Classification Of Human Cancers Research Articles

PND-Net: plant nutrition deficiency and disease classification using graph convolutional network

Crop yield production could be enhanced for agricultural growth if various plant nutrition deficiencies, and diseases are identified and detected at early stages. Hence, continuous health monitoring of plant is very crucial for handling plant stress. The deep learning methods have proven its superior performances in the automated detection of plant diseases and nutrition deficiencies from visual symptoms in leaves. This article proposes a new deep learning method for plant nutrition deficiencies and disease classification using a graph convolutional network (GNN), added upon a base convolutional neural network (CNN). Sometimes, a global feature descriptor might fail to capture the vital region of a diseased leaf, which causes inaccurate classification of disease. To address this issue, regional feature learning is crucial for a holistic feature aggregation. In this work, region-based feature summarization at multi-scales is explored using spatial pyramidal pooling for discriminative feature representation. Furthermore, a GCN is developed to capacitate learning of finer details for classifying plant diseases and insufficiency of nutrients. The proposed method, called Plant Nutrition Deficiency and Disease Network (PND-Net), has been evaluated on two public datasets for nutrition deficiency, and two for disease classification using four backbone CNNs. The best classification performances of the proposed PND-Net are as follows: (a) 90.00% Banana and 90.54% Coffee nutrition deficiency; and (b) 96.18% Potato diseases and 84.30% on PlantDoc datasets using Xception backbone. Furthermore, additional experiments have been carried out for generalization, and the proposed method has achieved state-of-the-art performances on two public datasets, namely the Breast Cancer Histopathology Image Classification (BreakHis 40×\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\ imes $$\\end{document}: 95.50%, and BreakHis 100×\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\ imes $$\\end{document}: 96.79% accuracy) and Single cells in Pap smear images for cervical cancer classification (SIPaKMeD: 99.18% accuracy). Also, the proposed method has been evaluated using five-fold cross validation and achieved improved performances on these datasets. Clearly, the proposed PND-Net effectively boosts the performances of automated health analysis of various plants in real and intricate field environments, implying PND-Net’s aptness for agricultural growth as well as human cancer classification.

Memetic micro-genetic algorithms for cancer data classification

Fast and precise medical diagnosis of human cancer is crucial for treatment decisions. Gene selection consists of identifying a set of informative genes from microarray data to allow high predictive accuracy in human cancer classification. This task is a combinatorial search problem, and optimisation methods can be applied for its resolution. In this paper, two memetic micro-genetic algorithms (MμV1 and MμV2) with different hybridisation approaches are proposed for feature selection of cancer microarray data. Seven gene expression datasets are used for experimentation. The comparison with stochastic state-of-the-art optimisation techniques concludes that problem-dependent local search methods combined with micro-genetic algorithms improve feature selection of cancer microarray data.

Leveraging the power for genomics for immuno-oncology

<div class="authors authors-lg"><div class="author"><div class="author-img"><img src="https://cdn.insights.bio/uploads/Golub headshot sq.jpg" data-image="1"></div><div class="author-det"> <!--StartFragment--><b>Todd Golub</b> is a founding core member of the Broad Institute and serves as the institute’s chief scientific officer and director of its Cancer Program. He is also the director of the Gerstner Center for Cancer Diagnostics at the Broad Institute. Dr Golub is a world leader in understanding the basis of cancer, by creating and applying tools of genomics. He has made fundamental discoveries in the molecular basis of childhood leukemia, and laid the foundation for the diagnosis and classification of human cancers using gene expression analysis. He also pioneered the development of chemical screening approaches based on gene expression. He joined the faculty of the Dana-Farber Cancer Institute and Harvard Medical School in 1997. At the same time he served as the leader of cancer genomics at the Whitehead Institute/MIT Center for Genome Research, which evolved into the Cancer Program at the Broad Institute. Dr Golub has directed that program since 2004, and is also currently the Charles A. Dana Investigator in Human Cancer Genetics at the Dana-Farber Cancer Institute and professor of pediatrics at Harvard Medical School. He is the recipient of multiple awards, including the Erasmus Hematology Award; the Richard and Hinda Rosenthal Memorial Award and the Outstanding Achievement Award from the American Association for Cancer Research; the Paul Marks Prize for Cancer Research; the E. Mead Johnson Award from the Society for Pediatric Research; and the Judson Daland Prize for Outstanding Achievement in Clinical Investigation from the American Philosophical Society. In 2014, he was elected as a member to the National Academy of Medicine (formerly the Institute of Medicine). Dr Golub has served on the scientific advisory boards of numerous organizations including serving as chair of St. Jude Children’s Research Hospital and the Board of Scientific Advisors of the National Cancer Institute. Golub received his B.A. from Carleton College and his M.D. from the University of Chicago Pritzker School of Medicine. He completed his internship, residency, and fellowship training at Boston Children’s Hospital and the Dana-Farber Cancer Institute.<!--EndFragment--> </div></div></div>

Complement System: Promoter or Suppressor of Cancer Progression?

Constituent of innate immunity, complement is present in the tumor microenvironment. The functions of complement include clearance of pathogens and maintenance of homeostasis, and as such could contribute to an anti-tumoral role in the context of certain cancers. However, multiple lines of evidence show that in many cancers, complement has pro-tumoral actions. The large number of complement molecules (over 30), the diversity of their functions (related or not to the complement cascade), and the variety of cancer types make the complement-cancer topic a very complex matter that has just started to be unraveled. With this review we highlight the context-dependent role of complement in cancer. Recent studies revealed that depending of the cancer type, complement can be pro or anti-tumoral and, even for the same type of cancer, different models presented opposite effects. We aim to clarify the current knowledge of the role of complement in human cancers and the insights from mouse models. Using our classification of human cancers based on the prognostic impact of the overexpression of complement genes, we emphasize the strong potential for therapeutic targeting the complement system in selected subgroups of cancer patients.

A study of data pre-processing techniques for imbalanced biomedical data classification

Biomedical data are widely accepted in developing prediction models for identifying a specific tumor, drug discovery and classification of human cancers. However, previous studies usually focused on different classifiers, and overlook the class imbalance problem in real-world biomedical datasets. There are a lack of studies on evaluation of data pre-processing techniques, such as resampling and feature selection, on imbalanced biomedical data learning. The relationship between data pre-processing techniques and the data distributions has never been analysed in previous studies. This article mainly focuses on reviewing and evaluating some popular and recently developed resampling and feature selection methods for class imbalance learning. We analyse the effectiveness of each technique from data distribution perspective. Extensive experiments have been done based on five classifiers, four performance measures, eight learning techniques across twenty real-world datasets. Experimental results show that: (1) resampling and feature selection techniques exhibit better performance using support vector machine (SVM) classifier. However, resampling and Feature Selection techniques perform poorly when using C4.5 decision tree and Linear discriminant analysis classifiers; (2) for datasets with different distributions, techniques such as Random undersampling and Feature Selection perform better than other data pre-processing methods with T Location-Scale distribution when using SVM and KNN (K-nearest neighbours) classifiers. Random oversampling outperforms other methods on Negative Binomial distribution using Random Forest classifier with lower level of imbalance ratio; (3) Feature Selection outperforms other data pre-processing methods in most cases, thus, Feature Selection with SVM classifier is the best choice for imbalanced biomedical data learning.

Genetic Analysis of Circulating Tumour Cells.

The classification of human cancers has traditionally relied on the tissue of origin, the histologic appearance and anatomical extent of disease, otherwise referred to as grade and stage. However, this system fails to explain the highly variable clinical behaviour seen for any one cancer. Molecular characterization through techniques such as next-generation sequencing (NGS) has led to an appreciation of the extreme genetic heterogeneity that underlies most human cancers. Because of the difficulties associated with fresh tissue biopsy, interest has increased in using circulating tumour material, such as circulating tumour cells (CTCs), as a non-invasive way to access tumour tissue. CTC enumeration has been demonstrated to have prognostic value in metastatic breast, colon and prostate cancers. Recent studies have also shown that CTCs are suitable material for molecular characterization, using techniques such as reverse transcription-polymerase chain reaction (RT-PCR), fluorescence in situ hybridization (FISH), array comparative genomic hybridization (aCGH) and NGS. Furthermore, genetic analysis of CTCs may be more suitable to study tumour heterogeneity and clonal evolution than fresh tissue biopsy. Whether blood-based biopsy techniques will be accepted as a replacement to fresh tissue biopsies remains to be seen, but there is reason for optimism. While significant barriers to this acceptance exist, blood-based biopsy techniques appear to be reliable and representative alternatives to fresh tissue biopsy.

65 YEARS OF THE DOUBLE HELIX: Classification of endocrine tumors in the age of integrated genomics.

The classification of human cancers represents one of the cornerstones of modern pathology. Over the last century, surgical pathologists established the current taxonomy of neoplasia using traditional histopathological parameters, which include tumor architecture, cytological features and cellular proliferation. This morphological classification is efficient and robust with high reproducibility and has served patients and health care providers well. The most recent decade has witnessed an explosion of genome-wide molecular genetic and epigenetic data for most cancers, including tumors of endocrine organs. The availability of this expansive multi-dimensional genomic data, collectively termed the cancer genome, has catalyzed a re-examination of the classification of endocrine tumors. Here, recent cancer genome studies of various endocrine tumors, including those of the thyroid, pituitary and adrenal glands, pancreas, small bowel, lung and skin, are presented with special emphasis on how genomic insights are impacting endocrine tumor classification.

Pathological bases for a robust application of cancer molecular classification.

Any robust classification system depends on its purpose and must refer to accepted standards, its strength relying on predictive values and a careful consideration of known factors that can affect its reliability. In this context, a molecular classification of human cancer must refer to the current gold standard (histological classification) and try to improve it with key prognosticators for metastatic potential, staging and grading. Although organ-specific examples have been published based on proteomics, transcriptomics and genomics evaluations, the most popular approach uses gene expression analysis as a direct correlate of cellular differentiation, which represents the key feature of the histological classification. RNA is a labile molecule that varies significantly according with the preservation protocol, its transcription reflect the adaptation of the tumor cells to the microenvironment, it can be passed through mechanisms of intercellular transference of genetic information (exosomes), and it is exposed to epigenetic modifications. More robust classifications should be based on stable molecules, at the genetic level represented by DNA to improve reliability, and its analysis must deal with the concept of intratumoral heterogeneity, which is at the origin of tumor progression and is the byproduct of the selection process during the clonal expansion and progression of neoplasms. The simultaneous analysis of multiple DNA targets and next generation sequencing offer the best practical approach for an analytical genomic classification of tumors.

Classification of Cancer Primary Sites Using Machine Learning and Somatic Mutations.

An accurate classification of human cancer, including its primary site, is important for better understanding of cancer and effective therapeutic strategies development. The available big data of somatic mutations provides us a great opportunity to investigate cancer classification using machine learning. Here, we explored the patterns of 1,760,846 somatic mutations identified from 230,255 cancer patients along with gene function information using support vector machine. Specifically, we performed a multiclass classification experiment over the 17 tumor sites using the gene symbol, somatic mutation, chromosome, and gene functional pathway as predictors for 6,751 subjects. The performance of the baseline using only gene features is 0.57 in accuracy. It was improved to 0.62 when adding the information of mutation and chromosome. Among the predictable primary tumor sites, the prediction of five primary sites (large intestine, liver, skin, pancreas, and lung) could achieve the performance with more than 0.70 in F-measure. The model of the large intestine ranked the first with 0.87 in F-measure. The results demonstrate that the somatic mutation information is useful for prediction of primary tumor sites with machine learning modeling. To our knowledge, this study is the first investigation of the primary sites classification using machine learning and somatic mutation data.

A feature selection method using fixed-point algorithm for DNA microarray gene expression data

As the performance of hardware is limited, the focus has been to develop objective, optimized and computationally efficient algorithms for a given task. To this extent, fixed-point and approximate algorithms have been developed and successfully applied in many areas of research. In this paper we propose a feature selection method based on fixed-point algorithm and show its application in the field of human cancer classification using DNA microarray gene expression data. In the fixed-point algorithm, we utilize between-class scatter matrix to compute the leading eigenvector. This eigenvector has been used to select genes. In the computation of the eigenvector, the eigenvalue decomposition of the scatter matrix is not required which significantly reduces its computational complexity and memory requirement.

A feature selection method using improved regularized linear discriminant analysis

Investigation of genes, using data analysis and computer-based methods, has gained widespread attention in solving human cancer classification problem. DNA microarray gene expression datasets are readily utilized for this purpose. In this paper, we propose a feature selection method using improved regularized linear discriminant analysis technique to select important genes, crucial for human cancer classification problem. The experiment is conducted on several DNA microarray gene expression datasets and promising results are obtained when compared with several other existing feature selection methods.

The specificity and patterns of staining in human cells and tissues of p16INK4a antibodies demonstrate variant antigen binding.

The validity of the identification and classification of human cancer using antibodies to detect biomarker proteins depends upon antibody specificity. Antibodies that bind to the tumour-suppressor protein p16INK4a are widely used for cancer diagnosis and research. In this study we examined the specificity of four commercially available anti-p16INK4a antibodies in four immunological applications. The antibodies H-156 and JC8 detected the same 16 kDa protein in western blot and immunoprecipitation tests, whereas the antibody F-12 did not detect any protein in western blot analysis or capture a protein that could be recognised by the H-156 antibody. In immunocytochemistry tests, the antibodies JC8 and H-156 detected a predominately cytoplasmic localised antigen, whose signal was depleted in p16INK4a siRNA experiments. F-12, in contrast, detected a predominately nuclear located antigen and there was no noticeable reduction in this signal after siRNA knockdown. Furthermore in immunohistochemistry tests, F-12 generated a different pattern of staining compared to the JC8 and E6H4 antibodies. These results demonstrate that three out of four commercially available p16INK4a antibodies are specific to, and indicate a mainly cytoplasmic localisation for, the p16INK4a protein. The F-12 antibody, which has been widely used in previous studies, gave different results to the other antibodies and did not demonstrate specificity to human p16INK4a. This work emphasizes the importance of the validation of commercial antibodies, aside to the previously reported use, for the full verification of immunoreaction specificity.

Blinded Comparator Study of Immunohistochemical Analysis versus a 92-Gene Cancer Classifier in the Diagnosis of the Primary Site in Metastatic Tumors

Accurate tumor classification is fundamental to inform predictive biomarker testing and optimize therapy. Gene expression-based tests are proposed as diagnostic aids in cases with uncertain diagnoses. This study directly compared the diagnostic accuracy of IHC analysis versus molecular classification using a 92-gene RT-PCR assay for determination of the primary tumor site. This prospectively defined blinded study of diagnostically challenging cases included 131 high-grade, primarily metastatic tumors. Cases were reviewed and reference diagnoses established through clinical correlation. Blinded FFPE sections were evaluated by either IHC/morphology analysis or the 92-gene assay. The final analysis included 122 cases. The 92-gene assay demonstrated overall accuracy of 79% (95% CI, 71% to 85%) for tumor classification versus 69% (95% CI, 60% to 76%) for IHC/morphology analysis (P = 0.019). Mean IHC use was 7.9 stains per case (median, 8; range, 2 to 15). IHC/morphology analysis accuracy was 79%, 80%, and 46% when 1 to 6 (n = 42), 7 to 9 (n = 41), and >9 (n = 39) IHC stains were used, respectively, versus 81%, 85%, and 69%, respectively, with the 92-gene assay. Results from this blinded series of high-grade metastatic cases demonstrate superior accuracy with the 92-gene assay versus standard-of-care IHC analysis and strongly support the diagnostic utility of molecular classification in difficult-to-diagnose metastatic cancer.

A Gene Selection Algorithm using Bayesian Classification Approach

In this study, we propose a new feature (or gene) selection algorithm using Bayes classification approach. The algorithm can find gene subset crucial for cancer classification problem. Problem statement: Gene identification plays important role in human cancer classification problem. Several feature selection algorithms have been proposed for analyzing and understanding influential genes using gene expression profiles. Approach: The feature selection algorithms aim to explore genes that are crucial for accurate cancer classification and also endure biological significance. However, the performance of the algorithms is still limited. In this study, we propose a feature selection algorithm using Bayesian classification approach. Results: This approach gives promising results on gene expression datasets and compares favorably with respect to several other existing techniques. Conclusion: The proposed gene selection algorithm using Bayes classification approach is shown to find important genes that can provide high classification accuracy on DNA microarray gene expression datasets.

Transgenic IGF-IR overexpression induces mammary tumors with basal-like characteristics, whereas IGF-IR-independent mammary tumors express a claudin-low gene signature

Molecular profiling has allowed a more precise classification of human cancers. With respect to breast cancer, this approach has been used to identify five subtypes; luminal A, luminal B, HER2-enriched, basal-like and claudin-low. In addition, this approach can be used to determine the type of tumor represented by particular cell lines or transgenic animal models. Therefore, this approach was utilized to classify the mammary tumors that develop in MTB-IGFIR transgenic mice. It was determined that the primary mammary tumors, which develop due to elevated expression of the type I insulin-like growth factor receptor (IGF-IR) in mammary epithelial cells, most closely resemble murine tumors with basal-like or mixed gene expression profiles and with human basal-like breast cancers. Downregulation of IGF-IR transgene in MTB-IGFIR tumor-bearing mice leads to the regression of most of the tumors, followed by tumor reappearance in some of the mice. These tumors that reappear following IGF-IR transgene downregulation do not express the IGF-IR transgene and cluster with murine mammary tumors that express a mesenchymal gene expression profile and with human claudin-low breast cancers. Therefore, IGF-IR overexpression in murine mammary epithelial cells induces mammary tumors with primarily basal-like characteristics, whereas tumors that develop following IGF-IR downregulation express a gene signature that most closely resembles human claudin-low breast tumors.

Performance and Clinical Evaluation of the 92-Gene Real-Time PCR Assay for Tumor Classification

Accurate determination of cancer origin is necessary to guide optimal treatment but remains a diagnostic challenge. Gene expression profiling technologies have aided the classification of tumors and, therefore, could be applied in conjunction with clinicopathologic correlates to improve accuracy. We report an expanded version of the previously described 92-gene assay to classify 30 main tumor types and 54 histological subtypes, with coverage of ≥95% of all solid tumors based on incidence. Increased tissue coverage was achieved through expansion of a reference tumor database containing 2,206 specimens, with a median of 62 samples per main tumor type. The 92-gene classification algorithm demonstrated sensitivities of 87% and 85% for 30 main types and 54 histological subtypes, respectively, in leave-one-out cross validation, and 83% in a test set of 187 tumors representing 28 of the 30 main cancer types. These findings provide further support that broad and diverse tumor classification can be performed using a relatively compact gene set. An additional 300 consecutive cases submitted for clinical testing were profiled to characterize clinical utility in a real-world setting: the 92-gene assay confirmed 78% of samples having a single suspected primary tumor and provided a single molecular prediction in 74% of cases with two or more differential diagnoses. Further development of the 92-gene RT-PCR assay has resulted in a significant expansion in reportable tumor types and histological features with strong performance characteristics and supports the use of molecular classification as an objective standardized adjunct to current methods.

Human Cancer Classification: A Systems Biology- Based Model Integrating Morphology, Cancer Stem Cells, Proteomics, and Genomics

Human cancer classification is currently based on the idea of cell of origin, light and electron microscopic attributes of the cancer. What is not yet integrated into cancer classification are the functional attributes of these cancer cells. Recent innovative techniques in biology have provided a wealth of information on the genomic, transcriptomic and proteomic changes in cancer cells. The emergence of the concept of cancer stem cells needs to be included in a classification model to capture the known attributes of cancer stem cells and their potential contribution to treatment response, and metastases. The integrated model of cancer classification presented here incorporates all morphology, cancer stem cell contributions, genetic, and functional attributes of cancer. Integrated cancer classification models could eliminate the unclassifiable cancers as used in current classifications. Future cancer treatment may be advanced by using an integrated model of cancer classification.

Performance enhancement of extreme learning machine for multi-category sparse data classification problems

This paper presents a performance enhancement scheme for the recently developed extreme learning machine (ELM) for multi-category sparse data classification problems. ELM is a single hidden layer neural network with good generalization capabilities and extremely fast learning capacity. In ELM, the input weights are randomly chosen and the output weights are analytically calculated. The generalization performance of the ELM algorithm for sparse data classification problem depends critically on three free parameters. They are, the number of hidden neurons, the input weights and the bias values which need to be optimally chosen. Selection of these parameters for the best performance of ELM involves a complex optimization problem. In this paper, we present a new, real-coded genetic algorithm approach called ‘RCGA-ELM’ to select the optimal number of hidden neurons, input weights and bias values which results in better performance. Two new genetic operators called ‘network based operator’ and ‘weight based operator’ are proposed to find a compact network with higher generalization performance. We also present an alternate and less computationally intensive approach called ‘sparse-ELM’. Sparse-ELM searches for the best parameters of ELM using K-fold validation. A multi-class human cancer classification problem using micro-array gene expression data (which is sparse), is used for evaluating the performance of the two schemes. Results indicate that the proposed RCGA-ELM and sparse-ELM significantly improve ELM performance for sparse multi-category classification problems.

Classification of human cancers based on DNA copy number amplification modeling

BackgroundDNA amplifications alter gene dosage in cancer genomes by multiplying the gene copy number. Amplifications are quintessential in a considerable number of advanced cancers of various anatomical locations. The aims of this study were to classify human cancers based on their amplification patterns, explore the biological and clinical fundamentals behind their amplification-pattern based classification, and understand the characteristics in human genomic architecture that associate with amplification mechanisms.MethodsWe applied a machine learning approach to model DNA copy number amplifications using a data set of binary amplification records at chromosome sub-band resolution from 4400 cases that represent 82 cancer types. Amplification data was fused with background data: clinical, histological and biological classifications, and cytogenetic annotations. Statistical hypothesis testing was used to mine associations between the data sets.ResultsProbabilistic clustering of each chromosome identified 111 amplification models and divided the cancer cases into clusters. The distribution of classification terms in the amplification-model based clustering of cancer cases revealed cancer classes that were associated with specific DNA copy number amplification models. Amplification patterns – finite or bounded descriptions of the ranges of the amplifications in the chromosome – were extracted from the clustered data and expressed according to the original cytogenetic nomenclature. This was achieved by maximal frequent itemset mining using the cluster-specific data sets. The boundaries of amplification patterns were shown to be enriched with fragile sites, telomeres, centromeres, and light chromosome bands.ConclusionsOur results demonstrate that amplifications are non-random chromosomal changes and specifically selected in tumor tissue microenvironment. Furthermore, statistical evidence showed that specific chromosomal features co-localize with amplification breakpoints and link them in the amplification process.

Differentially expressed genes are marked by histone 3 lysine 9 trimethylation in human cancer cells

Histone H3 lysine 9 trimethylation (H3K9Me3) has been associated with transcriptional repression, but recent findings implicate this chromatin modification in transcriptional activation and mRNA elongation by RNA polymerase II. Here, we applied immunoprecipitation (IP) with a custom DNA tiling microarray containing many transcription factors important in development and cancer (for example homeotic genes; N=683 total genes) to explore the relationship between H3K9Me3 and other histone modifications with the differential expression of genes. Cancer cell lines derived from different tissues (2 leukemia, 2 medulloblastoma) were characterized with IP antibodies to H3K9Me3, H3K4 dimethylation (H3K4Me2) and H3K9 acetylation (H3K9Ac). MV4-11 is known to overexpress the HOXA9 and MEIS1 genes, whereas D283 overexpresses the OTX2 homeobox gene. Gene expression was assessed by Affymetrix U133 array. Mapping the number and size of histone markings demonstrated significant colocalization of H3K9Ac and H3K4Me2 with H3K9Me3, indicating a pattern of putative 'activating' and 'repressive' markings. The median site size was 600-821 bp and 72-95% or 53-80% of chromatin signal sites were located within 1 kb or 500 bp of transcription start sites (TSS), respectively. A relatively small number of genes displayed additional H3K9Me3 sites in the 5'-region distant from the TSS. Comparing genes with modification sites to those without sites in their promoters confirmed the positive associations of H3K9Ac and H3K4Me2 with gene expression and revealed that H3K9Me3 is associated with active genes rather than being a repressive marking as previously thought. The positive regulatory effect of all three types of modifications were quantitatively correlated with site size, and applied to absolute gene expression within a single cell line as well as relative expression among pairs of cell lines. Extended patterns of H3K9Me3 upstream of some genes (for example HOXA9 and OTX2) may result from the action of multiple promoter elements. We found an inverse relationship between promoter DNA hypermethylation and H3K9Me3 in three studied genes (HOXA9, TMS1, RASSF1A). The localization of H3K9Me3 downstream of the TSSs of expressed genes and not within promoter regions of hypermethylated and silenced genes is consistent with the proposed coupling of H3K9Me3 with RNA polymerase II. Our results indicate a need for revising aspects of the histone code involving H3 lysine methylation. Awareness of H3K9Me3 as a mark of gene activity, not repression, is especially important for the classification of human cancer using chromatin and histone profiles.