https://www.genome.gov/Pages/Newsroom/Webcasts/2010ScienceReportersWorkshop/Panel5_integrated_genome_analysis_subtypes_of_glio.pdf

Integrated Genomic Analysis Identifies Clinically Relevant Subtypes of Glioblastoma Characterized by Abnormalities in PDGFRA, IDH1, EGFR, and NF1

The Cancer Genome Atlas Network recently cataloged recurrent genomic abnormalities in glioblastoma multiforme (GBM). We describe a robust gene expression-based molecular classification of GBM into Proneural, Neural, Classical, and Mesenchymal subtypes and integrate multidimensional genomic data to establish patterns of somatic mutations and DNA copy number. Aberrations and gene expression of EGFR, NF1, and PDGFRA/IDH1 each define the Classical, Mesenchymal, and Proneural subtypes, respectively. Gene signatures of normal brain cell types show a strong relationship between subtypes and different neural lineages. Additionally, response to aggressive therapy differs by subtype, with the greatest benefit in the Classical subtype and no benefit in the Proneural subtype. We provide a framework that unifies transcriptomic and genomic dimensions for GBM molecular stratification with important implications for future studies.

The epigenomics of cancer.

Most existing model-based approaches to anomaly detection construct a profile of normal instances, then identify instances that do not conform to the normal profile as anomalies. This paper proposes a fundamentally different model-based method that explicitly isolates anomalies instead of profiles normal points. To our best knowledge, the concept of isolation has not been explored in current literature. The use of isolation enables the proposed method, iForest, to exploit sub-sampling to an extent that is not feasible in existing methods, creating an algorithm which has a linear time complexity with a low constant and a low memory requirement. Our empirical evaluation shows that iForest performs favourably to ORCA, a near-linear time complexity distance-based method, LOF and random forests in terms of AUC and processing time, and especially in large data sets. iForest also works well in high dimensional problems which have a large number of irrelevant attributes, and in situations where training set does not contain any anomalies.

Isolation Forest

Gene transcription can be activated or inhibited by a reversible modification of the gene; this modification is termed an epigenetic change. This account of epigenetics in cancer reviews the mechan...

/pdf/epigenetics-in-cancer-4240m6az4l.pdf

Epigenetics in Cancer

Aberrations in post-translational modifications of histones have been shown to occur in cancer cells but only at individual promoters; they have not been related to clinical outcome. Other than being targeted to promoters, modifications of histones, such as acetylation and methylation of lysine and arginine residues, also occur over large regions of chromatin including coding regions and non-promoter sequences, which are referred to as global histone modifications. Here we show that changes in global levels of individual histone modifications are also associated with cancer and that these changes are predictive of clinical outcome. Through immunohistochemical staining of primary prostatectomy tissue samples, we determined the percentage of cells that stained for the histone acetylation and dimethylation of five residues in histones H3 and H4. Grouping of samples with similar patterns of modifications identified two disease subtypes with distinct risks of tumour recurrence in patients with low-grade prostate cancer. These histone modification patterns were predictors of outcome independently of tumour stage, preoperative prostate-specific antigen levels, and capsule invasion. Thus, widespread changes in specific histone modifications indicate previously undescribed molecular heterogeneity in prostate cancer and might underlie the broad range of clinical behaviour in cancer patients.

Global histone modification patterns predict risk of prostate cancer recurrence

A random forest (RF) predictor is an ensemble of individual tree predictors. As part of their construction, RF predictors naturally lead to a dissimilarity measure between the observations. One can also define an RF dissimilarity measure between unlabeled data: the idea is to construct an RF predictor that distinguishes the “observed” data from suitably generated synthetic data. The observed data are the original unlabeled data and the synthetic data are drawn from a reference distribution. Here we describe the properties of the RF dissimilarity and make recommendations on how to use it in practice.An RF dissimilarity can be attractive because it handles mixed variable types well, is invariant to monotonic transformations of the input variables, and is robust to outlying observations. The RF dissimilarity easily deals with a large number of variables due to its intrinsic variable selection; for example, the Addcl 1 RF dissimilarity weighs the contribution of each variable according to how dependent it is ...

http://www.genetics.ucla.edu/labs/horvath/RFclustering/RFclustering/RandomForestHorvath.pdf

Unsupervised Learning With Random Forest Predictors

Sexually dimorphic gene expression in mouse brain precedes gonadal differentiation

Identification of distinct molecular subtypes is a critical challenge for cancer biology. In this study, we used Affymetrix high-density oligonucleotide arrays to identify the global gene expression signatures associated with gliomas of different types and grades. Here, we show that the global transcriptional profiles of gliomas of different types and grades are distinct from each other and from the normal brain. To determine whether our data could be used to uncover molecular subtypes without prior knowledge of pathologic type and grade, we performed K-means clustering analysis and found evidence for three clusters with the aid of multidimensional scaling plots. These clusters corresponded to glioblastomas, lower grade astrocytomas and oligodendrogliomas (P<0.00001). A predictor constructed from the 170 genes that are most differentially expressed between the subsets correctly identified the type and grade of all samples, indicating that a relatively small number of genes can be used to distinguish between these molecular subtypes. These results further define molecular subsets of gliomas which may potentially be used for patient stratification, and suggest potential targets for treatment.

Gene expression profiling identifies molecular subtypes of gliomas.

Epidermal growth factor receptor (EGFR) overexpression occurs in nearly 50% of cases of glioblastoma (GBM), but its clinical and biological implications are not well understood. We have used Affymetrix high-density oligonucleotide arrays to demonstrate that EGFR-overexpressing GBMs (EGFR+) have a distinct global gene transcriptional profile. We show that the expression of 90 genes can distinguish EGFR+ from EGFR nonexpressing (EGFR-) GBMs, including a number of genes known to act as growth/survival factors for GBMs. We have also uncovered two additional novel molecular subtypes of GBMs, one of which is characterized by coordinate upregulation of contiguous genes on chromosome 12q13-15 and expression of both astrocytic and oligodendroglial genes. These results define distinct molecular subtypes of GBMs that may be important in disease stratification, and in the discovery and assessment of GBM treatment strategies.

Tao Shi

Papers

Global histone modification patterns predict risk of prostate cancer recurrence

Unsupervised Learning With Random Forest Predictors

Sexually dimorphic gene expression in mouse brain precedes gonadal differentiation

Gene expression profiling identifies molecular subtypes of gliomas.

Identification of molecular subtypes of glioblastoma by gene expression profiling