A5053196390- Protein Structure and Dynamics
- Machine Learning in Bioinformatics
- Genomics and Phylogenetic Studies
- Multimodal Machine Learning Applications
- Advanced Image and Video Retrieval Techniques
- Computational Drug Discovery Methods
- Enzyme Structure and Function
- RNA and protein synthesis mechanisms
- Machine Learning in Materials Science
- Domain Adaptation and Few-Shot Learning
- Speech Recognition and Synthesis
- Human Pose and Action Recognition
- Diamond and Carbon-based Materials Research
- Genomics and Chromatin Dynamics
- High voltage insulation and dielectric phenomena
- Advanced Materials Characterization Techniques
- RNA Research and Splicing
- Topic Modeling
- Glycosylation and Glycoproteins Research
- Generative Adversarial Networks and Image Synthesis
- Power Transformer Diagnostics and Insulation
- Plasmonic and Surface Plasmon Research
- Neural Networks and Applications
- Iron and Steelmaking Processes
- Galectins and Cancer Biology
Microsoft Research Asia (China)
2020-2024
Northeast Electric Power University
2024
Guangxi Normal University
2020-2023
Institute of Computing Technology
2015-2020
Chinese Academy of Sciences
2015-2020
Tongji University
2012-2019
University of Chinese Academy of Sciences
2016-2018
Toyota Technological Institute at Chicago
2018
China Three Gorges University
2009
Institute of Solid State Physics
2002
Abstract Advances in deep learning have greatly improved structure prediction of molecules. However, many macroscopic observations that are important for real-world applications not functions a single molecular but rather determined from the equilibrium distribution structures. Conventional methods obtaining these distributions, such as dynamics simulation, computationally expensive and often intractable. Here we introduce framework, called Distributional Graphormer (DiG), an attempt to...
Abstract Residue co-evolution has become the primary principle for estimating inter-residue distances of a protein, which are crucially important predicting protein structure. Most existing approaches adopt an indirect strategy, i.e., inferring residue based on some hand-crafted features, say, covariance matrix, calculated from multiple sequence alignment (MSA) target protein. This however, cannot fully exploit information carried by MSA. Here, we report end-to-end deep neural network,...
Advances in deep learning have greatly improved structure prediction of molecules. However, many macroscopic observations that are important for real-world applications not functions a single molecular structure, but rather determined from the equilibrium distribution structures. Traditional methods obtaining these distributions, such as dynamics simulation, computationally expensive and often intractable. In this paper, we introduce novel framework, called Distributional Graphormer (DiG),...
Template-based modeling, including homology modeling and protein threading, is a popular method for 3D structure prediction. However, alignment generation template selection sequences without close templates remain very challenging.We present new called DeepThreader to improve both selection, by making use of deep learning (DL) residue co-variation information. Our first employs DL predict inter-residue distance distribution from sequential information (e.g. sequence profile predicted...
Accurate recognition of protein fold types is a key step for template-based prediction structures. The existing approaches to mainly exploit the features derived from alignments query against templates. These have been shown be successful at family level, but usually failed superfamily/fold levels. To overcome this limitation, one points explore more structurally informative proteins. Although residue-residue contacts carry abundant structural information, how thoroughly these information...
Generally, most existing cross-modal retrieval methods only consider global or local semantic embeddings, lacking fine-grained dependencies between objects. At the same time, it is usually ignored that mutual transformation modalities also facilitates embedding of modalities. Given these problems, we propose a method called BiKA (Bidirectional Knowledge-assisted and Attention-based generation). The model uses bidirectional graph convolutional neural network to establish In addition, employs...
Abstract Summary: The protein structure prediction approaches can be categorized into template-based modeling (including homology and threading) free modeling. However, the existing threading tools perform poorly on remote homologous proteins. Thus, improving fold recognition for proteins remains a challenge. Besides, proteome-wide poses another challenge of increasing throughput. In this study, we presented FALCON@home as server focusing homologue identification. design is based observation...
Template-based modeling (TBM), including homology and protein threading, is one of the most reliable techniques for structure prediction. It predicts by building an alignment between query sequence under prediction templates with solved structures. However, it still very challenging to build optimal sequence-template alignment, especially when only distantly related are available. Here we report a novel deep learning approach ProALIGN that can predict much more accurate alignment. Like...
Accurate prediction of inter-residue contacts a protein is important to calculating its tertiary structure. Analysis co-evolutionary events among residues has been proved effective in inferring contacts. The Markov random field (MRF) technique, although being widely used for contact prediction, suffers from the following dilemma: actual likelihood function MRF accurate but time-consuming calculate; contrast, approximations likelihood, say pseudo-likelihood, are efficient calculate...
Abstract Linking cis -regulatory sequences to target genes has been a long-standing challenge. In this study, we introduce CREaTor, an attention-based deep neural network designed model patterns for genomic elements up 2 Mb from genes. Coupled with training strategy that predicts gene expression flanking candidate (cCREs), CREaTor can cell type-specific in new types without prior knowledge of cCRE-gene interactions or additional training. The zero-shot modeling capability, combined the use...
In order to improve the speaker recognition accuracy, pitch is applied GMM-based (SR). The circular average magnitude difference function (CAMDF) method used extract pitch. An endpoint detection based on proposed. following four features are selected as of SR: mel-frequency cepstral coefficient (MFCC) pitch, contour, first-order and changed rate. Experimental results show that rate using proposed improved 20% than conventional method. system 5% MFCC parameters only.
Residues in a protein might be buried inside or exposed to the solvent surrounding protein. The residues usually form hydrophobic cores maintain structural integrity of proteins while are tightly related functions. Thus, accurate prediction accessibility will greatly facilitate our understanding both structure and functionalities proteins. Most state-of-the-art approaches consider burial state each residue independently, thus neglecting correlations among residues.In this study, we present...
Protein functions are largely determined by the final details of their tertiary structures, and structures could be accurately reconstructed based on inter-residue distances. Residue co-evolution has become primary principle for estimating distances since residues in close spatial proximity tend to co-evolve. The widely-used approaches infer residue using an indirect strategy, i.e., they first extract from multiple sequence alignment (MSA) query protein some handcrafted features, say,...
Semiconducting nanohole arrays have been considered as a promising candidate for high-efficiency solar cells. In this paper, the optical absorption property of randomly rotated elliptical consisting 1×1, 2×2, and 4×4 cells has investigated. It is found that average ultimate efficiency increases with increase size supercell. The array highest efficiency, less sensitive to parameters random rotation angle than 1×1 2×2 arrays. comparison spectra three shows number peaks least, but peak...
Abstract Template-based modeling (TBM), including homology and protein threading, is one of the most reliable techniques for structure prediction. It predicts by building an alignment between query sequence under prediction templates with solved structures. However, it still very challenging to build optimal sequence-template alignment, especially when only distantly-related are available. Here we report a novel deep learning approach ProALIGN that can predict much more accurate alignment....
Abstract Linking cis -regulatory sequences to target genes has been a long-standing challenge. In this study, we introduce CREaTor, an attention-based deep neural network designed model patterns for genomic elements up 2Mb from genes. Coupled with training strategy that predicts gene expression flanking candidate (cCREs), CREaTor can cell type-specific in new types without prior knowledge of cCRE-gene interactions or additional training. The zero-shot modeling capability, combined the use...
This work is mainly focused on showing experimental results of speaker recognition with voice activity detection. A VAD algorithm based the finite state machine introduced firstly. The incorporated into two (SR)systems. mel frequency ceptral coefficients(MFCCs) are adopted as speech feature parameters in both systems. Vector quantization (VQ)and Gaussian mixture model (GMM) classifiers SR systems, respectively. show that improved performance systems small database. However, databases get...
Here is presented the spectroscopic study of evolution first buried interfaces a B<sub>4</sub>C capped Co/Mo<sub>2</sub>C multilayer mirror induced by thermal treatment up to 600°C. This kind typically performed simulate response optics working in extreme conditions, as for instance when irradiated new high brilliance sources Free Electron Lasers. In fact, efficiency multilayers related optical contrast between alternating and low density layers, then degree interdiffusion creation or...