Passenger mutations accurately classify human tumors.

Determining the cancer type and molecular subtype has important clinical implications. The primary site is however unknown for some malignancies discovered in the metastatic stage. Moreover liquid biopsies may be used to screen for tumoral DNA, which upon detection needs to be assigned to a site-of-origin. Classifiers based on genomic features are a promising approach to prioritize the tumor anatomical site, type and subtype. We examined the predictive ability of causal (driver) somatic mutations in this task, comparing it against global patterns of non-selected (passenger) mutations, including features based on regional mutation density (RMD). In the task of distinguishing 18 cancer types, the driver mutations-mutated oncogenes or tumor suppressors, pathways and hotspots-classified 36% of the patients to the correct cancer type. In contrast, the features based on passenger mutations did so at 92% accuracy, with similar contribution from the RMD and the trinucleotide mutation spectra. The RMD and the spectra covered distinct sets of patients with predictions. In particular, introducing the RMD features into a combined classification model increased the fraction of diagnosed patients by 50 percentage points (at 20% FDR). Furthermore, RMD was able to discriminate molecular subtypes and/or anatomical site of six major cancers. The advantage of passenger mutations was upheld under high rates of false negative mutation calls and with exome sequencing, even though overall accuracy decreased. We suggest whole genome sequencing is valuable for classifying tumors because it captures global patterns emanating from mutational processes, which are informative of the underlying tumor biology.

Sparse representation for tumor classification based on feature extraction using latent low-rank representation.

Accurate tumor classification is crucial to the proper treatment of cancer. To now, sparse representation (SR) has shown its great performance for tumor classification. This paper conceives a new SR-based method for tumor classification by using gene expression data. In the proposed method, we firstly use latent low-rank representation for extracting salient features and removing noise from the original samples data. Then we use sparse representation classifier (SRC) to build tumor classification model. The experimental results on several real-world data sets show that our method is more efficient and more effective than the previous classification methods including SVM, SRC, and LASSO.

Banco de tumores: la base para la investigación molecular en uro-oncología / Biobanking: The basis for molecular research in Uro-Oncology

OBJETIVO: Los avances en la investigación traslacional del cáncer, así como el estudio de sus cambios e interacciones, dependen fundamentalmente de la obtención de series de casos (individuos afectados) y controles (no afectados) que proporcionan muestras de alta calidad y otros datos asociados. Los biobancos han mostrado ser una herramienta indispensable en el avance de la investigación uro-oncológica. MÉTODOS: Revisión de la literatura basada en biobancos con interés en urología. RESULTADOS Y CONCLUSIONES: Grandes biobancos bien organizados suponen un elemento clave en investigación en uro-oncologica. La integración de estos recursos con ciencias moleculares y diversas “omics”, junto a las poderosas herramientas bioinformáticas disponibles posibilita el avance en el conocimiento del desarrollo de la enfermedad uro-oncológica con fuertes implicaciones a la hora de planteamientos terapéuticos muy avanzados. En España, sin embargo, estas valiosas colecciones de material tisular y fluidos biológicos suelen estar poco utilizadas, debido principalmente a su fragmentación, poca accesibilidad, falta de estrategias de gestión adecuadas, como es la falta de consenso sobre los procedimientos operativos estándar, limitadas políticas específicas de utilización y distribución, así como falta de una amplia base en la cual se reflejen las necesidades de investigación bajo un enfoque interdisciplinar y multi-institucional; hemos de añadir el frecuente desconocimiento por el mundo urológico del elevado potencial científico de estas instituciones. El desarrollo del plan nacional español de biobancos supone una luz para el mejor aprovechamiento de este material por la comunidad uro-oncológica. A continuación se presenta una visión general en el tema de biobancos que puede servir como modelo para futuros debates sobre su utilización en uro-oncología. Esta aproximación está basada en datos de la literatura y resultados de discusión en diversos foros internacionales (AU)

OBJECTIVES: The advances in cancer translational research, as well as the study of its changes and interactions, depend basically on the procurement of case series (individuals affected and non affected controls) which supply high quality samples and other associated data. Biobanks have shown they are indispensable tools for the advance of uro-oncological research. METHODS: Bibliographic review based on biobanks with focus on Urology. RESULTS AND CONCLUSIONS: Well-organized, large biobanks are a key element in research in Uro-oncology. The integration of theses resources with molecular sciences and various “omics”, together with powerful available bioinformatic tools enable the advance in the knowledge of development of uro-oncological diseases, with strong implications at the time of very advanced therapeutic strategies. However, in Spain, these valuable collections of tissue material and biological fluids are usually not much in use, mainly due to fragmentation, low accessibility, lack of proper management strategies (such as lack of consensus about standard operative procedures), limited specific policies of use and distribution, as well as lack of a comprehensive base in which the research needs are reflected under interdisciplinar and multi-institutional focus. We must add the frequent ignorance of the high scientific potential of these institutions in the urological world. The development of the Spanish National Plan of Biobanks brings light for the better use of these materials by the uro-oncological community. We present a general view on the biobank topic, which may serve as a model for future debates about their use in uro-oncology. This approach is based in data from the literature and results of discussions in various international forums (AU)

Breast cancer molecular profiling with single sample predictors: a retrospective analysis.

BACKGROUND: Microarray expression profiling classifies breast cancer into five molecular subtypes: luminal A, luminal B, basal-like, HER2, and normal breast-like. Three microarray-based single sample predictors (SSPs) have been used to define molecular classification of individual samples. We aimed to establish agreement between these SSPs for identification of breast cancer molecular subtypes. METHODS: Previously described microarray-based SSPs were applied to one in-house (n=53) and three publicly available (n=779) breast cancer datasets. Agreement was analysed between SSPs for the whole classification system and for the five molecular subtypes individually in each cohort. FINDINGS: Fair-to-substantial agreement between every pair of SSPs in each cohort was recorded (kappa=0.238-0.740). Of the five molecular subtypes, only basal-like cancers consistently showed almost-perfect agreement (kappa>0.812). The proportion of cases classified as basal-like in each cohort was consistent irrespective of the SSP used; however, the proportion of each remaining molecular subtype varied substantially. Assignment of individual cases to luminal A, luminal B, HER2, and normal breast-like subtypes was dependent on the SSP used. The significance of associations with outcome of each molecular subtype, other than basal-like and luminal A, varied depending on SSP used. However, different SSPs produced broadly similar survival curves. INTERPRETATION: Although every SSP identifies molecular subtypes with similar survival, they do not reliably assign the same patients to the same molecular subtypes. For molecular subtype classification to be incorporated into routine clinical practice and treatment decision making, stringent standardisation of methodologies and definitions for identification of breast cancer molecular subtypes is needed. FUNDING: Breakthrough Breast Cancer, Cancer Research UK.

DNA and kinetic heterogeneity during the clonal evolution of adrenocortical proliferative lesions.

Monoclonal adrenocortical lesions show inverse correlation between proliferation and apoptosis, with proliferation being the single most important criterion of malignancy in adrenal lesions. No study yet has evaluated the variability of proliferation regarding the clonal pattern and diagnosis in adrenocortical nodular hyperplasias (ACNHs), adrenocortical adenomas (ACAs), and adrenocortical carcinomas (ACCs). We studied 69 ACNHs, 64 ACAs, and 23 ACCs (World Health Organization criteria) from 156 females. Clonality HUMARA test (from microdissected DNA samples), DNA content and proliferation analysis (slide and flow cytometry), and mitotic figure (MF) counting/50 high-power fields (HPFs) were performed in the same areas. Heterogeneity was assessed by 5cER (percentage of nonoctaploid cells with DNA content exceeding 5c) and standard deviation of MF/HPF. Statistics included analysis of variance/Student t tests regarding the clonal patterns and diagnosis. Polyclonal patterns were observed in 48 of 62 informative ACNHs and 7 of 56 informative ACAs, and monoclonal in 14 of 62 ACNHs, 49 of 56 ACAs, and 21 of 21 ACCs, with all hyperdiploid lesions (14 ACCs and 13 ACAs) being monoclonal. The standard deviation of MF/HPF progressively increased in ACNH-ACA-ACC (0.048 +/- 0.076, 0.110 +/- 0.097, 0.506 +/- 0.291, respectively; P = .0023), but did not differentiate ACNH/ACA. Only tetraploid percentage (P = .0496) and 5cER (P = .0352) distinguished polyclonal (3.64 +/- 2.20 and 0.14 +/- 0.15) from monoclonal (7.25 +/- 7.52 and 1.00 +/- 1.74) benign lesions. Malignancy significantly correlated with a low diploid percentage and high tetraploid percentage. Cell kinetic heterogeneity is the hallmark of adrenocortical neoplasms: tetraploid/hypertetraploid cell accumulation characterizes monoclonal lesions (suggesting nondisjunctional mitoses), whereas heterogeneously distributed mitotic figures and decreased diploid percentage define ACCs.

A comparison of cluster analysis methods using DNA methylation data.

MOTIVATION: Aberrant DNA methylation is common in cancer. DNA methylation profiles differ between tumor types and subtypes and provide a powerful diagnostic tool for identifying clusters of samples and/or genes. DNA methylation data obtained with the quantitative, highly sensitive MethyLight technology is not normally distributed; it frequently contains an excess of zeros. Established tools to analyze this type of data do not exist. Here, we evaluate a variety of methods for cluster analysis to determine which is most reliable. RESULTS: We introduce a Bernoulli-lognormal mixture model for clustering DNA methylation data obtained using MethyLight. We model the outcomes using a two-part distribution having discrete and continuous components. It is compared with standard cluster analysis approaches for continuous data and for discrete data. In a simulation study, we find that the two-part model has the lowest classification error rate for mixture outcome data compared with other approaches. The methods are illustrated using DNA methylation data from a study of lung cancer cell lines. Compared with competing hierarchical clustering methods, the mixture model approaches have the lowest cross-validation error for detecting lung cancer subtype (non-small versus small cell). The Bernoulli-lognormal mixture assigns observations to subgroups with the lowest uncertainty. AVAILABILITY: Software is available upon request from the authors. SUPPLEMENTARY INFORMATION: http://www-rcf.usc.edu/~kims/SupplementaryInfo.html

Marker identification and classification of cancer types using gene expression data and SIMCA.

OBJECTIVES: High-throughput technologies are radically boosting the understanding of living systems, thus creating enormous opportunities to elucidate the biological processes of cells in different physiological states. In particular, the application of DNA micro-arrays to monitor expression profiles from tumor cells is improving cancer analysis to levels that classical methods have been unable to reach. However, molecular diagnostics based on expression profiling requires addressing computational issues as the overwhelming number of variables and the complex, multi-class nature of tumor samples. Thus, the objective of the present research has been the development of a computational procedure for feature extraction and classification of gene expression data. METHODS: The Soft Independent Modeling of Class Analogy (SIMCA) approach has been implemented in a data mining scheme, which allows the identification of those genes that are most likely to confer robust and accurate classification of samples from multiple tumor types. RESULTS: The proposed method has been tested on two different microarray data sets, namely Golub's analysis of acute human leukemia and the small round blue cell tumors study presented by Khan et al.. The identified features represent a rational and dimensionally reduced base for understanding the biology of diseases, defining targets of therapeutic intervention, and developing diagnostic tools for classification of pathological states. CONCLUSIONS: The analysis of the SIMCA model residuals allows the identification of specific phenotype markers. At the same time, the class analogy approach provides the assignment to multiple classes, such as different pathological conditions or tissue samples, for previously unseen instances.

Numerous novel annotations of the human genome sequence supported by a 5'-end-enriched cDNA collection.

A collection of 90,000 human cDNA clones generated to increase the fraction of "full-length" cDNAs available was analyzed by sequence alignment on the human genome assembly. Five hundred fifty-two gene models not found in LocusLink, with coding regions of at least 300 bp, were defined by using this collection. Exon composition proposed for novel genes showed an average of 4.7 exons per gene. In 20% of the cases, at least half of the exons predicted for new genes coincided with evolutionary conserved regions defined by sequence comparisons with the pufferfish Tetraodon nigroviridis. Among this subset, CpG islands were observed at the 5' end of 75%. In-frame stop codons upstream of the initiator ATG were present in 49% of the new genes, and 16% contained a coding region comprising at least 50% of the cDNA sequence. This cDNA resource also provided candidate small protein-coding genes, usually not included in genome annotations. In addition, analysis of a sample from this cDNA collection indicates that approximately 380 gene models described in LocusLink could be extended at their 5' end by at least one new exon. Finally, this cDNA resource provided an experimental support for annotations based exclusively on predictions, thus representing a resource substantially improving the human genome annotation.

Improving reliability of gene selection from microarray functional genomics data.

Constructing a classifier based on microarray gene expression data has recently emerged as an important problem for cancer classification. Recent results have suggested the feasibility of constructing such a classifier with reasonable predictive accuracy under the circumstance where only a small number of cancer tissue samples of known type are available. Difficulty arises from the fact that each sample contains the expression data of a vast number of genes and these genes may interact with one another. Selection of a small number of critical genes is fundamental to correctly analyze the otherwise overwhelming data. It is essential to use a multivariate approach for capturing the correlated structure in the data. However, the curse of dimensionality leads to the concern about the reliability of selected genes. Here, we present a new gene selection method in which error and repeatability of selected genes are assessed within the context of M-fold cross-validation. In particular, we show that the method is able to identify source variables underlying data generation.

Statistical modeling and visualization of molecular profiles in cancer.

Current cancer classifications using morphological criteria produce heterogeneous classes with variable prognosis and clinical course. By measuring gene expression for thousands of genes in a single hybridization experiment, microarrays have the potential to contribute to more effective classifications based on molecular information. This gives hope to improve both prognosis and treatment. Statistical methods for molecular classification have focused on using high dimensional representations of molecular profiles to identify subclasses. These can be noisy, unstable, and highly platform-specific. In this article, we emphasize the notion of molecular profiles based on latent categories signifying under-, over-, and baseline expression. Following this approach, we can generate results that are more easily interpretable, more easily translated into clinical tools, more robust to noise, and less platform-dependent. We illustrate both the methods and the associated software for molecular class discovery on a data set of 244 microarrays comprising six known leukemia classes.

Microarray-based cancer diagnosis with artificial neural networks.

In recent years, the advent of experimental methods to probe gene expression profiles of cancer on a genome-wide scale has led to widespread use of supervised machine learning algorithms to characterize these profiles. The main applications of these analysis methods range from assigning functional classes of previously uncharacterized genes to classification and prediction of different cancer tissues. This article surveys the application of machine learning algorithms to classification and diagnosis of cancer based on expression profiles. To exemplify the important issues of the classification procedure, the emphasis of this article is on one such method, namely artificial neural networks. In addition, methods to extract genes that are important for the performance of a classifier, as well as the influence of sample selection on prediction results are discussed.

Simple rules underlying gene expression profiles of more than six subtypes of acute lymphoblastic leukemia (ALL) patients.

MOTIVATIONS AND RESULTS: For classifying gene expression profiles or other types of medical data, simple rules are preferable to non-linear distance or kernel functions. This is because rules may help us understand more about the application in addition to performing an accurate classification. In this paper, we discover novel rules that describe the gene expression profiles of more than six subtypes of acute lymphoblastic leukemia (ALL) patients. We also introduce a new classifier, named PCL, to make effective use of the rules. PCL is accurate and can handle multiple parallel classifications. We evaluate this method by classifying 327 heterogeneous ALL samples. Our test error rate is competitive to that of support vector machines, and it is 71% better than C4.5, 50% better than Naive Bayes, and 43% better than k-nearest neighbour. Experimental results on another independent data sets are also presented to show the strength of our method. AVAILABILITY: Under http://sdmc.lit.org.sg/GEDatasets/, click on Supplementary Information.

Characteristic attributes in cancer microarrays.

Rapid advances in genome sequencing and gene expression microarray technologies are providing unprecedented opportunities to identify specific genes involved in complex biological processes, such as development, signal transduction, and disease. The vast amount of data generated by these technologies has presented new challenges in bioinformatics. To help organize and interpret microarray data, new and efficient computational methods are needed to: (1) distinguish accurately between different biological or clinical categories (e.g., malignant vs. benign), and (2) identify specific genes that play a role in determining those categories. Here we present a novel and simple method that exhaustively scans microarray data for unambiguous gene expression patterns. Such patterns of data can be used as the basis for classification into biological or clinical categories. The method, termed the Characteristic Attribute Organization System (CAOS), is derived from fundamental precepts in systematic biology. In CAOS we define two types of characteristic attributes ('pure' and 'private') that may exist in gene expression microarray data. We also consider additional attributes ('compound') that are composed of expression states of more than one gene that are not characteristic on their own. CAOS was tested on three well-known cancer DNA microarray data sets for its ability to classify new microarray samples. We found CAOS to be a highly accurate and robust class prediction technique. In addition, CAOS identified specific genes, not emphasized in other analyses, that may be crucial to the biology of certain types of cancer. The success of CAOS in this study has significant implications for basic research and the future development of reliable methods for clinical diagnostic tools.

Molecular validation of the modified Vienna classification of colorectal tumors.

Although the Vienna classification has been introduced to resolve discrepancies in histological diagnoses of colorectal tumors between Western and Japanese pathologists, practical applications of this classification scheme have been problematic because invasion of the lamina propria of tumor cells is often difficult to recognize. Therefore, the following refinements of the classification criteria are needed: category 3, low-grade adenoma/dysplasia; category 4, intramucosal borderline neoplasia; 4-a, high-grade adenoma/dysplasia; 4-b, well-differentiated adenocarcinoma; category 5, definite carcinoma; 5-a, intramucosal moderately-differentiated adenocarcinoma; and 5-b, submucosal carcinoma. We attempted to test whether molecular genetic alterations are related to the modified classification scheme and whether they may help to further categorize the various intramucosal neoplasia grades of colorectal tumors. Two-hundred-thirty-two colorectal tumors were examined using flow cytometric analysis of DNA content, polymerase chain reaction microsatellite assays, and single-strand conformational polymorphism assays to detect abnormalities of DNA content, chromosomal allelic loss, and Ki-ras and p53 gene mutations. Microsatellite instability (MSI) was also examined. Frequencies of genetic alterations and DNA aneuploid states increased with an increase in the grade assigned according to the modified Vienna classification. MSI was a rare event in colorectal adenomas and their frequency of MSI did not correlate with tumor grade. The combined genetic and DNA ploidy data support the conclusion that analysis of genetic alterations and DNA aneuploid states may help in appropriate categorization of colorectal tumors according to the modified Vienna scheme. In addition, MSI-positive tumors may represent a specific subtype of colorectal adenomas.

Block principal component analysis with application to gene microarray data classification.

We propose a block principal component analysis method for extracting information from a database with a large number of variables and a relatively small number of subjects, such as a microarray gene expression database. This new procedure has the advantage of computational simplicity, and theory and numerical results demonstrate it to be as efficient as the ordinary principal component analysis when used for dimension reduction, variable selection and data visualization and classification. The method is illustrated with the well-known National Cancer Institute database of 60 human cancer cell lines data (NCI60) of gene microarray expressions, in the context of classification of cancer cell lines.