Interplay among GATA transcription factors is an important determinant of cell fate during hematopoiesis. Although GATA-2 regulates hematopoietic stem cell function, mechanisms controlling GATA-2 expression are undefined. Of particular interest is the repression of GATA-2, because sustained GATA-2 expression in hematopoietic stem and progenitor cells alters hematopoiesis. GATA-2 transcription is derepressed in erythroid precursors lacking GATA-1, but the underlying mechanisms are unknown. Using chromatin immunoprecipitation analysis, we show that GATA-1 binds a highly restricted upstream region of the Ϸ70-kb GATA-2 domain, despite >80 GATA sites throughout the domain. GATA-2 also binds this region in the absence of GATA-1. Genetic complementation studies in GATA-1-null cells showed that GATA-1 rapidly displaces GATA-2, which is coupled to transcriptional repression. GATA-1 also displaces CREB-binding protein (CBP), despite the fact that GATA-1 binds CBP in other contexts. Repression correlates with reduced histone acetylation domain-wide, but not altered methylation of histone H3 at lysine 4. The GATA factor-binding region exhibited cell-type-specific enhancer activity in transient transfection assays. We propose that GATA-1 instigates GATA-2 repression by means of disruption of positive autoregulation, followed by establishment of a domain-wide repressive chromatin structure. Such a mechanism is predicted to be critical for the control of hematopoiesis.

In this paper, we address the 3D object detection task by capturing multi-level contextual information with the selfattention mechanism and multi-scale feature fusion. Most existing 3D object detection methods recognize objects individually, without giving any consideration on contextual information between these objects. Comparatively, we propose Multi-Level Context VoteNet (MLCVNet) to recognize 3D objects correlatively, building on the state-of-the-art VoteNet. We introduce three context modules into the voting and classifying stages of VoteNet to encode contextual information at different levels. Specifically, a Patch-to-Patch Context (PPC) module is employed to capture contextual information between the point patches, before voting for their corresponding object centroid points. Subsequently, an Object-to-Object Context (OOC) module is incorporated before the proposal and classification stage, to capture the contextual information between object candidates. Finally, a Global Scene Context (GSC) module is designed to learn the global scene context. We demonstrate these by capturing contextual information at patch, object and scene levels. Our method is an effective way to promote detection accuracy, achieving new state-of-the-art detection performance on challenging 3D object detection datasets, i.e., SUN RGBD and ScanNet. We also release our code at https://github.com/NUAAXQ/MLCVNet.

Elucidating mechanisms controlling nuclear processes requires an understanding of the nucleoprotein structure of genes at endogenous chromosomal loci. Traditional approaches to measuring protein-DNA interactions in vitro have often failed to provide insights into physiological mechanisms. Given that most transcription factors interact with simple DNA sequence motifs, which are abundantly distributed throughout a genome, it is essential to pinpoint the small subset of sites bound by factors in vivo. Signaling mechanisms induce the assembly and modulation of complex patterns of histone acetylation, methylation, phosphorylation, and ubiquitination, which are crucial determinants of chromatin accessibility. These seemingly complex issues can be directly addressed by a powerful methodology termed the chromatin immunoprecipitation (ChIP) assay. ChIP analysis involves covalently trapping endogenous proteins at chromatin sites, thereby yielding snapshots of protein-DNA interactions and histone modifications within living cells. The chromatin is sonicated to generate small fragments, and an immunoprecipitation is conducted with an antibody against the desired factor or histone modification. Crosslinks are reversed, and polymerase chain reaction (PCR) is used to assess whether DNA sequences are recovered immune-specifically. Chromatin-domain scanning coupled with quantitative analysis is a powerful means of dissecting mechanisms by which signaling pathways target genes within a complex genome.

The process whereby the primitive vascular network develops into the mature vasculature, known as angiogenic vascular remodeling, is controlled by the Notch signaling pathway. Of the two mammalian Notch receptors expressed in vascular endothelium, Notch1 is broadly expressed in diverse cell types, whereas Notch4 is preferentially expressed in endothelial cells. As mechanisms that confer Notch4 expression were unknown, we investigated how NOTCH4 transcription is regulated in human endothelial cells and in transgenic mice. The NOTCH4 promoter and the 5 portion of NOTCH4 assembled into an endothelial cell-specific histone modification pattern. Analysis of NOTCH4 primary transcripts in human umbilical vein endothelial cells by RNA fluorescence in situ hybridization revealed that 36% of the cells transcribed one or both NOTCH4 alleles. The NOTCH4 promoter was sufficient to confer endothelial cell-specific transcription in transfection assays, but intron 1 or upstream sequences were required for expression in the vasculature of transgenic mouse embryos. Cell-type-specific activator protein 1 (AP-1) complexes occupied NOTCH4 chromatin and conferred endothelial cell-specific transcription. Vascular angiogenic factors activated AP-1 and reprogrammed the endogenous NOTCH4 gene in HeLa cells from a repressed to a transcriptionally active state. These results reveal an AP-1-Notch4 pathway, which we propose to be crucial for transducing angiogenic signals and to be deregulated upon aberrant signal transduction in cancer.

Tree boosting, which combines weak learners (typically decision trees) to generate a strong learner, is a highly effective and widely used machine learning method. However, the development of a high performance tree boosting model is a time-consuming process that requires numerous trial-and-error experiments. To tackle this issue, we have developed a visual diagnosis tool, BOOSTVis, to help experts quickly analyze and diagnose the training process of tree boosting. In particular, we have designed a temporal confusion matrix visualization, and combined it with a t-SNE projection and a tree visualization. These visualization components work together to provide a comprehensive overview of a tree boosting model, and enable an effective diagnosis of an unsatisfactory training process. Two case studies that were conducted on the Otto Group Product Classification Challenge dataset demonstrate that BOOSTVis can provide informative feedback and guidance to improve understanding and diagnosis of tree boosting algorithms.