DeepTracer-Denoising: Deep Learning for 3D Electron Density Map Denoising

The pyruvate dehydrogenase multienzyme complexes are among the largest multifunctional catalytic machines in cells, catalyzing the production of acetyl CoA from pyruvate. We have previously reported the molecular architecture of an 11-MDa subcomplex comprising the 60-mer icosahedral dihydrolipoyl acetyltransferase (E2) decorated with 60 copies of the heterotetrameric (alpha(2)beta(2)) 153-kDa pyruvate decarboxylase (E1) from Bacillus stearothermophilus (Milne, J. L. S., Shi, D., Rosenthal, P. B., Sunshine, J. S., Domingo, G. J., Wu, X., Brooks, B. R., Perham, R. N., Henderson, R., and Subramaniam, S. (2002) EMBO J. 21, 5587-5598). An annular gap of approximately 90 A separates the acetyltransferase catalytic domains of the E2 from an outer shell formed of E1 tetramers. Using cryoelectron microscopy, we present here a three-dimensional reconstruction of the E2 core decorated with 60 copies of the homodimeric 100-kDa dihydrolipoyl dehydrogenase (E3). The E2E3 complex has a similar annular gap of approximately 75 A between the inner icosahedral assembly of acetyltransferase domains and the outer shell of E3 homodimers. Automated fitting of the E3 coordinates into the map suggests excellent correspondence between the density of the outer shell map and the positions of the two best fitting orientations of E3. As in the case of E1 in the E1E2 complex, the central 2-fold axis of the E3 homodimer is roughly oriented along the periphery of the shell, making the active sites of the enzyme accessible from the annular gap between the E2 core and the outer shell. The similarities in architecture of the E1E2 and E2E3 complexes indicate fundamental similarities in the mechanism of active site coupling involved in the two key stages requiring motion of the swinging lipoyl domain across the annular gap, namely the synthesis of acetyl CoA and regeneration of the dithiolane ring of the lipoyl domain.

High-resolution electron crystallography of protein molecules

Cryo-electron microscopy (cryo-EM) has transformed structural biology by enabling the determination of macromolecular structures at near-atomic resolution. The recent Nobel Prizes awarded to John Hopfield and Geoffrey Hinton in Physics, and to David Baker, Demis Hassabis, and John Jumper in Chemistry, highlight the growing convergence of artificial intelligence (AI) and computational innovation in advancing scientific discovery. Cryo-EM has emerged as a powerful tool for elucidating the architectures and dynamics of complex biological assemblies. Its ability to capture multiple conformational states provides mechanistic insights that are critical for understanding molecular function. However, traditional cryo-EM data processing pipelines face challenges in handling large, heterogeneous datasets. Manual particle selection, extensive computational demands, and uneven sampling of conformational states often limit resolution, especially for rare or transient conformations. Recent advances in AI-driven methods are reshaping this landscape. Deep neural networks now facilitate automated particle picking, denoising, classification, and three-dimensional reconstruction, substantially reducing human bias and computational bottlenecks. Machine learning models have also enhanced the accuracy of map interpretation and atomic modeling, accelerating the transition from raw micrographs to high-resolution structures. These developments represent a paradigm shift, positioning AI as an integral component of the “resolution revolution” in structural biology. This comprehensive article examines the state of the art in cryo-electron microscopy, highlighting how artificial intelligence is redefining every stage of the workflow-from automated particle picking and map reconstruction to structure modeling and high-resolution refinement. The integration of deep learning and neural network-based tools has transformed cryo-EM into a data-driven and highly precise discovery platform. By showcasing these innovations, we emphasize AI’s pivotal role in advancing next-generation cryo-EM research and accelerating molecular insights with profound implications for structural biology and drug discovery. Not applicable.

Determining the structure of a protein complex using electron microscopy requires the calculation of a 3D density map from 2D images of single particles. Since the individual images are taken at low electron dose to avoid radiation damage, they are noisy and difficult to align with each other. This can result in incorrect maps, making validation essential. Pairs of electron micrographs taken at known angles to each other (tilt-pairs) can be used to measure the accuracy of assigned projection orientations and verify the soundness of calculated maps. Here we establish a statistical framework for evaluating images and density maps using tilt-pairs. The directional distribution of such angular data is modelled using a Fisher distribution on the unit sphere. This provides a simple, quantitative and easily comparable metric, the concentration parameter κ, for evaluating the quality of datasets and density maps that is independent of the data collection and analysis methods. A large κ is indicative of good agreement between the particle images and the 3D density map. For structure validation, we recommend and a p-value <0.01. The statistical framework herein allows one to objectively answer the question: Is a reconstructed density map correct within a particular confidence interval?

The cryo-EM structure of ETB-Gi complex, obtained using the Fusion-G system, provides insights into the distinct activation mechanism and G-protein promiscuity of ETB, aiding in the design of ETB agonists for therapeutic applications.

Cryo-EM in molecular and cellular biology

High-resolution structures of membrane and soluble proteins can be obtained by cryo-electron crystallography, given highly-ordered cryo-preparations of perfectly flat 2D crystals are available. Studies of membrane proteins, which are embedded into a lipid membrane, mimicking the native cell membrane, are of particular biological interest. However there are multiple reasons why electron crystallography is far from being a mainstream protein structure determination technique. In this thesis we address three major difficulties of electron crystallography: (i) resolution loss due to not perfectly flat crystals, (ii) reliable high-throughput automatic image processing and (iii) correction of electron beam-induced motion of the sample.&#13;\n&#13;\nThe conventional electron crystallography image processing procedure assumes perfectly flat 2D crystals, which are almost impossible to obtain. Our new processing approach, described in Chapter 2, weakens this assumption tremendously. Traditional processing assigns the same tilt geometry to all proteins of one 2D crystal. Thus local tilt geometry variations, due to not perfectly flat crystals, are neglected. We developed an algorithm that optimizes the tilt geometry of each protein separately, while exploiting the correlation between neighboring proteins. The new method proves the feasibility of this approach, improves the achieved resolution and opens the doors to new studies, i.e. structural studies of membrane proteins embedded into lipid vesicles.&#13;\n&#13;\nRecently a new generation of digital detectors tremendously changed the cryo-electron microscopy field. Beside a significantly increased signal-to-noise ratio these detectors record dose-fractionated movies of the sample under the electron beam. Previous cameras only recorded one image instead. This new exposure mode enables the computational correction for beam-induced sample movements. In Chapter 3 we describe a new “real-time” automation pipeline for electron crystallography using direct electron detectors. The novel pipeline automatically corrects for homogeneous sample drift on frame level and processes the acquired images automatically. Both, the time-to-solution and the quality of the obtained 3D reconstructions are significantly improved.&#13;\n&#13;\nHeterogeneous beam-induced sample movements are the most severe resolution- limiting factor in modern cryo-electron crystallography. In Chapter 4 we present an algorithm, termed movie-mode unbending, which corrects for inhomogeneous beam- induced sample drift. In contrast to the previous homogeneous drift-correction, the novel algorithm can correct for locally varying beam-induced sample motion. This novel approach significantly increases the resolution of electron cryo-crystallographic studies recorded on the latest detectors.&#13;\n&#13;\nChapter 5 covers multiple successful applications of the methods introduced above. For instance the real-time drift-correction module of the developed automation pipeline was used for multiple near-atomic resolution single particle and helical image processing projects. The high-throughput crystal image processing was applied to different kinds of 2D crystals and enabled the qualitative assessment of different sample preparation methods. Additionally we describe the implementation of a high-throughput single particle automation pipeline, which will enable the generation of near-atomic resolution single particle cryo-electron microscopy density maps on a daily basis.

Objective To compare the rat head MRI image quality of small aperture orthogonal phased-array creatural experimental coil and the C3 surface coil.Methods Ten male SD rats underwent both coils for head MRI.The signal intensities and background noise values of the cerebral cortex,cerebellum, brain stem, muscle and aqueous humor in eye were measured, and signal to noise ratio (SNR)was calculated.Image quality scores were obtained from 2 radiologists, t test was performed for statistical differences.Results SNR values over the cerebral cortex, cerebellum, brain stem, muscle and aqueous humor in eye were obtained from the images of all protocols.Statistically significant differences of the two coils were achieved in the cerebral cortex in the transverse T1WI (SNR sense-body=41.3±23.5; SNR C3=10.3±0.5; t=-3.21), sagittal T2WI (SNR sense-body=63.8±16.6; SNR C3=37.9±4.4; (=-4.00) and coronal T2WI (SNR sense-body=91.6±23.8, SNR C3=38.5±11.8; t=-5.69)(P＜0.01), so were in cerebellum, brain stem and aqueous humor in eye.Statistical differences of muscular tissue were observed merely in the transverse axial TiWI (SNR sense-body=39.9±21.1; SNR C3=8.3±1.7; t=-3.64) and the coronal T2WI (SNR sense-body=23.5±9.8; SNR C3=1 1.9±4.1; t=-3.10)(P＜0.01).Coronal T2WI score achieved the top in the small aperture orthogonal phased array coil, and the image qualities of diffusion-weighted imaging(DWI)in both two coils were low.Conclusion In head MRI of rats, the SNR value and head image quality of small aperture orthogonal phased-array coil is superior to that of the frequently used C3 surface coil. Key words: Magnetic resonance imaging; Small aperture orthogonal phased-array coil; Surface coils; Rat; Image quality

Cryo-EM structure and functional analyses show that human iron exporter ferroportin (Fpn) can bind and transport calcium ions and the transport activity is reduced when the concentration of ferrous ions is high.

Volta phase plate data collection facilitates image processing and cryo-EM structure determination.

Cryo-electron microscopy (cryo-EM) single-particle image analysis is a powerful technique to resolve structures of biomacromolecules, while the challenge is that the cryo-EM image is of a low signal-to-noise ratio. For both two-dimensional image analysis and three-dimensional density map analysis, image alignment is an important step to improve the precision of the image distance calculation. In this paper, we introduce a new algorithm for performing two-dimensional pairwise alignment for cryo-EM particle images, which is based on the Fourier transform and power spectrum analysis. Compared to the existing heuristic iterative alignment methods, our method utilizes the signal distribution and signal feature on images' power spectrum to directly compute the alignment parameters. It does not require iterative computations and is robust against the cryo-EM image noise. Both theoretical analysis and experimental results suggest that our power-spectrum-feature-based alignment method is highly computational-efficient and is capable of offering effective alignment results. This new alignment algorithm is publicly available at: www.csbio.sjtu.edu.cn/bioinf/EMAF/for academic use.

We discuss recent advancements in structural biology methods for investigating sites of protein-protein interactions. We will inform readers outside the field of structural biology about techniques beyond crystallography, and how these different technologies can be utilized for drug development. Advancements in cryo-electron microscopy (cryoEM) and micro-electron diffraction (microED) may change how we view atomic resolution structural biology, such that well-ordered macrocrystals of protein complexes are not required for interface identification. However, some drug discovery applications, such as lead peptide compound generation, may not require atomic resolution; mass spectrometry techniques can provide an expedited path to generation of lead compounds. New crosslinking compounds, more user-friendly data analysis, and novel protocols such as protein painting can advance drug discovery programs, even in the absence of atomic resolution structural data. Finally, artificial intelligence and machine learning methods, while never truly replacing experimental methods, may provide rational ways to stratify potential druggable regions identified with mass spectrometry into higher and lower priority candidates. Electron diffraction of nanocrystals combines the benefits of both x-ray diffraction and cryoEM, and may prove to be the next generation of atomic resolution protein-protein interface identification. However, in situations such as peptide drug discovery, mass spectrometry techniques supported by advancements in computational methods will likely prove sufficient to support drug discovery efforts. In addition, these methods can be significantly faster than any crystallographic or cryoEM methods for identification of interacting regions.

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.

Subwavelength ultrasonic imaging using a deep convolutional neural network trained on structural noise

Temporal processing is critical to a wide variety of everyday listening tasks, including speech perception. Although the importance of signal to noise ratio (SNR) is well documented in speech perception experiments, it is less explored in temporal processing experiments. The present study examined the effect of SNR on temporal processing abilities using Gap Detection Threshold (GDT) in children and adults. The study included a total of 45 subjects, where in, 25 children (Group-1) and 20 adults (Group-2) with pure-tone thresholds ranging from 0 to 25dB HL at frequency range 250-8000Hz. The GDT was measured at presentation level 50 dBSL. All the measurements were performed in 5 different conditions: 'Quiet', ' + 10dB SNR', ' + 5dB SNR', ' + 0dB SNR' and ' - 5dB SNR'. Gap Detection Thresholds are significantly higher from + 10 to - 5dB SNR when compared to quiet condition in young-adults and all sub-groups of children, whereas at + 10dB SNR, thresholds were not significantly different from quiet condition in young-adults and all sub-groups of children except for sub-group A of children, and were significantly different for all the five conditions. It was revealed that, as the signal to noise ratio (SNR) was decreased from + 10dB SNR to 0dB SNR there was a significant increase in Gap Detection Thresholds. There was a significant increase in Gap Detection Thresholds from + 10dB SNR to - 5dB SNR in both children and adults. The results also suggest that the performance on temporal processing task in the presence of background noise achieves young-adult like pattern by the age of 10-11years. Background noise affect temporal processing in both children and young-adults. Background noise impairs temporal processing in children more than the adults, which could be because of poor temporal resolving abilities in children.