the presence of detectable phosphorylated ERK1/2 in the primary tumor is associated with biochemical failure after radical prostatectomy independent of clinicopathologic features. ERK1 is the immediate downstream target of MEK1/2, which is druggable with trametinib, an approved therapeutic for melanoma. Trametinib elicited a profound biochemical and clinical response in a patient who had failed multiple prior treatments for mCRPC. CONCLUSIONS: We conclude that pharmacologic targeting of the MEK/ERK pathway may be a viable treatment strategy for patients with refractory metastatic prostate cancer. An ongoing Phase II trial tests this hypothesis.

A Community Challenge for Inferring Genetic Predictors of Gene Essentialities through Analysis of a Functional Screen of Cancer Cell Lines.

We report the results of a DREAM challenge designed to predict relative genetic essentialities based on a novel dataset testing 98,000 shRNAs against 149 molecularly characterized cancer cell lines. We analyzed the results of over 3,000 submissions over a period of 4 months. We found that algorithms combining essentiality data across multiple genes demonstrated increased accuracy; gene expression was the most informative molecular data type; the identity of the gene being predicted was far more important than the modeling strategy; well-predicted genes and selected molecular features showed enrichment in functional categories; and frequently selected expression features correlated with survival in primary tumors. This study establishes benchmarks for gene essentiality prediction, presents a community resource for future comparison with this benchmark, and provides insights into factors influencing the ability to predict gene essentiality from functional genetic screens. This study also demonstrates the value of releasing pre-publication data publicly to engage the community in an open research collaboration.

Integrative Analysis Identifies Four Molecular and Clinical Subsets in Uveal Melanoma.

Comprehensive multiplatform analysis of 80 uveal melanomas (UM) identifies four molecularly distinct, clinically relevant subtypes: two associated with poor-prognosis monosomy 3 (M3) and two with better-prognosis disomy 3 (D3). We show that BAP1 loss follows M3 occurrence and correlates with a global DNA methylation state that is distinct from D3-UM. Poor-prognosis M3-UM divide into subsets with divergent genomic aberrations, transcriptional features, and clinical outcomes. We report change-of-function SRSF2 mutations. Within D3-UM, EIF1AX- and SRSF2/SF3B1-mutant tumors have distinct somatic copy number alterations and DNA methylation profiles, providing insight into the biology of these low- versus intermediate-risk clinical mutation subtypes.

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.

A basal stem cell signature identifies aggressive prostate cancer phenotypes.

Evidence from numerous cancers suggests that increased aggressiveness is accompanied by up-regulation of signaling pathways and acquisition of properties common to stem cells. It is unclear if different subtypes of late-stage cancer vary in stemness properties and whether or not these subtypes are transcriptionally similar to normal tissue stem cells. We report a gene signature specific for human prostate basal cells that is differentially enriched in various phenotypes of late-stage metastatic prostate cancer. We FACS-purified and transcriptionally profiled basal and luminal epithelial populations from the benign and cancerous regions of primary human prostates. High-throughput RNA sequencing showed the basal population to be defined by genes associated with stem cell signaling programs and invasiveness. Application of a 91-gene basal signature to gene expression datasets from patients with organ-confined or hormone-refractory metastatic prostate cancer revealed that metastatic small cell neuroendocrine carcinoma was molecularly more stem-like than either metastatic adenocarcinoma or organ-confined adenocarcinoma. Bioinformatic analysis of the basal cell and two human small cell gene signatures identified a set of E2F target genes common between prostate small cell neuroendocrine carcinoma and primary prostate basal cells. Taken together, our data suggest that aggressive prostate cancer shares a conserved transcriptional program with normal adult prostate basal stem cells.

Multiplatform analysis of 12 cancer types reveals molecular classification within and across tissues of origin.

Recent genomic analyses of pathologically defined tumor types identify "within-a-tissue" disease subtypes. However, the extent to which genomic signatures are shared across tissues is still unclear. We performed an integrative analysis using five genome-wide platforms and one proteomic platform on 3,527 specimens from 12 cancer types, revealing a unified classification into 11 major subtypes. Five subtypes were nearly identical to their tissue-of-origin counterparts, but several distinct cancer types were found to converge into common subtypes. Lung squamous, head and neck, and a subset of bladder cancers coalesced into one subtype typified by TP53 alterations, TP63 amplifications, and high expression of immune and proliferation pathway genes. Of note, bladder cancers split into three pan-cancer subtypes. The multiplatform classification, while correlated with tissue-of-origin, provides independent information for predicting clinical outcomes. All data sets are available for data-mining from a unified resource to support further biological discoveries and insights into novel therapeutic strategies.

Impact of the National Breast and Cervical Cancer Early Detection Program on cervical cancer mortality among uninsured low-income women in the U.S., 1991-2007.

BACKGROUND: The benefits of the National Breast and Cervical Cancer Early Detection Program (NBCCEDP) on cervical cancer screening for participating uninsured low-income women have never been measured. PURPOSE: To estimate the benefits in life-years (LYs) gained; quality-adjusted life-years (QALYs) gained; and deaths averted. METHODS: A cervical cancer simulation model was constructed based on an existing cohort model. The model was applied to NBCCEDP participants aged 18-64 years. Screening habits for uninsured low-income women were estimated using National Health Interview Survey data from 1990 to 2005 and NBCCEDP data from 1991 to 2007. The study was conducted during 2011-2012 and covered all 68 NBCCEDP grantees in 50 states, the District of Columbia, five U.S. territories, and 12 tribal organizations. Separate simulations were performed for the following three scenarios: (1) women who received NBCCEDP (Program) screening; (2) women who received screening without the program (No Program); and (3) women who received no screening (No Screening). RESULTS: Among 1.8 million women screened in 1991-2007, the Program added 10,369 LYs gained compared to No Program, and 101,509 LYs gained compared to No Screening. The Program prevented 325 women from dying of cervical cancer relative to No Program, and 3,829 relative to No Screening. During this time period, the Program accounted for 15,589 QALYs gained when compared with No Program, and 121,529 QALYs gained when compared with No Screening. CONCLUSIONS: These estimates suggest that NBCCEDP cervical cancer screening has reduced mortality among medically underserved low-income women who participated in the program.

Allocating HIV prevention funds in the United States: recommendations from an optimization model.

The Centers for Disease Control and Prevention (CDC) had an annual budget of approximately $327 million to fund health departments and community-based organizations for core HIV testing and prevention programs domestically between 2001 and 2006. Annual HIV incidence has been relatively stable since the year 2000 and was estimated at 48,600 cases in 2006 and 48,100 in 2009. Using estimates on HIV incidence, prevalence, prevention program costs and benefits, and current spending, we created an HIV resource allocation model that can generate a mathematically optimal allocation of the Division of HIV/AIDS Prevention's extramural budget for HIV testing, and counseling and education programs. The model's data inputs and methods were reviewed by subject matter experts internal and external to the CDC via an extensive validation process. The model projects the HIV epidemic for the United States under different allocation strategies under a fixed budget. Our objective is to support national HIV prevention planning efforts and inform the decision-making process for HIV resource allocation. Model results can be summarized into three main recommendations. First, more funds should be allocated to testing and these should further target men who have sex with men and injecting drug users. Second, counseling and education interventions ought to provide a greater focus on HIV positive persons who are aware of their status. And lastly, interventions should target those at high risk for transmitting or acquiring HIV, rather than lower-risk members of the general population. The main conclusions of the HIV resource allocation model have played a role in the introduction of new programs and provide valuable guidance to target resources and improve the impact of HIV prevention efforts in the United States.

Estimated effects of the National Breast and Cervical Cancer Early Detection Program on breast cancer mortality.

BACKGROUND: The National Breast and Cervical Cancer Early Detection Program (NBCCEDP) provides breast cancer screening to medically underserved, low-income women aged 40-64 years. No study has evaluated NBCCEDP's effect on breast cancer mortality. PURPOSE: This study estimates life-years saved by NBCCEDP breast cancer screening compared with screening in the absence of NBCCEDP and with no screening. METHODS: A breast cancer simulation model based on existing Cancer Intervention and Surveillance Modeling Network models was constructed. The screening module from these models was modified to reflect screening frequency for NBCCEDP participants. Screening data for uninsured women represented what would have happened without the program. Separate simulations were performed for women who received NBCCEDP (Program) screening, women who potentially received screening without the program (No Program), and women who received no screening (No Screening). The impact of NBCCEDP was estimated as the difference in life-years between the Program and No Program, and the Program and No Screening scenarios. The analysis was performed in 2008-2009. RESULTS: Among 1.8 million women who were screened between 1991 and 2006, the Program saved 100,800 life-years compared with No Program and 369,000 life-years compared with No Screening. Per woman screened, the Program saved 0.056 life-years (95% CI=0.031, 0.081) compared with No Program and 0.206 life-years (95% CI=0.177, 0.234) compared with No Screening. Per woman with invasive breast cancer and screen-detected invasive cancer, the Program saved 0.41 and 0.71 life-years, respectively, compared with No Program. CONCLUSIONS: These estimates suggest that NBCCEDP breast cancer screening has reduced mortality among medically uninsured and underinsured low-income women.

A model for allocating CDC's HIV prevention resources in the United States.

The Division of HIV/AIDS Prevention (DHAP) at the Centers for Disease Control and Prevention has an annual budget of approximately $325 million for funding HIV prevention programs in the U.S. The purpose of this paper is to thoroughly describe the methods used to develop a national HIV resource allocation model intended to inform DHAP on allocation strategies that might improve the overall effectiveness of HIV prevention efforts. The HIV prevention resource allocation problem consists of choosing how to apportion prevention resources among interventions and populations so that HIV incidence is minimized, given a budget constraint. We developed an epidemic model that projects HIV infections over time given a specific allocation scenario. The epidemic model is then embedded in a nonlinear mathematical optimization program to determine the allocation scenario that minimizes HIV incidence over a 5-year horizon. In our model, we consider the general U.S. population and specific at-risk populations. The at-risk populations include 15 subgroups structured by gender, race/ethnicity and HIV transmission risk group. HIV transmission risk groups include high-risk heterosexuals, men who have sex with men and injection drug users. We consider HIV screening interventions and interventions to reduce HIV-related risk behaviors. The output of the model is the optimal funding scenario indicating the amounts to be allocated to all combinations of populations and interventions. For illustrative purposes only, we provide a sample application of the model. In this example, the optimal allocation scenario is compared to the current baseline funding scenario to highlight how the current allocation of funds could be improved. In the baseline allocation, 29% of the annual budget is aimed at the general population, while the model recommends targeting 100% of the budget to the at-risk populations with no allocation targeted to the general population. Within the allocation to behavioral interventions the model recommends an increase in targeting diagnosed positives. Also, the model allocation suggests a greater focus on MSM and IDUs with a 72% of the annual budget allocated to them, while the baseline allocation for MSM and IDUs totals 37%. Incorporating future epidemic trends in the decision-making process informs the selection of populations and interventions that should be targeted. Improving the use of funds by targeting the interventions and population subgroups at greatest risk may lead to improved HIV outcomes. These models can also direct research by pointing to areas where the development of cost-effective interventions can have the most impact on the epidemic.