Narrow Results

This paper confirms the applicability of a newly developed efficient algorithm, the direct electrifying method, for identifying backbone for 3D site and bond percolating networks. This algorithm is based on the current-carrying definition of backbone and carried out on the predetermined spanning cluster, which is assumed to be a resistor network. The scaling exponents so obtained for backbone mass, red bonds, and conductivity are in very good agreement with some existing results. The perfectly balanced bonds in 3D backbone structures are predicted first time to be 0.00179 ± 0.00009 and 0.00604 ± 0.00008 of the backbone mass for bond and site percolations, respectively.

A matching algorithm for the identification of backbones in percolation problems is introduced. Using this procedure, percolation backbones are studied in two- to five-dimensional systems containing 1.7×10⁷ sites, two orders of magnitude larger than was previously possible using burning algorithms.

We study the optimized version of the multiple invasion percolation model. Some topological aspects as the behavior of the acceptance profile, coordination number and vertex type abundance were investigated and compared to those of the ordinary invasion. Our results indicate that the clusters show a very high degree of connectivity, spoiling the usual nodes-links-blobs geometrical picture.

Object detection occupies a very important position in the fishing operation and autonomous navigation of underwater vehicles. At present, most deep-learning object detection approaches, such as R-CNN, SPPNet, R-FCN, etc., have two stages and are based on anchors. However, the previous methods generally have the problems of weak generalization ability and not high enough computational efficiency due to the generation of anchors. As a well-known one-stage anchor-free method, CenterNet can accelerate the inference speed by omitting the step of generating anchors, whereas it is difficult to extract sufficient global information because of the residual structure at the bottom layer, which leads to low detection precision for the overlapping targets. Dilation convolution makes the kernel obtain a larger reception field and access more information. Multi-branch structure can not only preserve the whole area information, but also efficiently separate foreground and background. By combining the dilation convolution and multi-branch structure, multi-branch dilation convolution is proposed and applied to the Hourglass backbone network in CenterNet, then an improved CenterNet named multi-branch dilation convolution CenterNet (MDC-CenterNet) is built, which has a stronger ability of object detection. The proposed method is successfully utilized for detection of underwater organisms including holothurian, scallop, echinus and starfish, and the comparison result shows that it outperforms the original CenterNet and the classical object detection network. Moreover, with the MS-COCO and PASCAL VOC datasets, a number of comparative experiments are performed for showing the advancement of our method compared to other best methods.

Residual Dipolar Couplings (RDCs) are a source of NMR data that can provide a powerful set of constraints on the orientation of inter-nuclear vectors, and are quickly becoming a larger part of the experimental toolset for molecular biologists. However, few reliable protocols exist for the determination of protein backbone structures from small sets of RDCs. DynaFold is a new dynamic programming algorithm designed specifically for this task, using minimal sets of RDCs collected in multiple alignment media. DynaFold was first tested utilizing synthetic data generated for the N–H, C_α–H_α, and C–N vectors of 1BRF, 1F53, 110M, and 3LAY proteins, with up to ±1 Hz error in three alignment media, and was able to produce structures with less than 1.9 Å of the original structures. DynaFold was then tested using experimental data, obtained from the Biological Magnetic Resonance Bank, for proteins PDBID:1P7E and 1D3Z using RDC data from two alignment media. This exercise yielded structures within 1.0 Å of their respective published structures in segments with high data density, and less than 1.9 Å over the entire protein. The same sets of RDC data were also used in comparisons with traditional methods for analysis of RDCs, which failed to match the accuracy of DynaFold's approach to structure determination.

Two challenges for G-protein coupled receptor (GPCR)-ligand docking with flexibility by computational method are: (1) how to accurately model the perturbations of the backbone when docking ligand to the receptor and (2) how to sample the backbone near the binding pocket. To address the two challenges, we propose a combined refinement (coREF) approach working on the flexibility of the docking interface by integrating two different refinement protocols based on RosettaLigand (RL) platform. The first refinement protocol generates multiple initial receptors by parallelling different movements to mimic the observed simultaneous fluctuations of backbones near the binding site without ligand conformation. The second refinement protocol adds a backbone refinement step for interface residues with ligand conformation, which extending the RL docking algorithm. To validate the performance of coREF, we refined 20 GPCR Docking 2008/2010 models submitted by participants. Eight of 10 CXCR4 targets and half of D3 targets showed lower LRMSD values than the submitted models. Nine in 12 results were slightly better than those refined using the RL docking algorithm.