In the last decade, advances in high-throughput technologies such as DNA microarrays have made it possible to simultaneously measure the expression levels of tens of thousands of genes and proteins. This has resulted in large amounts of biological data requiring analysis and interpretation. Nonnegative matrix factorization (NMF) was introduced as an unsupervised, parts-based learning paradigm involving the decomposition of a nonnegative matrix V into two nonnegative matrices, W and H, via a multiplicative updates algorithm. In the context of a p×n gene expression matrix V consisting of observations on p genes from n samples, each column of W defines a metagene, and each column of H represents the metagene expression pattern of the corresponding sample. NMF has been primarily applied in an unsupervised setting in image and natural language processing. More recently, it has been successfully utilized in a variety of applications in computational biology. Examples include molecular pattern discovery, class comparison and prediction, cross-platform and cross-species analysis, functional characterization of genes and biomedical informatics. In this paper, we review this method as a data analytical and interpretive tool in computational biology with an emphasis on these applications.

Background The tumor immune microenvironment has become the focus of research in clear cell renal cell carcinoma (ccRCC) due to its important role in immune surveillance post nephrectomy. This study investigates the correlation of tumor infiltrating immune cell characteristics with rates of recurrence following surgery in localized ccRCC. Methods We morphologically identified and scored tumor infiltrating lymphocytes (TILs) in hematoxylin and eosin (H&E) stained slides of patients with localized ccRCC (stage ≥T1b excluding stage IV). The University of Alabama at Birmingham (UAB) dataset (n = 159) was used to discover and the Fox Chase Cancer Center (FCCC) dataset (n = 198) was used to validate the results of morphologic immune cell analysis. We then performed gene expression analysis using the Immune Profile panel by NanoString in the UAB cohort and identified immune cells and pathways associated with recurrence, followed by validation in the Cancer Genome Atlas (TCGA) ccRCC dataset. Infiltrating immune cell types were identified by gene expression deconvolution. Results The presence of TILs identified by morphology correlated with higher T cell, Th1, CD8+ T and Treg gene signatures. Recurrence was associated with lower T cells and higher neutrophils. Higher Teffector (Teff)/Treg ratio correlated with lower rate of recurrence and was validated in the TCGA dataset. Genes associated with adaptive immune response were downregulated in tumors that recurred. Unsupervised hierarchical clustering identified a subset of patients with over-expression of adaptive response genes including CD8, CD3, GZMA/B, PRF1, IDO1, CTLA4, PDL1, ICOS and TIGIT. These patients had higher morphologic lymphocyte infiltration and T cell gene expression. Higher levels of TILs identified by morphology correlated with higher rates of recurrence in our discovery dataset but not in our validation set. Conclusions Recurrence of ccRCC following surgery was associated with lower T cell infiltrate, lower adaptive immune response and higher neutrophil gene expression. Presence of higher Teff/Treg ratio correlated with lower recurrence.

Nonnegative matrix factorization (NMF) by the multiplicative updates algorithm is a powerful machine learning method for decomposing a high-dimensional nonnegative matrix V into two nonnegative matrices, W and H, where [ ]. It has been successfully applied in the analysis and interpretation of large-scale data arising in neuroscience, computational biology, and natural language processing, among other areas. A distinctive feature of NMF is its nonnegativity constraints that allow only additive linear combinations of the data, thus enabling it to learn parts that have distinct physical representations in reality. In this letter, we describe an information-theoretic approach to NMF for signal-dependent noise based on the generalized inverse gaussian model. Specifically, we propose three novel algorithms in this setting, each based on multiplicative updates, and prove monotonicity of updates using the EM algorithm. In addition, we develop algorithm-specific measures to evaluate their goodness of fit on data. Our methods are demonstrated using experimental data from electromyography studies, as well as simulated data in the extraction of muscle synergies, and compared with existing algorithms for signal-dependent noise.

Recent genome-wide profiling by sequencing and distinctive chromatin signatures has identified thousands of long non-coding RNA (lncRNA) species (>200 nt). LncRNAs have emerged as important regulators of gene expression, involving in both developmental and pathological processes. While altered expression of lncRNAs has been observed in breast cancer development, their roles in breast cancer progression and metastasis are still poorly understood. To identify novel breast cancer-associated lncRNA candidates, we employed a high-density SNP array-based approach to uncover intergenic lncRNA genes that are aberrantly expressed in breast cancer. We first evaluated the potential value as a breast cancer prognostic biomarker for one breast cancer-associated lncRNA, LincIN, using a breast cancer cohort retrieved from The Cancer Genome Atlas (TCGA) Data Portal. Then we characterized the role of LincIN in breast cancer progression and metastasis by in vitro invasion assay and a mouse tail vein injection metastasis model. To study the action of LincIN, we identified LincIN-interacting protein partner(s) by RNA pull-down experiments followed with protein identification by mass spectrometry. High levels of LincIN expression are frequently observed in tumors compared to adjacent normal tissues, and are strongly associated with aggressive breast cancer. Importantly, analysis of TCGA data further suggest that high expression of LincIN is associated with poor overall survival in patients with breast cancer (P = 0.044 and P = 0.011 after adjustment for age). The functional experiments demonstrate that knockdown of LincIN inhibits tumor cell migration and invasion in vitro, which is supported by the results of transcriptome analysis in the LincIN-knockdown cells. Furthermore, knockdown of LincIN diminishes lung metastasis in a mouse tail vein injection model. We also identified a LincIN-binding protein, NF90, through which overexpression of LincIN may repress p21 protein expression by inhibiting its translation, and upregulation of p21 by LincIN knockdown may be associated with less aggressive metastasis phenotypes. Our studies provide clear evidence to support LincIN as a new regulator of tumor progression-metastasis at both transcriptional and translational levels and as a promising prognostic biomarker for breast cancer.