The Gaussian graphical model (GGM) is widely applied to learn genetic network since it defines an undirected graph decoding the conditional dependence between genes. Many algorithms based on the GGM have been proposed for learning genetic network structures. Since the number of gene variables is typically far more than the number of samples collected, and a real genetic network is typically sparse, the graphical lasso implementation of GGM becomes a popular tool for inferring the conditional interdependence among genes. In this study, based on the guidance of specific types of human cancer pathways in the Kyoto Encyclopedia of Genes and Genomes (KEGG), I extracted the genes involved in a specific KEGG pathway and the corresponding RNA-seq expression levels in cancer and normal tissues from The Cancer Genome Atlas (TCGA), and constructed two types of small gene expression datasets: normal and cancer gene expression datasets corresponding to gene sets of different types of human cancers. I directly applied graphical lasso to the gene expression datasets of the genes to infer their genetic conditional dependences. By integrated analysis and comparison on these inferred normal and cancer networks, the results reveal highly conditional dependences among the genes at the RNA-seq expression levels and further confirm the essential roles played by the genes that encode proteins involved in the two-key signaling pathways phosphoinositide 3-kinase (PI3K)/AKT/mTOR and Ras/Raf/MEK/ERK in human carcinogenesis. These highly conditional dependences elucidate the expression level interactions among the genes that are implicated in many different human cancers. The inferred genetic networks were examined to further identify and characterize a collection of gene interactions that are unique to cancers. The cross-cancer genetic interactions revealed from our study provide another set of knowledge for cancer biologists to propose strong hypotheses, so further biological investi (open full item for complete abstract)

Gliomas are the most common primary malignant brain tumor. Glioblastoma (GBM), World Health Organization (WHO) Grade IV, is the most common, aggressive, and deadly glioma and has extremely poor prognosis. Lower grade gliomas (LGG), comprised of WHO Grades II and III tumors (sometimes referred to as low and intermediate grade gliomas), are invasive and may progress to higher-grade lesions. In 2016, the WHO reclassified the definition of GBM, dividing these tumors into isocitrate dehydrogenase (IDH)-wildtype and IDH-mutant GBM, where the vast majority are IDH-wildtype. A nomogram accounts is an easily accessible tool for physicians to use on behalf of their patients for predicting survival, developing individualized cancer prognosis, and deciding the interval for follow-up and/or imaging. Sex disparities in cancer survival have been established for several cancers but have remained inconclusive for gliomas. Therefore, we aimed to (1) develop and independently validate a nomogram for patients with newly diagnosed LGG, (2) develop and independently validate a nomogram for patients with newly diagnosed IDH-wildtype GBM, and (3) determine whether sex differences exist in glioma survival. The LGG nomogram was trained using data from The Cancer Genome Atlas (TCGA) and validated using OBTS data. The GBM nomogram was trained using data from the Ohio Brain Tumor Study (OBTS) and validated using data from the University of California San Francisco (UCSF). For both nomograms, survival was assessed using Cox proportional hazards regression, random survival forests, and recursive partitioning analysis. Nomograms provide an individualized estimate of survival rather than a group estimate and can be useful tools to patients and healthcare providers for counseling patients and their families regarding treatment decisions, follow-up, and prognosis. Free software for implementing these nomograms have been developed. To assess sex differences in glioma survival, data were obt (open full item for complete abstract)

Classically much of the focus in cancer biology has been on driver genes, i.e. tumor-suppressors and oncogenes, and how somatic mutations in these genes influence tumor phenotype. Large scale profiling studies like The Cancer Genome Atlas (TCGA) have produced massive repositories of genomic and transcriptomic data. This data has facilitated discovery and characterization of somatic drive mutations across cancers. In addition to these driver mutations cancer also have onco-passenger mutations that are passively acquired. These mutations are particularly common in tumors with extensive structural variations resulting in chromosomal deletions or amplifications. These variants are thought to target driver genes but result in copy number changes in hundreds of genes surrounding the driver genes- these genes are onco-passenger genes. Though individual instances of onco-passenger events have been characterized, a systematic understanding of their role in cancers is lacking. In addition to somatic changes, tumors also carry common heritable variants or germline polymorphisms. Recent studies have shown that these polymorphisms play critical roles in modulating gene expression across tissues. These regulatory polymorphisms are also related to numerous loci associated with predisposition to complex diseases. These studies indicate that germline polymorphisms could have a significant role in cancer biology. However, in contrast to somatic mutations we lack a systematic understanding of the functional role that germline polymorphisms play in cancers. The body of work presented here leverages the TCGA dataset, and using rigorous computational and statistical analysis hope to provide a more comprehensive picture of the role that onco-passenger genes and germline polymorphisms play in cancers.

The high-throughput next generation sequencing revolutionized the genomic sequencing techniques. It allows the study of thousands of genes and even the entire exome in a given organism simultaneously. This as well as other high-throughput technologies such as DNA microarray has broadened the genomic sequencing applications and changed biomedical research in a profound way. Comparing with microarray, the big data generated from next generation sequencing is considerably more reliable. As such, the technique has rapidly emerged as a major tool to obtain gene mutation and expression profiles of human cancers. The availability of these big genomic data presents unique scientific challenges and opportunities. One such challenge is to understand and characterize the patterns of genomic mutation and gene expression in different cancer types presented in the datasets. Many data mining approaches have already been developed to analyze the large datasets for feature selections and sample classifications. Since mutation and gene expression profiles are noisy due to both biological and technical variations in the data, it is clear that the effectiveness and robustness of a machine learning based classification system significantly depends upon the nature of the input data. In this study, we explore the DNA mutation and gene expression patterns in lung cancer using support vector machines with embedded parameter tuning. Two datasets used are derived from somatic mutation data and RNA-seq gene expression profiles presented in TCGA (The Cancer Genome Atlas). The embedded parameter tuning is based on data mining the training dataset using validation techniques and concepts of committee voting approach. We show that the support vector machines with tuning significantly improve the robustness and the classification accuracy when they are compared to the regular support vector machines. The approach was applied to the two datasets to explore the mutation patterns in lung adenocarcin (open full item for complete abstract)