Adversarially Learning Occlusions by Backpropagation for Face Recognition.

With the accomplishment of deep neural networks, face recognition methods have achieved great success in research and are now being applied at a human level. However, existing face recognition models fail to achieve state-of-the-art performance in recognizing occluded face images, which are common scenarios captured in the real world. One of the potential reasons for this is the lack of large-scale training datasets, requiring labour-intensive and costly labelling of the occlusions. To resolve these issues, we propose an Adversarially Learning Occlusions by Backpropagation (ALOB) model, a simple yet powerful double-network framework used to mitigate manual labelling by contrastively learning the corrupted features against personal identity labels, thereby maximizing the loss. To investigate the performance of the proposed method, we compared our model to the existing state-of-the-art methods, which function under the supervision of occlusion learning, in various experiments. Extensive experimentation on LFW, AR, MFR2, and other synthetic masked or occluded datasets confirmed the effectiveness of the proposed model in occluded face recognition by sustaining better results in terms of masked face recognition and general face recognition. For the AR datasets, the ALOB model outperformed other advanced methods by obtaining a 100% recognition rate for images with sunglasses (protocols 1 and 2). We also achieved the highest accuracies of 94.87%, 92.05%, 78.93%, and 71.57% TAR@FAR = 1 × 10-3 in LFW-OCC-2.0 and LFW-OCC-3.0, respectively. Furthermore, the proposed method generalizes well in terms of FR and MFR, yielding superior results in three datasets, LFW, LFW-Masked, and MFR2, and producing accuracies of 98.77%, 97.62%, and 93.76%, respectively.

Physical Activity and Factors Affecting Its Maintenance Among Patients With Coronary Heart Disease Not Undergoing Cardiac Rehabilitation in China.

BACKGROUND: The level of physical activity (PA) among patients with coronary heart disease (CHD) living in Chinese communities who do not participate in cardiac rehabilitation programs and the factors contributing to patient maintenance of PA are unclear. OBJECTIVE: This cross-sectional study, guided by the Transtheoretical Model, evaluated (1) the maintenance of PA in Chinese patients with CHD 12 months after hospital discharge and (2) the demographic, clinical, and psychological characteristics associated with maintenance of PA. METHODS: A total of 1162 patients completed 6 questionnaires at 12 months posthospitalization to assess their maintenance of PA, stage of change, symptoms of depression and anxiety, and health-related quality of life and sleep. RESULTS: Only 40% of patients with CHD maintained regular PA 12 months after hospital discharge. Walking was their primary PA. Thirty-seven percent of patients reported no intention of having regular PA. Male sex (odds ratio [OR], 1.69), awareness of PA's cardiac benefit (OR, 4.12), a history of regular PA before the cardiac event (OR, 6.08), history of chronic disease (OR, 1.43), mild depressive symptoms (OR, 1.40), moderate and severe depressive symptoms (OR, 0.41), smoking (OR, 0.54), and years of CHD (OR, 0.96) were related to maintenance of regular PA. Patients with CHD who maintained regular PA had better quality of life and sleep (P < .001) and fewer unplanned clinic visits (P = .001) and cardiac cause readmissions (P = .012) and reported fewer declines in PA capacity (P < .001). CONCLUSIONS: Walking is the most common form of PA 12 months posthospitalization among patients with CHD in China. Patient education and counseling about the cardiac benefits of PA, taking into account stage of change, are important considerations to improve maintenance of PA.

Use of ¹9F NMR methods to probe conformational heterogeneity and dynamics of exchange in functional RNA molecules.

Functional RNA molecules are often very plastic and often undergo changes in base-pairing patterns to achieve alternative secondary and tertiary conformations associated with their roles in multiple events in gene expression. Solution NMR techniques are an excellent tool for the analysis of conformational heterogeneity and dynamic exchange. In this work, we measure the rates associated with spontaneous interconversion between major conformers in folded RNA sequences by use of a (19)F-(19)F EXSY NMR experiment, taking advantage of RNA samples carrying a single 5-(19)F-pyrimidine label. We first utilize this approach to determine kinetic exchange rates between conformers in a model RNA stem loop capable of adopting two conformations. We then probe the dynamics of conformational rearrangements in a larger RNA construct, the U2-U6 snRNA complex of the human spliceosome. In the case of the U2-U6 snRNA complex, such a rearrangement in the context of the intact spliceosome may have critical implications in splicing activity.

Measurement of chemical exchange between RNA conformers by 19F NMR.

Many noncoding RNA molecules adopt alternative secondary and tertiary conformations that are critical for their roles in gene expression. Although many of these rearrangements are mediated by other biomolecular components, it is important to evaluate the equilibrium relationship of the conformers. To measure the spontaneous interconversion in a bi-stable RNA stem loop sequence into which a single (19)F-uridine label was incorporated, a (19)F-(19)F EXSY experiment was employed. The kinetic exchange rate measured from EXSY experiments for this system was 37.3±2.8s(-1). The advantage of this approach is that exchange kinetics can be monitored in any RNA sequence into which a single (19)F nucleotide is incorporated by commercial synthesis. This method is therefore suitable for application to biologically significant systems in which dynamic conformational rearrangement is important for function and may therefore facilitate studies of RNA structure-function relationships.

Conformational heterogeneity of the protein-free human spliceosomal U2-U6 snRNA complex.

The complex formed between the U2 and U6 small nuclear (sn)RNA molecules of the eukaryotic spliceosome plays a critical role in the catalysis of precursor mRNA splicing. Here, we have used enzymatic structure probing, (19)F NMR, and analytical ultracentrifugation techniques to characterize the fold of a protein-free biophysically tractable paired construct representing the human U2-U6 snRNA complex. Results from enzymatic probing and (19)F NMR for the complex in the absence of Mg(2+) are consistent with formation of a four-helix junction structure as a predominant conformation. However, (19)F NMR data also identify a lesser fraction (up to 14% at 25°C) of a three-helix conformation. Based upon this distribution, the calculated ΔG for inter-conversion to the four-helix structure from the three-helix structure is approximately -4.6 kJ/mol. In the presence of 5 mM Mg(2+), the fraction of the three-helix conformation increased to ∼17% and the Stokes radius, measured by analytical ultracentrifugation, decreased by 2%, suggesting a slight shift to an alternative conformation. NMR measurements demonstrated that addition of an intron fragment to the U2-U6 snRNA complex results in displacement of U6 snRNA from the region of Helix III immediately 5' of the ACAGAGA sequence of U6 snRNA, which may facilitate binding of the segment of the intron adjacent to the 5' splice site to the ACAGAGA sequence. Taken together, these observations indicate conformational heterogeneity in the protein-free human U2-U6 snRNA complex consistent with a model in which the RNA has sufficient conformational flexibility to facilitate inter-conversion between steps of splicing in situ.

Development of a microscopy-based assay for protein kinase Czeta activation in human breast cancer cells.

Protein kinase Czeta (PKCzeta) plays a critical role in cancer cell chemotaxis. Upon activation induced by epidermal growth factor (EGF) or chemoattractant SDF-1alpha, PKCzeta redistributes from cytosol to plasma membrane. Based on this property, we developed a rapid cell-based assay for inhibitors of ligand-induced PKCzeta activation. PKCzeta green fluorescent protein (GFP) was transfected into human breast cancer cells, MDA-MB-231, to establish a stable cell line, PKCzeta-GFP/MDA-MB-231. PKCzeta-GFP/MDA-MB-231 maintained phenotypes, such as chemotaxis, adhesion, and cell migration, similar to those of its parental cell line. Therefore it could be used as a representative cancer cell line. EGF induced translocation of PKCzeta-GFP to plasma membrane in a pattern similar to that of endogenous PKCzeta, indicative of activation of PKCzeta Translocation of PKCzeta-GFP could be easily and directly recorded by an inverted fluorescence microscope. Inhibitors of chemotaxis also impaired the translocation of PKCzeta-GFP, which further validated the biological relevance of our assay. Taken together, we have developed a simple, rapid, and reliable assay to detect the ligand-induced activation of PKCzeta in human cancer cells. This assay can be used in screening for inhibitors of PKCzeta activation, which is critically required for cancer cell chemotaxis.