Exploring the spatial reasoning ability of neural models in human IQ tests.

Although neural models have performed impressively well on various tasks such as image recognition and question answering, their reasoning ability has been measured in only few studies. In this work, we focus on spatial reasoning and explore the spatial understanding of neural models. First, we describe the following two spatial reasoning IQ tests: rotation and shape composition. Using well-defined rules, we constructed datasets that consist of various complexity levels. We designed a variety of experiments in terms of generalization, and evaluated six different baseline models on the newly generated datasets. We provide an analysis of the results and factors that affect the generalization abilities of models. Also, we analyze how neural models solve spatial reasoning tests with visual aids. We hope that our work can encourage further research into human-level spatial reasoning and provide a new direction for future work.

Enhancing the interpretability of transcription factor binding site prediction using attention mechanism.

Transcription factors (TFs) regulate the gene expression of their target genes by binding to the regulatory sequences of target genes (e.g., promoters and enhancers). To fully understand gene regulatory mechanisms, it is crucial to decipher the relationships between TFs and DNA sequences. Moreover, studies such as GWAS and eQTL have verified that most disease-related variants exist in non-coding regions, and highlighted the necessity to identify such variants that cause diseases by interrupting TF binding mechanisms. To do this, it is necessary to build a prediction model that precisely predicts the binding relationships between TFs and DNA sequences. Recently, deep learning based models have been proposed and have shown competitive results on a transcription factor binding site prediction task. However, it is difficult to interpret the prediction results obtained from the previous models. In addition, the previous models assumed all the sequence regions in the input DNA sequence have the same importance for predicting TF-binding, although sequence regions containing TF-binding-associated signals such as TF-binding motifs should be captured more than other regions. To address these challenges, we propose TBiNet, an attention based interpretable deep neural network for predicting transcription factor binding sites. Using the attention mechanism, our method is able to assign more importance on the actual TF binding sites in the input DNA sequence. TBiNet outperforms the current state-of-the-art methods (DeepSea and DanQ) quantitatively in the TF-DNA binding prediction task. Moreover, TBiNet is more effective than the previous models in discovering known TF-binding motifs.

KLF5 enhances SREBP-1 action in androgen-dependent induction of fatty acid synthase in prostate cancer cells.

KLF5 (Krüppel-like factor 5) is a zinc-finger transcription factor that plays a critical role in the regulation of cellular signalling involved in cell proliferation, differentiation and oncogenesis. In the present study, we showed that KLF5 acts as a key regulator controlling the expression of FASN (fatty acid synthase) through an interaction with SREBP-1 (sterol-regulatory-element-binding protein-1) in the androgen-dependent LNCaP prostate cancer cell line. The mRNA level of KLF5 increased when cells were treated with a synthetic androgen, R1881. Furthermore, KLF5 bound to SREBP-1 and enhanced the SREBP-1-mediated increase in FASN promoter activity. The results also demonstrated that the expression of KLF5 in LNCaP prostate cancer cells enhanced FASN expression, whereas silencing of KLF5 by small interfering RNA down-regulated FASN expression. The proximal promoter region and the first intron of the FASN gene contain multiple CACCC elements that mediate the transcriptional regulation of the gene by KLF5. However, other lipogenic and cholesterogenic genes, such as those encoding acetyl-CoA carboxylase, ATP-citrate lyase, the LDL (low-density lipoprotein) receptor, HMG-CoA (3-hydroxy-3-methylglutaryl-CoA) synthase and HMG-CoA reductase are irresponsive to KLF5 expression, owing to the absence of CACCC elements in their promoter regions. Taken together, these results suggest that the FASN gene is activated by the synergistic action of KLF5 and SREBP-1, which was induced by androgen in androgen-dependent prostate cancer cells.

Lipin1 is a key factor for the maturation and maintenance of adipocytes in the regulatory network with CCAAT/enhancer-binding protein alpha and peroxisome proliferator-activated receptor gamma 2.

Lipin1 expression was induced at a late stage of differentiation of 3T3-L1 preadipocytes and maintained at high levels in mature adipocytes. Knockdown of expression of lipin1 by small interfering RNA in 3T3-L1 preadipocytes almost completely inhibited differentiation into adipocytes, whereas overexpression of lipin1 accelerated adipocyte differentiation, demonstrating that lipin1 is required for adipocyte differentiation. In mature adipocytes, transfection of lipin1-small interfering RNA decreased the expression of adipocyte functional genes, indicating the involvement of lipin1 in the maintenance of adipocyte function. Lipin1 increases the transcription-activating function of peroxisome proliferator-activated receptor gamma(2) (PPAR gamma(2)) via direct physical interaction, whereas lipin1 did not affect the function of other adipocyte-related transcription factors such as C/EBP alpha, liver X-activated receptor alpha, or sterol regulatory element binding protein 1c. In mature adipocytes, lipin1 was specifically recruited to the PPAR gamma-response elements of the phosphoenolpyruvate carboxykinase gene, an adipocyte-specific gene. C/EBP alpha up-regulates lipin1 transcription by directly binding to the lipin1 promoter. Based on the existence of a positive feedback loop between C/EBP alpha and PPAR gamma(2), we propose that lipin1 functions as an amplifier of the network between these factors, resulting in the maintenance of high levels of the specific gene expression that are required for adipogenesis and mature adipocyte functions.

Transcriptional activation of SHP by PPAR-gamma in liver.

The mechanism of how PPARgamma decrease gluconeogenic gene expressions in liver is still unclear. Since PPARgamma is a transcriptional activator, it requires a mediator to decrease the transcription of gluconeogenic genes. Recently, SHP has been shown to mediate the bile acid-dependent down regulation of gluconeogenic gene expression in liver. This led us to explore the possibility that SHP may mediate the antigluconeogenic effect of PPARgamma. In the present study, we have identified and characterized the presence of functional PPRE in human SHP promoter. We show the binding of PPARgamma/RXRalpha heterodimer to the PPRE and increased SHP expression by rosiglitazone in primary rat hepatocytes. Taken together with the previous reports about the function of SHP on gluconeogenesis, our results indicate that SHP can mediate the acute antigluconeogenic effect of PPARgamma.

Up-regulation of acetyl-CoA carboxylase alpha and fatty acid synthase by human epidermal growth factor receptor 2 at the translational level in breast cancer cells.

Expression of the HER2 oncogene is increased in approximately 30% of human breast carcinomas and is closely correlated with the expression of fatty acid synthase (FASN). In the present study, we determined the mechanism by which FASN and acetyl-CoA carboxylase alpha (ACCalpha) could be induced by HER2 overexpression. SK-BR-3 and BT-474 cells, breast cancer cells that overexpress HER2, expressed higher levels of FASN and ACCalpha compared with MCF-7 and MDA-MB-231 breast cancer cells in which HER2 expression is low. The induction of FASN and ACCalpha in BT474 cells were not mediated by the activation of SREBP-1. Exogenous HER2 expression in MDA-MB-231 cells induced the expression of FASN and ACCalpha, and the HER2-mediated increase in ACCalpha and FASN was inhibited by both LY294002, a phosphatidylinositol 3-kinase inhibitor, and rapamycin, a mammalian target of rapamycin (mTOR) inhibitor. In addition, the activation of mTOR by the overexpression of RHEB in MDA-MB-231 cells increased the synthetic rates of both FASN and ACCalpha. On the other hand, FASN and ACCalpha were reduced in BT-474 cells by a blockade of the mTOR signaling pathway. These changes observed in their protein levels were not accompanied by changes in their mRNA levels. The 5'- and 3'-untranslated regions of both FASN and ACCalpha mRNAs were involved in selective translational induction that was mediated by mTOR signal transduction. These results strongly suggest that the major mechanism of HER2-mediated induction of FASN and ACCalpha in the breast cancer cells used in this study is translational regulation primarily through the mTOR signaling pathway.