Many solid tissues consist of two distinct anatomical compartments: the glandular epithelium and its surrounding stroma. Grossly dissected tumor samples include varying amounts of adjacent stroma, which may provide important clues to tumor initiation and progression;also, matching samples of normal tissue from the same individual include stromal, epithelial and other cells. Studies considering multiple tissue compartments for each patient allow for a deeper level of scientific investigation than normally seen in genomic analyses, but also pose unique challenges. Bioinformatics tools to address even the most basic scientific questions posed by these studies are lacking. Currently available methods to computationally separate expression from the different tissue compartments have a limited utility in addressing these questions, as they do not retain patients'individual and uniqu gene expression profile. This significantly limits our present ability to reproducibly infer tumor and stroma driven cancer molecular subtypes, and hence hampers downstream analysis of predicting personalized therapeutic targets. This proposal is to develop from the ground up the data analytic tools to address these two important challenges, and to demonstrate the utilization of these tools by investigating mechanisms by which obesity may affect the tumor-stroma interaction in prostate cancer patients. One of the proposed tools will provide the ability to dissect computationally the signals from individual cell types. This would accelerate research on the role of the surrounding environment (the microenvironment) across all cancer types, because it would permit the utilization of mixed samples to interrogate, at least partially, the transcriptional programs of multiple tissue compartments. Today, researchers must apply time-consuming approaches such as laser-capture microdissection (LCM) to physically dissect specimens if they want pure cell populations for expression profiling. The other proposed tool addresses the cross-talk question: what is the relationship between the transcriptional programs in the tumor and the surrounding (say stromal) cells? Is the activation of any stromal pathway associated with the activation of the same or different pathway in the tumor? Are specific combinations of pathway activities in the stroma and pathway activities in the tumor associated with worse prognosis? Are these combinations associated with treatment response? Are stromal gene signatures, alone or in conjunction with tumor information, predictive of progression and response to therapy? These are questions for which no statistical tools are available. We propose simple and effective analysis tools to address them. Lastly, our methods will allow investigation of the effect of obesity on tumor-stroma cross-talk in prostate cancer. It would use an outstanding existing resource, it would be the first of its kind, and has the potentia to generate important new hypotheses on the underlying mechanisms linking obesity and lethal prostate cancer.
In cancer research it is common to investigate samples representing a mixture of different cell types, unless a time consuming micro-dissection is performed prior to analysis. On the one hand this makes it difficult to understand the relation between features of these samples and disease subtypes or clinical outcomes, but on the other it also offers a completely untapped opportunity to better investigate the role of the cells immediately surrounding the tumor. This proposal is to develop from the ground up the data analytic tools to these issues, and to demonstrate the utilization of the tools by investigating mechanisms by which obesity may affect the tumor-stroma interaction in prostate cancer patients.
|Fan, Yu; Xi, Liu; Hughes, Daniel S T et al. (2016) MuSE: accounting for tumor heterogeneity using a sample-specific error model improves sensitivity and specificity in mutation calling from sequencing data. Genome Biol 17:178|
|Palculict, Timothy Blake; Ruteshouser, E Cristy; Fan, Yu et al. (2016) Identification of germline DICER1 mutations and loss of heterozygosity in familial Wilms tumour. J Med Genet 53:385-8|
|Nikooienejad, Amir; Wang, Wenyi; Johnson, Valen E (2016) Bayesian variable selection for binary outcomes in high-dimensional genomic studies using non-local priors. Bioinformatics 32:1338-45|
|Lefterova, Martina I; Shen, Peidong; Odegaard, Justin I et al. (2016) Next-Generation Molecular Testing of Newborn Dried Blood Spots for Cystic Fibrosis. J Mol Diagn 18:267-82|
|Peng, Gang; Fan, Yu; Wang, Wenyi (2014) FamSeq: a variant calling program for family-based sequencing data using graphics processing units. PLoS Comput Biol 10:e1003880|
|Ahn, Jaeil; Liu, Suyu; Wang, Wenyi et al. (2013) Bayesian latent-class mixed-effect hybrid models for dyadic longitudinal data with non-ignorable dropouts. Biometrics 69:914-24|
|Ahn, Jaeil; Yuan, Ying; Parmigiani, Giovanni et al. (2013) DeMix: deconvolution for mixed cancer transcriptomes using raw measured data. Bioinformatics 29:1865-71|