Generalized linear models with nonlinear feature transformations are widely used for large-scale regression and classification problems with sparse inputs. Memorization of feature interactions through a wide set of cross-product feature transformations are effective and interpretable, while generalization requires more feature engineering effort. With less feature engineering, deep neural networks can generalize better to unseen feature combinations through low-dimensional dense embeddings learned for the sparse features. However, deep neural networks with embeddings can over-generalize and recommend less relevant items when the user-item interactions are sparse and high-rank. In this paper, we present Wide & Deep learning-jointly trained wide linear models and deep neural networks-to combine the benefits of memorization and generalization for recommender systems. We productionized and evaluated the system on Google Play, a commercial mobile app store with over one billion active users and over one million apps. Online experiment results show that Wide & Deep significantly increased app acquisitions compared with wide-only and deep-only models. We have also open-sourced our implementation in TensorFlow.

Purpose: The aim of this study was to evaluate the microRNA expression patterns in squamous cell carcinoma (SCC) of the tongue. Experimental Design: Expression levels of 156 human mature microRNAs were examined using real-time quantitative PCR (Taq Man MicroRNA Assays; Human Panel) on laser microdissected cells of 4 tongue carcinomas and paired normal tissues. Expression of mature miR-184 was further validated in 20 paired tongue SCC and the normal tissues. Potential oncogenic functions of miR-184 were evaluated in tongue SCC cell lines (Cal27, HN21B, and HN96) with miR-184 inhibitor. Plasma miR-184 levels were evaluated using real-time quantitative PCR. Results: Using 3-fold expression difference as a cutoff level, we identified 24 up-regulated mature miRNAs including miR-184, miR-34c, miR-137, miR-372, miR-124a, miR-21, miR-124b, miR-31, miR-128a, miR-34b, miR-154, miR-197, miR-132, miR-147, miR-325, miR-181c, miR-198, miR-155, miR-30a-3p, miR-338, miR-17-5p, miR-104, miR-134, and miR-213; and 13 down-regulated mature miRNAs including miR-133a, miR-99a, miR-194, miR-133b, miR-219, miR-100, miR-125b, miR-26b, miR-138, miR-149, miR-195, miR-107, and miR-139. Overexpression of miR-184 was further validated in 20 paired tongue SCC and normal tissues (P = 0.002). Inhibition of miR-184 in tongue SCC cell lines could reduce cell proliferation rate. Down-regulation of c-Myc was observed in two cell lines in response to miR-184 inhibitor. Suppressing miR-184 could induce apoptosis in all three cell lines. Plasma miR-184 levels were significantly higher in tongue SCC patients in comparison with normal individuals, and the levels were significantly reduced after surgical removal of the primary tumors. Conclusions: Overexpression of miR-184 might play an oncogenic role in the antiapoptotic and proliferative processes of tongue SCC. In addition, plasma miR-184 levels were associated with the presence of primary tumor. Further studies on the aberrantly expressed miRNAs in tongue SCC as well as using plasma miRNAs as novel tumor markers are warranted.

MicroRNAs (miRNAs) are noncoding RNAs with specific regulatory role in gene expression. Recent reports suggested their involvement in human malignancies. Currently, there is no information concerning miRNA expression and functions in squamous cell carcinoma (SCC) of tongue. In this study, we evaluated the expression patterns of 156 mature miRNAs in tongue SCC using Taqman-based microRNA assays. Of these 156 miRNAs, miR133a and miR-133b were significantly reduced in tongue SCC cells in comparison with the paired normal epithelial cells. Tongue SCC cell lines transfected with miR-133a and miR-133b precursors displayed reduction in proliferation rate. In addition, the number of apoptotic cells was increased in response to the introduction of precursors. Computational target gene prediction suggested that both miR-133a and miR-133b are targeting transcript of pyruvate kinase type M2 (PKM2), a potential oncogene in solid cancers. In tongue SCC cell lines, PKM2 expression was reduced in response to miR-133a and miR-133b precursors transfection. Immunohistochemical staining results of tongue SCC tissues suggested that PKM2 was overexpressed in tongue SCC and was associated with the downregulation of miR-133a and miR-133b. Our results suggested that aberrant reduction of miR-133a and miR133b was associated with the dysregulation of PKM2 in SCC of tongue. ' 2008 Wiley-Liss, Inc.Key words: microRNAs; miR-133a; miR-133b; tumor suppressor; squamous cell carcinoma of tongue MicroRNAs (miRNAs) are endogenous and noncoding RNA molecules. The primary miRNAs are long RNA molecules. This long RNA molecules are processed by endogeneous nucleases and are actively transported into the cytoplasm.1 Cytoplasmic miRNAs are further processed by Dicer into mature and functional miRNAs (single-stranded; 22 nucleotides). Mature miRNAs are associated with a cellular complex that is similar to the RNA-induced silencing complex that participates in RNA interference.2,3 Mature miRNAs could bind onto specific mRNA molecules. The specific bindings could promote degradation of target mRNA and/or hinder the translation process. Thus, the dynamic of miRNA expression is directly influencing the mRNA expression patterns of the cells.Recent studies revealed the important role of miRNA in cancer development and malignant transformation. In cancer cells, miRNA expression patterns are usually altered.4 miRNAs with tumor-suppressive function are usually suppressed/reduced in cancer cells. In chronic lymphatic leukemia, expression levels of tumor-suppressing miR-15-1 and miR-16-1 were significantly reduced. miR-15-1 and miR-16-1 function as tumor suppressors by regulating BCL2. The reduction of miR-15-1 and miR-16-1 could confer selective advantages to cancer cells in the transforming process. 5 miRNA expression patterns are tissue-specific. Using the miRNA expression patterns, Visone et al. could distinguish papillary thyroid carcinoma from normal thyroid cells. 6 We therefore suggested that there is a distinctive miRNA expression patterns in carcinoma of...