Search

Ligand molecules naturally constitute a graph structure. Recently, many excellent deep graph learning (DGL) methods have been proposed and used to model ligand bioactivities, which is critical for the virtual screening of drug hits from compound databases in interest. However, pharmacists can find that these well-trained DGL models usually are hard to achieve satisfying performance in real scenarios for virtual screening of drug candidates. The main challenges involve that the datasets for training models were small-sized and biased, and the inner active cliff cases would worsen model performance. These challenges would cause predictors to overfit the training data and have poor generalization in real virtual screening scenarios. Thus, we proposed a novel algorithm named adversarial feature subspace enhancement (AFSE). AFSE dynamically generates abundant representations in new feature subspace via bi-directional adversarial learning, and then minimizes the maximum loss of molecular divergence and bioactivity to ensure local smoothness of model outputs and significantly enhance the generalization of DGL models in predicting ligand bioactivities. Benchmark tests were implemented on seven state-of-the-art open-source DGL models with the potential of modeling ligand bioactivities, and precisely evaluated by multiple criteria. The results indicate that, on almost all 33 GPCRs datasets and seven DGL models, AFSE greatly improved their enhancement factor (top-10%, 20% and 30%), which is the most important evaluation in virtual screening of hits from compound databases, while ensuring the superior performance on RMSE and $r^2$. The web server of AFSE is freely available at http://noveldelta.com/AFSE for academic purposes.

For the problem of low accuracy and efficiency in criminal investigation image retrieval tasks, proposed a retrieval method based on deep hash coding. Adopting ResNet as backbone network for feature extraction, and adding a pyramid pooling layer to enhance the quality of image features; In order to improve the retrieval efficiency, a deep hash layer is introduced to adaptively extract binary hash code for the image. By means of image-pair training method, and Combination of classification loss and contrast loss for joint learning, optimize training effect. Experiments show that the accuracy and efficiency of image retrieval by this method are significantly improved when compared with the existing benchmark algorithms.

Aiming at the disadvantages of traditional manual attendance, this paper proposes a face recognition based attendance scheme. Through mobile platform and face recognition technology to optimize the manual attendance process. This design divides into the face recognition system of check on work attendance information input, attendance sign-in and attendance record three function module, and introduces a principle of face detection and classification, analyses the process of the construction of the face recognition classifier, the last on the Android platform design and implement a face recognition system of check on work attendance, by comparing the experiment results of face recognition accuracy, verify the feasibility of this scheme.

By integrating diagenesis and sedimentary facies, the distribution of diagenetic alterations and their impacts on reservoir quality were investigated within a lacustrine fan delta depositional environment in the Lower Cretaceous of the Western sub-sag in the Chagan Sag, Yin-E Basin, northern China. Core observation and analyses of petrography and geochemistry of the sandstones revealed that the eogenetic alterations display spatial and temporal distribution patterns associated with sedimentary facies, including distributary channels of a fan delta front (DC), distributary bays (DB) and sheet sands (SS) and mouth bar (MB) (SS and MB are collectively referred to as S). Percolation of meteoric waters occurring in high permeability DC sandstones resulted in leaching of feldspar, the formation of kaolinite, as well as mechanical infiltrated clays around detrital grains. Conversely, the DB deposits containing abundant ductile lithic fragments were subjected to mechanical compaction and thus the development of a pseudomatrix. During a long burial residence time (the Early Cretaceous period) in a lacustrine environment, abundant early carbonate cements were precipitated, especially in the distal S sandstones. The sandstones lacking eogenetic cements were subjected to stronger mechanical compaction. Additionally, eogenetic alterations have an important impact on the distribution of the mesogenetic alterations. Sandstones containing few eogenetic cements or thin or discontinuous infiltrated clay rims around the detrital grains were subjected to quartz cementation. However, during the hydrocarbon generation stage, the most efficient percolation of meteoric waters and organic acid dissolved feldspar minerals and carbonate cements, resulted in DC sandstones having more intragranular and intergranular porosity. Owing to high matrix contents and early calcite cement contents, the DB and S sandstones had lower permeability and were rarely dissolved. During the late mesogenetic stage, late carbonate cements occurred in all sedimentary facies and iron and magnesium ions were released from the transformation from kaolinite to illite or to chlorite. The results from this study show possible diagenetic evolutionary pathways in the reservoir sandstones within the fan delta depositional environment, which in turn provides some insights into the controls on reservoir potential. Copyright © 2016 John Wiley & Sons, Ltd.