Machine learning with maximization (support) of separating margin (vector), called support vector machine (SVM) learning, is a powerful classification tool that has been used for cancer genomic classification or subtyping. Today, as advancements in high-throughput technologies lead to production of large amounts of genomic and epigenomic data, the classification feature of SVMs is expanding its use in cancer genomics, leading to the discovery of new biomarkers, new drug targets, and a better understanding of cancer driver genes. Herein we reviewed the recent progress of SVMs in cancer genomic studies. We intend to comprehend the strength of the SVM learning and its future perspective in cancer genomic applications.

Cancer is a genetic disease where genetic variations cause abnormally functioning genes that appear to alter expression. Proteins, the final products of gene expression, determine the phenotypes and biological processes. Therefore, detecting gene expression levels can be used for cancer diagnosis, prognosis, and treatment prediction in a clinical setting. In this review, we investigated six gene expression assay systems (qRT-PCR, DNA microarray, nCounter, RNA-Seq, FISH, and tissue microarray) that are currently being used in clinical cancer studies. Some of these methods are also commonly used in a modified way; for example, detection of DNA content or protein expression. Herein, we discuss their principles, sample preparation, design, quantification and sensitivity, data analysis, time for sample preparation and processing, and cost. We also compared these methods according to their sample selection, particularly for the feasibility of using formalin-fixed paraffin-embedded (FFPE) samples, which are routinely archived for clinical cancer studies. We intend to provide a guideline for choosing an assay method with respect to its oncological applications in a clinical setting.

Transcriptional regulation is impacted by multiple layers of genome organization. A general feature of transcriptionally active chromatin is sensitivity to DNase I and association with acetylated histones. However, very few of these active DNase I-sensitive domains, such as the chicken erythrocyte β-globin domain, have been identified and characterized. In chicken polychromatic erythrocytes, dynamically acetylated histones associated with DNase I-sensitive, transcriptionally active chromatin prevent histone H1/H5-induced insolubility at physiological ionic strength. Here, we identified and mapped out all the transcriptionally active chromosomal domains in the chicken polychromatic erythrocyte genome by combining a powerful chromatin fractionation method with next-generation DNA and RNA sequencing. Two classes of transcribed chromatin organizations were identified on the basis of the extent of solubility at physiological ionic strength. Highly transcribed genes were present in multigenic salt-soluble chromatin domains ranging in length from 30 to over 150 kb. We identified over 100 highly expressed genes that were organized in broad dynamically highly acetylated, salt-soluble chromatin domains. Highly expressed genes were associated with H3K4me3 and H3K27ac and produced discernible antisense transcripts. The moderately- and low-expressing genes had highly acetylated, salt-soluble chromatin regions confined to the 5′ end of the gene. Our data provide a genome-wide profile of chromatin signatures in relation to expression levels in chicken polychromatic erythrocytes.

Breast cancer is a heterogeneous disease and personalized medicine is the hope for the improvement of the clinical outcome. Multi-gene signatures for breast cancer stratification have been extensively studied in the past decades and more than 30 different signatures have been reported. A major concern is the minimal overlap of genes among the reported signatures. We investigated the breast cancer signature genes to address our hypothesis that the genes of different signature may share common functions, as well as to use these previously reported signature genes to build better prognostic models. A total of 33 signatures and the corresponding gene lists were investigated. We first examined the gene frequency and the gene overlap in these signatures. Then the gene functions of each signature gene list were analysed and compared by the KEGG pathways and gene ontology (GO) terms. A classifier built using the common genes was tested using the METABRIC (Molecular Taxonomy of Breast Cancer International Consortium) data. The common genes were also tested for building the Yin Yang gene mean expression ratio (YMR) signature using public datasets (GSE1456 and GSE2034). Among a total of 2239 genes collected from the 33 breast cancer signatures, only 238 genes overlapped in at least two signatures; while from a total of 1979 function terms enriched in the 33 signature gene lists, 429 terms were common in at least two signatures. Most of the common function terms were involved in cell cycle processes. While there is almost no common overlapping genes between signatures developed for ER-positive (e.g. 21-gene signature) and those developed for ER-negative (e.g. basal signatures) tumours, they have common function terms such as cell death, regulation of cell proliferation. We used the 62 genes that were common in at least three signatures as a classifier and subtyped 1141 METABRIC cases including 144 normal samples into nine subgroups. These subgroups showed different clinical outcome. Among the 238 common genes, we selected those genes that are more highly expressed in normal breast tissue than in tumours as Yang genes and those more highly expressed in tumours than in normal as Yin genes and built a YMR model signature. This YMR showed significance in risk stratification in two datasets (GSE1456 and GSE2034). The lack of significant numbers of overlapping genes among most breast cancer signatures can be partially explained by our discovery that these signature genes represent groups with similar functions. The genes collected from these previously reported signatures are valuable resources for new model development. The subtype classifier and YMR signature built from the common genes showed promising results.

Background: Emerging evidence has suggested that dysbiosis of the microbiota may play vital roles in tumorigenesis. However, the interplay between the microbiome and lung cancer remains undetermined. In this study, we characterize the microbiome in the early stage of lung cancer, which presented as ground-glass nodules (GGNs). Methods: We sequenced the whole genomes from 10 GGN lesions and 5 adjacent normal lung tissue samples. After being filtered with human genome sequences, the sequence reads were mapped to prokaryotic genomes refSeq and non-redundant protein database for taxa and gene functions profiling, respectively. Results: Mycobacterium, Corynebacterium, and Negativicoccus were the core microbiota found in all GGNs and the normal tissue samples. The microbiota composition did not show significant difference between GGNs and normal tissues except the adenocarcinoma (AD) (P=0.047). A significant β diversity in microbiome gene functions was found among different patients. Two individual gene functions, the Secondary Metabolism (1.32 fold with P=0.01) and the Serine Threonine protein kinase (4.23 fold, P<0.001), were significantly increased in GGNs over normal tissue samples. Conclusions: This study helps shed light on the implication of the microbiome in early stage lung cancer, which encourages the further study and development of innovative strategies for early prevention and treatment of lung cancer.

Triple Negative Breast Cancers (TNBCs) lack the appropriate targets for currently used breast cancer therapies, conferring an aggressive phenotype, more frequent relapse and poorer survival rates. The biological heterogeneity of TNBC complicates the clinical treatment further. We have explored and compared the biological pathways in TNBC and other subtypes of breast cancers, using an in silico approach and the hypothesis that two opposing effects (Yin and Yang) pathways in cancer cells determine the fate of cancer cells. Identifying breast subgroup specific components of these opposing pathways may aid in selecting potential therapeutic targets as well as further classifying the heterogeneous TNBC subtype. Gene expression and patient clinical data from The Cancer Genome Atlas (TCGA) and the Molecular Taxonomy of Breast Cancer International Consortium (METABRIC) were used for this study. Gene Set Enrichment Analysis (GSEA) was used to identify the more active pathways in cancer (Yin) than in normal and the more active pathways in normal (Yang) than in cancer. The clustering analysis was performed to compare pathways of TNBC with other types of breast cancers. The association of pathway classified TNBC sub-groups to clinical outcomes was tested using Cox regression model. Among 4729 curated canonical pathways in GSEA database, 133 Yin pathways (FDR < 0.05) and 71 Yang pathways (p-value <0.05) were discovered in TNBC. The FOXM1 is the top Yin pathway while PPARα is the top Yang pathway in TNBC. The TNBC and other types of breast cancers showed different pathways enrichment significance profiles. Using top Yin and Yang pathways as classifier, the TNBC can be further subtyped into six sub-groups each having different clinical outcomes. We first reported that the FOMX1 pathway is the most upregulated and the PPARα pathway is the most downregulated pathway in TNBC. These two pathways could be simultaneously targeted in further studies. Also the pathway classifier we performed in this study provided insight into the TNBC heterogeneity.