CR-I-TASSER: assemble protein structures from cryo-EM density maps using deep convolutional neural networks.

The rigid-body fitting of predicted structural models into cryo-electron microscopy (cryo-EM) density maps is a necessary procedure for density map-guided protein structure determination and prediction. We proposed a novel multiobjective optimization protocol, MOFIT, which performs a rigid-body density-map fitting based on particle swarm optimization (PSO). MOFIT was tested on a large set of 292 nonhomologous single-domain proteins. Starting from structural models predicted by I-TASSER, MOFIT achieved an average coordinate root-mean-square deviation of 2.46 Å, which was 1.57, 2.79, and 3.95 Å lower than three leading single-objective function-based methods, where the differences were statistically significant with p-values of 1.65 × 10-6, 6.36 × 10-8, and 6.44 × 10-11 calculated using two-tail Student's t tests. Detailed analyses showed that the major advantages of MOFIT lie in the multiobjective protocol and the extensive PSO search simulations guided by the composite objective functions, which integrates complementary correlation coefficients from the global structure, local fragments, and individual residues with the cryo-EM density maps.

The two major challenges in protein structure prediction problems are (1) the lack of an accurate energy function and (2) the lack of an efficient search algorithm. A protein energy function accurately describing the interaction between residues is able to supervise the optimization of a protein conformation, as well as select native or native-like structures from numerous possible conformations. An efficient search algorithm must be able to reduce a conformational space to a reasonable size without missing the native conformation. My PhD research studies focused on these two directions. A protein energy function—the distance and orientation dependent energy function of amino acid key blocks (DOKB), containing a distance term, an orientation term, and a highly packed term—was proposed to evaluate the stability of proteins. In this energy function, key blocks of each amino acids were used to represent each residue; a novel reference state was used to normalize block distributions. The dependent relationship between the orientation term and the distance term was revealed, representing the preference of different orientations at different distances between key blocks. Compared with four widely used energy functions using six general benchmark decoy sets, the DOKB appeared to perform very well in recognizing native conformations. Additionally, the highly packed term in the DOKB played its important role in stabilizing protein structures containing highly packed residues. The cluster potential adjusted the reference state of highly packed areas and significantly improved the recognition of the native conformations in the ig_structal data set. The DOKB is not only an alternative protein energy function for protein structure prediction, but it also provides a different view of the interaction between residues. The top-k search algorithm was optimized to be used for proteins containing both α-helices and β-sheets. Secondary structure elements (SSEs) are visible in cryo-electron microscopy (cryo-EM) density maps. Combined with the SSEs predicted in a protein sequence, it is feasible to determine the topologies referring to the order and direction of the SSEs in the cryo-EM density map with respect to the SSEs in the protein sequence. Our group member Dr. Al Nasr proposed the top-k search algorithm, searching the top-k possible topologies for a target protein. It was the most effective algorithm so far. However, this algorithm only works well for pure a-helix proteins due to the complexity of the topologies of β-sheets. Based on the known protein structures in the Protein Data Bank (PDB), we noticed that some topologies in β-sheets had a high preference; on the contrary, some topologies never appeared. The preference of different topologies of β-sheets was introduced into the optimized top-k search algorithm to adjust the edge weight between nodes. Compared with the previous results, this optimization significantly improved the performance of the top-k algorithm in the proteins containing both α-helices and…

Structural modeling of proteins from cryo-electron microscopy (cryo-EM) density maps is one of the challenging issues in structural biology. De novo modeling combined with flexible fitting refinement (FFR) has been widely used to build a structure of new proteins. In de novo prediction, artificial conformations containing local structural errors such as chirality errors, cis peptide bonds, and ring penetrations are frequently generated and cannot be easily removed in the subsequent FFR. Moreover, refinement can be significantly suppressed due to the low mobility of atoms inside the protein. To overcome these problems, we propose an efficient scheme for FFR, in which the local structural errors are fixed first, followed by FFR using an iterative simulated annealing (SA) molecular dynamics protocol with the united atom (UA) model in an implicit solvent model; we call this scheme “SAUA-FFR”. The best model is selected from multiple flexible fitting runs with various biasing force constants to reduce overfitting. We apply our scheme to the decoys obtained from MAINMAST and demonstrate an improvement of the best model of eight selected proteins in terms of the root-mean-square deviation, MolProbity score, and RWplus score compared to the original scheme of MAINMAST. Fixing the local structural errors can enhance the formation of secondary structures, and the UA model enables progressive refinement compared to the all-atom model owing to its high mobility in the implicit solvent. The SAUA-FFR scheme realizes efficient and accurate protein structure modeling from medium-resolution maps with less overfitting.

Recent breakthroughs in deep learning-based protein structure prediction show that it is possible to obtain highly accurate models for a wide range of difficult protein targets for which only the amino acid sequence is known. The availability of accurately predicted models from sequences can potentially revolutionise many modelling approaches in structural biology, including the interpretation of cryo-EM density maps. Although atomic structures can be readily solved from cryo-EM maps of better than 4 Å resolution, it is still challenging to determine accurate models from lower-resolution density maps. Here, we report on the benefits of models predicted by AlphaFold2 (the best-performing structure prediction method at CASP14) on cryo-EM refinement using the Phenix refinement suite for AlphaFold2 models. To study the robustness of model refinement at a lower resolution of interest, we introduced hybrid maps (i.e. experimental cryo-EM maps) filtered to lower resolutions by real-space convolution. The AlphaFold2 models were refined to attain good accuracies above 0.8 TM scores for 9 of the 13 cryo-EM maps. TM scores improved for AlphaFold2 models refined against all 13 cryo-EM maps of better than 4.5 Å resolution, 8 hybrid maps of 6 Å resolution, and 3 hybrid maps of 8 Å resolution. The results show that it is possible (at least with the Phenix protocol) to extend the refinement success below 4.5 Å resolution. We even found isolated cases in which resolution lowering was slightly beneficial for refinement, suggesting that high-resolution cryo-EM maps might sometimes trap AlphaFold2 models in local optima.

Recent breakthroughs in deep learning-based protein structure prediction show that it is possible to obtain highly accurate models for a wide range of difficult protein targets for which only the amino acid sequence is known. The availability of accurately predicted models from sequences can potentially revolutionise many modelling approaches in structural biology, including the interpretation of cryo-EM density maps. Although atomic structures can be readily solved from cryo-EM maps of better than 4 Å resolution, it is still challenging to determine accurate models from lower-resolution density maps. Here, we report on the benefits of models predicted by AlphaFold2 (the best-performing structure prediction method at CASP14) on cryo-EM refinement using the Phenix refinement suite for AlphaFold2 models. To study the robustness of model refinement at a lower resolution of interest, we introduced hybrid maps (i.e. experimental cryo-EM maps) filtered to lower resolutions by real-space convolution. The AlphaFold2 models were refined to attain good accuracies above 0.8 TM scores for 9 of the 13 cryo-EM maps. TM scores improved for AlphaFold2 models refined against all 13 cryo-EM maps of better than 4.5 Å resolution, 8 hybrid maps of 6 Å resolution, and 3 hybrid maps of 8 Å resolution. The results show that it is possible (at least with the Phenix protocol) to extend the refinement success below 4.5 Å resolution. We even found isolated cases in which resolution lowering was slightly beneficial for refinement, suggesting that high-resolution cryo-EM maps might sometimes trap AlphaFold2 models in local optima.

Recent breakthroughs in deep learning-based protein structure prediction show that it is possible to obtain highly accurate models for a wide range of difficult protein targets for which only the amino acid sequence is known. The availability of accurately predicted models from sequences can potentially revolutionise many modelling approaches in structural biology, including the interpretation of cryo-EM density maps. Although atomic structures can be readily solved from cryo-EM maps of better than 4 Å resolution, it is still challenging to determine accurate models from lower-resolution density maps. Here, we report on the benefits of models predicted by AlphaFold2 (the best-performing structure prediction method at CASP14) on cryo-EM refinement using the Phenix refinement suite for AlphaFold2 models. To study the robustness of model refinement at a lower resolution of interest, we introduced hybrid maps (i.e. experimental cryo-EM maps) filtered to lower resolutions by real-space convolution. The AlphaFold2 models were refined to attain good accuracies above 0.8 TM scores for 9 of the 13 cryo-EM maps. TM scores improved for AlphaFold2 models refined against all 13 cryo-EM maps of better than 4.5 Å resolution, 8 hybrid maps of 6 Å resolution, and 3 hybrid maps of 8 Å resolution. The results show that it is possible (at least with the Phenix protocol) to extend the refinement success below 4.5 Å resolution. We even found isolated cases in which resolution lowering was slightly beneficial for refinement, suggesting that high-resolution cryo-EM maps might sometimes trap AlphaFold2 models in local optima.

Recent breakthroughs in deep learning-based protein structure prediction show that it is possible to obtain highly accurate models for a wide range of difficult protein targets for which only the amino acid sequence is known. The availability of accurately predicted models from sequences can potentially revolutionise many modelling approaches in structural biology, including the interpretation of cryo-EM density maps. Although atomic structures can be readily solved from cryo-EM maps of better than 4 Å resolution, it is still challenging to determine accurate models from lower-resolution density maps. Here, we report on the benefits of models predicted by AlphaFold2 (the best-performing structure prediction method at CASP14) on cryo-EM refinement using the Phenix refinement suite for AlphaFold2 models. To study the robustness of model refinement at a lower resolution of interest, we introduced hybrid maps (i.e. experimental cryo-EM maps) filtered to lower resolutions by real-space convolution. The AlphaFold2 models were refined to attain good accuracies above 0.8 TM scores for 9 of the 13 cryo-EM maps. TM scores improved for AlphaFold2 models refined against all 13 cryo-EM maps of better than 4.5 Å resolution, 8 hybrid maps of 6 Å resolution, and 3 hybrid maps of 8 Å resolution. The results show that it is possible (at least with the Phenix protocol) to extend the refinement success below 4.5 Å resolution. We even found isolated cases in which resolution lowering was slightly beneficial for refinement, suggesting that high-resolution cryo-EM maps might sometimes trap AlphaFold2 models in local optima.

Atomic structure modeling is a crucial step in determining the structures of protein complexes using cryo-electron microscopy (cryo-EM). This work introduces DEMO-EMol, an improved server that integrates deep learning-based map segmentation and chain fitting to accurately assemble protein-nucleic acid (NA) complex structures from cryo-EM density maps. Starting from a density map and independently modeled chain structures, DEMO-EMol first segments protein and NA regions from the density map using deep learning. The overall complex is then assembled by fitting protein and NA chain models into their respective segmented maps, followed by domain-level fitting and optimization for protein chains. The output of DEMO-EMol includes the final assembled complex model along with overall and residue-level quality assessments. DEMO-EMol was evaluated on a comprehensive benchmark set of cryo-EM maps with resolutions ranging from 1.96 to 12.77 Å, and the results demonstrated its superior performance over the state-of-the-art methods for both protein-NA and protein-protein complex modeling. The DEMO-EMol web server is freely accessible at https://zhanggroup.org/DEMO-EMol/.

Deep learning (also known as deep structured learning, hierarchical learning, or deep machine learning) is a branch of machine learning based on a set of algorithms that attempt to model high-level abstractions in data by using a deep graph with multiple processing layers, composed of multiple linear and nonlinear transformations. Deep learning has been characterized as a class of machine learning algorithms with the following characteristics [257]: • They use a cascade of many layers of nonlinear processing units for feature extraction and transformation. Each successive layer uses the output from the previous layer as input. The algorithms may be supervised or unsupervised and applications include pattern analysis (unsupervised) and classification (supervised). • They are based on the (unsupervised) learning of multiple levels of features or representations of the data. Higher-level features are derived from lower-level features to form a hierarchical representation. • They are part of the broader machine learning field of learning representations of data. • They learn multiple levels of representations that correspond to different levels of abstraction; the levels form a hierarchy of concepts. These definitions have in common: multiple layers of nonlinear processing units and the supervised or unsupervised learning of feature representations in each layer, with the layers forming a hierarchy from low-level to high-level features. Various deep learning architectures such as deep neural networks, convolutional deep neural networks, deep belief networks (DBN), and recurrent neural networks have been applied to fields like computer vision, automatic speech recognition, natural language processing, audio recognition, and bioinformatics where they have been shown to produce state-of-the-art results on various tasks. In this chapter we start in Section 7.1 with an introduction, giving a brief history of this field, the relevant literature, and its applications. Then we study some basic concepts of deep learning such as convolutional neural networks, recurrent neural networks, backpropagation algorithm, restricted Boltzmann machines, and deep learning networks in Section 7.2. Then we illustrate three examples for Apache Spark implementation for mobile big data (MBD), user moving pattern extraction, and combination with nonparametric Bayesian learning, respectively, in Sections 7.3 through 7.5. Finally, we have summary in Section 7.6.

Cryo-electron microscopy (Cryo-EM) has become central to the study of large-scale molecular interactions and has produced three-dimensional (3D) density maps at various resolutions. Secondary structure element (SSE) identification from volumetric protein density maps is critical for de novo backbone structure derivation in cryo-EM. Multiple methods have been developed to detect helix and β-sheet from density maps at medium resolutions (~5-10A). β-barrel as a special β-sheet structure has been found in many proteins. However, currently there are no methods aimed to automatically and accurately extract the β-barrel as a complete chunk of density from cryo-EM map of β-barrel protein. We present an effective approach, BarrelMiner, to automatically extract the β-barrel region from cryo-EM density maps based on the conspicuous cylindrical shape of β-barrel. BarrelMiner has been tested using ten simulated density maps at 10A resolution and seven experimental cryo-EM maps between 5.5A and 8.25A resolution. The result suggests that BarrelMiner can be used for the detection of β-barrel from medium resolution cryo-EM density maps of β-barrel proteins.

Automated image classification is an essential task of the computer vision field. The tagging of images into a set of predefined groups is referred to as image classification. The implementation of computer vision to automate image classification would be beneficial because manual image evaluation and identification can be time-consuming, particularly when there are many images of different classes. Deep learning approaches are proven to overperform existing machine learning techniques in many fields in recent years, and computer vision is one of the most notable examples. The very deep neural network (VDCNN) is a powerful deep learning model for image classification, and this paper examines it briefly using MNIST handwritten digit dataset. This dataset is used to prove the efficacy of very deep neural networks over other deep learning models. The proposed study aims to comprehend the very deep neural network architecture used to accomplish a handwritten digit recognition task. The feasibility of the proposed model is evaluated using mean accuracy, validation accuracy, and standard deviation. The study results of the very deep neural network model are compared to a convolutional neural network and convolutional neural network with batch normalization. According to the results of the comparison study, very deep neural networks achieve high accuracy of 99.1% for a handwritten dataset. The outcome of the proposed work is used to interpret how well a very deep neural network performs when compared to the other two models of deep neural networks. This proposed architecture may be used to automate the classification of handwritten digits dataset.

PurposeThe aim of this study was to analyse and review deep learning convolutional neural networks for detecting and diagnosing early-stage dental caries on periapical radiographs.Materials and MethodsIn order to conduct this review, the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines were followed. Studies published from 2015 to 2021 under the keywords (deep convolutional neural network) AND (caries), (deep learning caries) AND (convolutional neural network) AND (caries) were systematically reviewed.ResultsWhen dental caries is improperly diagnosed, the lesion may eventually invade the enamel, dentin, and pulp tissue, leading to loss of tooth function. Rapid and precise detection and diagnosis are vital for implementing appropriate prevention and treatment of dental caries. Radiography and intraoral images are considered to play a vital role in detecting dental caries; nevertheless, studies have shown that 20% of suspicious areas are mistakenly diagnosed as dental caries using this technique; hence, diagnosis via radiography alone without an objective assessment is inaccurate. Identifying caries with a deep convolutional neural network-based detector enables the operator to distinguish changes in the location and morphological features of dental caries lesions. Deep learning algorithms have broader and more profound layers and are continually being developed, remarkably enhancing their precision in detecting and segmenting objects.ConclusionClinical applications of deep learning convolutional neural networks in the dental field have shown significant accuracy in detecting and diagnosing dental caries, and these models hold promise in supporting dental practitioners to improve patient outcomes.

Gait recognition is a noncontact biometric procedure that determines the identity or health status of a person by analyzing his or her walking posture and habits, including skeletal and joint movements. The most remarkable feature of this method is the possibility of conducting recognition without demanding much cooperation from participants. Therefore, this recognition technique has attracted much attention from scholars. Additionally, because of the rapid development of graphics processing unit technology, related hardware and computation performance, the applications of deep-learning technology are considerably enhanced. The objective of this study was to apply a deep neural network (DNN), which employs deep-learning technology, to achieve gait-based automatic pedestrian detection and recognition. In contrast to using wearable devices to precisely capture skeletal and joint movements, pedestrian color-image sequences were used as input in this study. Subsequently, a pretraining convolutional neural network (CNN) was employed to capture pedestrian location and extract pedestrian dense optical flow to serve as concrete low-level feature inputs. Then, a finely-tuned DNN based on the wide residual network was employed to extract high-level abstract features. In addition, to overcome the difficulty of obtaining local temporal features by using a 2D CNN, part of the 3D convolutional structure was introduced into the CNN. This design enabled use of limited memory to acquire more effective features and enhance the DNN performance. The experimental results show that the proposed method has exceptional performance for pedestrian detection and recognition.

Self-mixing interferometry (SMI) is superior to other laser interferometry methods due to its simplicity and compactness. However, SMI signals are often complex and difficult to process due to interference in the form of variations in the effective reflectivity of the target, noisy signals, complex signal shapes, and other dependencies. Deep neural networks have been a very popular area of research in computer artificial intelligence in recent years, allowing more implicit features to be uncovered than traditional shallow machine learning. It has been shown that convolutional and back-propagation neural networks can be used for SMI signal processing. There are also studies that have used machine learning genetic algorithms for absolute distance measurement. Based on the above, this study used convolutional neural networks to form a deep neural network for absolute distance measurement based on SMI technology. We first trained the deep convolutional neural network at different feedback strengths. The results of the Convolutional Neural Network (CNN) model showed a coefficient of determination of 0.9987. which is consistent with the required model. The trained network was then used to estimate absolute distances with and without the addition of noise. The comparison proves that the proposed method is noise-proof and has high adaptability for measurements under different conditions.

Breast density is one of the most significant factors for cancer risk. In this study, we proposed a supervised deep learning approach for automated estimation of percentage density (PD) on digital mammography (DM). The deep convolutional neural network (DCNN) was trained to estimate a probability map of breast density (PMD). PD was calculated as the ratio of the dense area to the breast area based on the probability of each pixel belonging to dense region or fatty region at a decision threshold of 0.5. The DCNN estimate was compared to a feature-based statistical learning approach, in which gray level, texture and morphological features were extracted from each ROI and the least absolute shrinkage and selection operator (LASSO) was used to select and combine the useful features to generate the PMD. The reference PD of each image was provided by two experienced MQSA radiologists. With IRB approval, we retrospectively collected 347 DMs from patient files at our institution. The 10-fold cross-validation results showed a strong correlation r=0.96 between the DCNN estimation and interactive segmentation by radiologists while that of the feature-based statistical learning approach vs radiologists’ segmentation had a correlation r=0.78. The difference between the segmentation by DCNN and by radiologists was significantly smaller than that between the feature-based learning approach and radiologists (p < 0.0001) by two-tailed paired t-test. This study demonstrated that the DCNN approach has the potential to replace radiologists’ interactive thresholding in PD estimation on DMs.