Biology subjects

Huang, S.-Y.

Publications and source records attributed to Huang, S.-Y..

At least 19 recordsLinked to original sources

bioRxiv2026

EM3DFold: accurate de novo protein and nucleic acid model building for cryo-EM maps using language model-powered deep learning

Cryo-electron microscopy (cryo-EM) has become one of the most powerful techniques for macromolecular structure determination. However, accurate model building from cryo-EM maps remains challenging, particularly for nucleic acids. Here, we present EM3DFold, a unified de novo model-building framework for accurate structure determination of proteins, nucleic acids, and protein-nucleic acid complexes from cryo-EM maps using a density-aware, large language model-powered three-track attention (TTA) network. The TTA network effectively integrates sequence, density, and structural information to enable accurate all-atom model building. EM3DFold was extensively evaluated on independent benchmarks of 298 experimental cryo-EM maps at < 4.0 [A] resolutions, and achieves an unprecedentedly high median accuracy of 75% completeness (88% coverage and 94% sequence accuracy) for 178 nucleic acid maps, 90% completeness (95% coverage and 96% sequence accuracy) for 124 protein-nucleic acid complexes, and 95% completeness (97% coverage and 98% sequence accuracy) for 120 protein-only targets, substantially outperforming state-of-the-art methods including ModelAngelo, EM2NA, CryoREAD, and EMProt. In addition, EM3DFold also produces models with superior model-to-map fit and stereochemical quality, providing a robust and reliable solution for automated cryo-EM model building. The EM3DFold package is freely available at https://github.com/huang-laboratory/EM3DFold.

bioinformatics↗

bioRxiv2026

HDOCK-Multimer: integrating docking and combinatorial assembly for structure prediction of large protein complexes

Deep learning methods, such as AlphaFold and RosettaFold, achieve high accuracy in protein structure prediction. However, predicting the structure of large protein complexes remains challenging due to their large size and intricate multi-chain interactions. Docking-based methods can handle large proteins, but are limited by the huge combinatorial binding space of multichains. Assembly-based approaches offer an alternative, but their accuracy critically relies on the precision of predicted subcomponents. Addressing the challenges, we propose HDOCK-Multimer (HDM), a structure prediction framework of large protein complexes by integrating ab initio docking and combinatorial assembly. HDM can efficiently reduce reliance on subcom-ponent accuracy through docking process, while leveraging the pairwise interactions of subcom-ponents through assembly strategy. HDM is extensively validated on three benchmarks of 35 large heteromeric complexes, 172 large protein complexes, and 7 CASP15 targets, and compared with state-of-the-art methods including MoLPC, CombFold, AlphaFold-Multimer (AFM), and AlphaFold3 (AF3). It is shown that HDOCK-Multimer substantially outperforms the other methods. In addition, HDM also shows ability to predict the stoichiometry and model the complex without stoichiometry input. It is anticipated that HDM will serve as a powerful tool for studying large protein complexes or molecular machines. The HDM package is freely available at https://github.com/huang-laboratory/HDOCK-Multimer/.

bioinformatics↗

bioRxiv2026

Validation of a high throughput fluorescent Capillary Electrophoresis Sodium Dodecyl Sulfate method for monoclonal antibody size heterogeneity assessment

Fluorescent capillary gel electrophoresis (CGE) with sodium dodecyl sulfate (CE-SDS) provides a powerful, high-sensitivity alternative to ultraviolet (UV)-based detection for characterizing therapeutic monoclonal antibodies (mAb). Regulatory and standards organizations, such as the United States Pharmacopeia (USP), include only UV based CE-SDS methods, hindering adoption, of alternative detection methods. There is growing opportunity to expand beyond exclusively UV-based CE-SDS methods. In this study, we present a full analytical validation of a light-emitting diode (LED) fluorescence-based parallel CE-SDS method for both non-reduced and reduced analysis of therapeutic antibodies. Using the NISTmAb reference material as a model system, size heterogeneity critical quality attributes (CQAs) including monomeric purity, percent glycosylation, and percent thioether were assessed. The fluorescence method demonstrated high specificity and precision with relative standard deviation (RSD) values <1% for monomeric purity and glycosylation, and <3% for thioether), as well as robust performance across variations in injection voltage, electrophoresis voltage, labeling temperature, and Labeling Buffer concentration. Ruggedness testing across users and reagent lots confirmed reproducibility, and accuracy assessments showed strong agreement with reported values from the National Institute of Standards (NIST) and traditional UV detection measurements. Linearity studies yielded coefficient of determination (R2) values >0.995 for both non-reduced and reduced analyses. These results highlight the high sensitivity, stable baseline performance, and suitability of LED fluorescence-based parallel CE-SDS as a validated, higher-throughput alternative to traditional UV-based methods for mAb quality control (QC).

biochemistry↗

bioRxiv2026

EMAlign: accurate alignment of cryo-EM maps through main-chain probability using deep learning

Accurate alignment of cryo-EM density maps is essential for comparing conformational states, searching map libraries, and guiding atomic model building, but remains challenging for noisy ex-perimental maps and partially overlapping structures. Existing alignment methods are often based on raw maps, which may result in reduced accuracy due to the density noise, or require manual intervention for local alignment, which suffers from limited general applicability. Addressing the limitations, we present EMAlign, an automatic global and local cryo-EM map alignment with predicted main-chain probability using deep learning. First, EMAlign predicts main-chain prob-ability maps from raw cryo-EM density maps using a BiMCUNet network. Then, a fast Fourier transform (FFT)-based search strategy is used to globally search the accurate alignment between cryo-EM maps based on predicted main-chain probability maps. As such, the main-chain prob-ability map overcomes the noisy raw map problem, and the FFT-based exhaustive global search ensures the general applicability of alignment. EMAlign is evaluated on 64 global map pairs, 195 local map pairs, and 60 structure-to-map pairs at 3-10 [A] resolution and compared with gm-fit, fitmap, VESPER, and CryoAlign. It is shown that EMAlign outperforms the other methods in both global and local alignment, achieving mean RMSDs of 1.03 [A] (global), 2.56 [A] (lo-cal), and 0.82 [A] (structure-to-map), with success rates of 100.0%, 100.0%, and 98.3% under the criterion of RMSD < 10 [A] . The EMAlign package is freely available at https://github.com/huang-laboratory/EMAlign/.

bioinformatics↗

bioRxiv2026

Nonequilibrium States Promote One-Pot Nonenzymatic Carbon Fixation in the Reverse Tricarboxylic Acid Cycle and Amino Acid Synthesis

Several metabolites within the reductive tricarboxylic acid (rTCA) cycle have been found to form prebiotically. However, how these metabolites connect to each other and form rTCA cycle remains unresolved. The rTCA cycle is an ancient route and is considered significant for the emergence of life, since it connects to the routes of amino acids and nucleobases synthesis. A major challenge to complete the rTCA cycle under prebiotic conditions is the thermodynamically unfavorable reductive carboxylation of succinate to -ketoglutarate. Here, we address this challenge by using the nature of energy: nonequilibrium conditions. By calculating the changes in free energy, {Delta}G, of succinate to -ketoglutarate, and its downstream reactions: -ketoglutarate to glutamate and -ketoglutarate to isocitrate under different nonequilibrium conditions, we find that these two-step reactions are exergonic under nonequilibrium conditions at a 10000:1 reactant-to-product ratio at 1.013 bar, pH 10 and 70{degrees}C. To prove the concept, we catalyze succinate to glutamate at a 10000:1 reactant-to-product ratio, with NH2OH and sodium dithionite. The process is catalyzed by Fe(0), Fe3O4, and artificial proto-[4Fe4S] clusters in 1M NaCl at pH 10 and 70{degrees}C under 1 atm of 13CO2 for 48 hours. This nonequilibrium condition and one-pot system successfully promote the formation of -ketoglutarate through carbon fixation with succinate and its subsequent conversion to glutamate. These findings demonstrate nonequilibrium states enable -ketoglutarate formation through succinate and CO2, and suggest that a tendency toward natural thermodynamics may serve as a driving force for autocatalysis in the origin of life. ImportanceHow life began remains open, metabolism provides a key framework for origins. We use a simple and robust energetic principle to show that non-equilibrium conditions can drive the highly endergonic carboxylation step of the reverse tricarboxylic acid (rTCA) cycle, enabling one-pot synthesis of glutamate. This is work bridges the gap between protometabolites and protometabolsim, suggesting that metabolites may have accumulated first, creating concentration gradients that drove reactions and ultimately enabled the emergence of protometabolism. These findings provide a plausible pathway from prebiotic chemistry to the emergence of metabolism.

biochemistry↗

bioRxiv2026

Development of autotrophy in Escherichia coli through adaptive laboratory evolution

Enabling heterotrophic Escherichia coli to use CO2 as its only carbon source remains a great challenge, and previous studies approached autotrophy conversion by metabolic engineering. Although its native carbon fixation routes were identified, the potential to reach autotrophy by itself has long been overlooked. In this study, autotrophy in E. coli was developed through adaptive laboratory evolution. After 1,000 days of consecutive inorganic subculturing, missense mutations were found in isocitrate dehydrogenase icd and isocitrate dehydrogenase kinase/phosphatase aceK genes, determining the metabolic switch between the citrate cycle and the glyoxylate shunt. By transcriptomic comparison of the adapted E. coli between inorganic and organic cultivations, two CO2 fixing enzymes activated in autotrophic mode were found, including the upregulated pyruvate:ferredoxin oxidoreductase YdbK and phosphoenolpyruvate carboxykinase Pck. Connected by the upregulated phosphoenolpyruvate synthase PpsA, a carbon fixation module was constituted, which was the shared foundation of the aspartate-threonine cycle and the citrate-glyoxylate-methylcitrate cycle, and thus integrating into an autotrophic network. By comparing the 13C enrichment patterns in inorganic cultivations between the adapted and initial E. coli, the favorable direction of the autotrophic network was confirmed. IMPORTANCEThis is the first study to accomplish autotrophy in E. coli through long-term evolution alone. Besides missense mutations in icd and aceK genes, adapted E. coli also actively regulated its gene expression to respond to inorganic environment, such as directing the metabolic switch towards the glyoxylate shunt. For biomass formation, a carbon fixation module consisted of the upregulated YdbK, PpsA, and Pck produced pyruvate and oxaloacetate as precursors for two cycles. The aspartate-threonine cycle with a replenishment side loop further accumulated these precursors, and the citrate-glyoxylate-methylcitrate cycle was driven by four overexpressed enzymes to catalyze six reactions. These metabolic pathways were integrated into a novel autotrophic network, and by understanding the nature of E. coli, rational designs for its carbon fixation optimization become attainable by using compatible mechanisms.

microbiology↗

bioRxiv2025

Accurate protein structure determination from cryo-EM maps using deep learning and structure prediction

Cryo-electron microscopy (cryo-EM) has become the main-stream technique for macromolecular structure determination. However, due to intrinsic resolution heterogeneity, accurate modeling of all-atom structure from cryo-EM maps remains challenging even for maps at near-atomic resolutions. Addressing the challenge, we present EMProt, a fully automated method for accurate protein structure determination from cryo-EM maps by efficiently integrating map information and structure prediction with a three-track attention network. EMProt is extensively evaluated on a diverse test set of 177 experimental cryo-EM maps with up to 54 chains in a case at <4.0 [A] resolutions, and compared with state-of-the-art methods including DeepMainmast, ModelAngelo, phenix.dock_and_rebuild, and AlphaFold3. It is shown that EMProt significantly outperforms the existing methods in recovering the protein structure and building the complete structure. In addition, the built models by EMProt also exhibit a high accuracy in model-to-map fit and structure validations. EMProt is freely available at http://huanglab.phys.hust.edu.cn/EMProt/.

bioinformatics↗

bioRxiv2025

Improving cryo-EM maps by resolution-dependent and heterogeneity-aware deep learning

Cryo-electron microscopy (cryo-EM) has emerged as a leading technology for determining the structures of biological macromolecules. However, map quality issues such as noise and loss of contrast hinder accurate map interpretation. Traditional and deep learning-based post-processing methods offer improvements but face limitations particularly in handling map heterogeneity. Here, we present a generalist Mamba-based deep learning model for improving cryo-EM maps, named EMReady2. EMReady2 introduces a fast Mamba-based dual-branch UNet architecture to jointly capture local and global features. In addition, EMReady2 also uses a local resolution-guided learning strategy to address map heterogeneity, and significantly extends the training set. These advances render EMReady2 applicable to a broader range of cryo-EM maps, including those containing nucleic acids, medium-resolution maps, and cryo-electron tomography (cryo-ET) maps, while substantially reducing computational cost. EMReady2 is extensively evaluated on 136 diverse maps at 2.0-10.0 [A] resolutions, and compared with existing map post-processing methods. It is shown that EMReady2 exhibits state-of-the-art performance in both map quality and map interpretability improvement. EMReady2 is freely available at https://github.com/huang-laboratory/EMReady2/.

bioinformatics↗

bioRxiv2025

ZapA employs a two-pronged mechanism to facilitate Z ring formation in Escherichia coli

The tubulin-like protein FtsZ assembles into the Z ring which leads to the assembly and activation of the division machinery in most bacteria. ZapA, a widely conserved protein that interacts with FtsZ, plays a pivotal role in organizing FtsZ filaments into a coherent Z ring. Previous studies revealed that ZapA forms a dumbbell-like tetramer that binds cooperatively to FtsZ filaments and aligns them in parallel, leading to the straightening and organization of FtsZ filament bundles. However, how ZapA interacts with FtsZ remains obscure. In this study, we uncover how ZapA interacts with FtsZ to facilitate Z ring formation in Escherichia coli. We find that mutations affecting surface exposed residues at the junction between adjacent FtsZ subunits in a filament as well as in an N-terminal motif of FtsZ weaken its interaction with ZapA in vivo and in vitro, indicating that ZapA binds to these regions of FtsZ. Consistent with this, ZapA prefers FtsZ polymers over monomeric FtsZ molecules and site-specific crosslinking confirmed that the dimer head domain of ZapA is in contact with the junction of FtsZ subunits. As a result, disruption of the putative interaction interfaces between FtsZ and ZapA abolishes the midcell localization of ZapA. Taken together, our results suggest that ZapA tetramers grab the N-terminal tails of FtsZ and then bind to the junctions between FtsZ subunits in the filament to straighten and crosslink parallel FtsZ filaments into the Z ring. SignificanceZapA is a widely conserved FtsZ-associated protein that promotes the organization of the Z ring, the key cytoskeletal element in the bacterial divisome. Although ZapA is known to crosslink FtsZ filaments, how it interacts with FtsZ remains enigmatic. In this study, we find that E. coli ZapA utilizes a dual binding mode in which it binds the junction between FtsZ subunits in a filament and to an N-terminal motif in FtsZ so that it can straighten and crosslink parallel FtsZ filaments simultaneously. Since the junction is formed when FtsZ polymerizes and falls apart when FtsZ depolymerizes, this interaction mode indicates that ZapA employs the polymerization dynamics of FtsZ to organize the filaments into the Z ring.

microbiology↗

bioRxiv2024

FOXO-regulated Deaf1 controls muscle regeneration through autophagy

The commonality between various muscle diseases is the loss of muscle mass, function, and regeneration, which severely restricts mobility and impairs the quality of life. With muscle stem cells (MuSCs) playing a key role in facilitating muscle repair, targeting regulators of muscle regeneration has been shown to be a promising therapeutic approach to repair muscles. However, the underlying molecular mechanisms driving muscle regeneration are complex and poorly understood. Here, we identified a new regulator of muscle regeneration, Deformed epidermal autoregulatory factor 1 (Deaf1) - a transcriptional factor downstream of FOXO signaling. We showed that Deaf1 is transcriptionally repressed by FOXOs and that Deaf1 targets to PI3KC3 and Atg16l1 promoter regions and suppresses their expressions. Deaf1 depletion therefore induces autophagy, which in turn blocks MuSC survival and differentiation. In contrast, Deaf1 overexpression inactivates autophagy in MuSCs, leading to increased protein aggregation and cell death. Interestingly, Deaf1 depletion and overexpression both lead to defects in muscle regeneration, highlighting the importance of fine tuning Deaf1-regulated autophagy during muscle regeneration. We further showed that Deaf1 expression is altered in aging and cachectic MuSCs. Remarkably, manipulation of Deaf1 expression can attenuate muscle atrophy and restore muscle regeneration in aged mice or mice with cachectic cancers. Together, our findings unveil an evolutionarily conserved role for Deaf1 in muscle regeneration, providing insights into the development of new therapeutic strategies against muscle atrophy.

cell biology↗

bioRxiv2023

A cyclic dipeptide for salinity stress alleviation and the trophic flexibility of an endophyte reveal niches in salt marsh plant-microbe interactions

In response to climate change, the nature of endophytes and their applications in sustainable agriculture has attracted the attention of academia and agro-industries. We focused on the endophytic halophiles of the endangered Taiwanese salt marsh plant, Bolboschoenus planiculmis, and evaluated the functions of the isolates through in planta salinity stress alleviation assay using Arabidopsis. An endophytic strain Priestia megaterium BP01R2 that could promote plant growth and salinity tolerance was further characterised through multi-omics approaches. The transcriptomics results suggested that BP01R2 could function by tuning hormone signal transduction, energy-producing metabolism, multiple stress responses, etc. In addition, a cyclodipeptide, cyclo(L-Ala-Gly), identified by metabolomics analysis was later confirmed to contribute to salinity stress alleviation in stressed plants by exogenous supplementation. Here we provide a new perspective on host-microbe interactions in the wetland biome based on the multi-omics investigation and mixotrophic character of BP01R2. This study revealed a biostimulant-based plant-endophyte symbiosis with potential application in sustainable agriculture and facilitated our understanding of those enigmatic cross-kingdom relationships.

microbiology↗

bioRxiv2023

Manipulation of Bacterial ROS Production Leads to Self-escalating DNA Damage and Resistance-resistant Lethality for Intracellular Mycobacteria

The high prevalence of drug resistance in mycobacteria calls for antimicrobial mechanisms that suppresses the development of resistance. As a structurally conserved multi-site bio-macromolecule, DNA is presumed to be an ideal candidate for such resistance-resistant drug target. However, survey of marketed and investigational DNA interactors indicates that they are not immune to resistance development. Here, we report our strategy to achieve real resistance-resistant DNA targeting by incurring "catastrophic" DNA damage with an organoruthenium-natural product hybrid. The dual-mode DNA damage, in the form of strong tri-valent binding and concomitant oxidative modification, is achieved by manipulating of bacterias native endogenous ROS production mechanism upon lethal stress (such as DNA binding). Such self-escalating DNA damage, together with precise targeting of intracellular bacteria via vacuole fusion, thus endows the hybrids resistance-resistant lethality against mycobacteria and in vivo efficacy in animal models.

microbiology↗

bioRxiv2022

Residue communities reveal evolutionary signatures of γδ T-Cell receptor

Naturally co-occurring amino acids, term coevolution, in a protein family play a significant role in both protein engineering and folding, and it is expanding in recent years from the studies of the effects of single-site mutations to the complete re-design of a protein and its folding, especially in three-dimensional structure prediction. Here, to better characterize such coevolving interactions, we in silico decipher evolutionary couplings from massive homologous sequences using spectral analysis to capture signatures that are important for specific molecular interactions and binding activities. We implement the present approach on the G7 gamma delta T-cell receptor to identify functionally important residues that contribute to its highly distinct binding mode. The analysis indicates the evolutionary signatures (highly ordered networks of coupled amino acids, termed residue communities) of the protein confirm previously identified functional sites that are relevant to dock the receptor underneath the major histocompatibility complex class I-related protein-1 (MR1) antigen presenting groove. Moreover, we analyze the correlation of inter-residue contacts with the activation states of receptors and show that contact patterns closely correlating with activation indeed coincide with these sites. The theoretical results demonstrate our method provides an alternative path towards bridging protein sequence with its function at residue-level without requiring its tertiary structure or highly accurate measurement of its biological activities in vivo/vitro.

bioinformatics↗

bioRxiv2022

Virulence, Aggressiveness, and Fungicide Sensitivity of Phytophthora spp. Associated with Soybean in Illinois

Phytophthora root and stem rot (PRR), caused by Phytophthora sojae, is one of the most devastating oomycete diseases of soybean in Illinois. Single resistant genes (Rps) are used to manage this pathogen, but P. sojae has adapted to Rps causing failure of resistance in many regions. In addition to P. sojae, recent reports indicate that Phytophthora sansomeana could also cause root rot in soybean. Soil samples and symptomatic plants were collected across 40 Illinois counties between 2016 and 2018. Phytophthora sojae (77%) was more abundant than P. sansomeana (23%) across Illinois fields. Both species were characterized by virulence, aggressiveness, and fungicide sensitivity. Virulence of all isolates was evaluated using the hypocotyl inoculation technique in 13 soybean differentials. Aggressiveness was evaluated in the greenhouse by inoculating a susceptible cultivar and measuring root and shoot dry weight. On average, P. sojae isolates were able to cause disease on six soybean differentials. P. sojae was more aggressive than P. sansomeana. All isolates were sensitive to azoxystrobin, ethaboxam, mefenoxam, and metalaxyl. The characterization of the population of species associated with PRR will inform management decisions for this disease in Illinois.

plant biology↗

bioRxiv2022

Equilibrative Nucleoside Transporter 3 is an IFN-stimulated Gene that Facilitates Viral Genome Release

An increasing body of evidence emphasizes the role of metabolic reprogramming in immune cells to fight off infections. However, little is known about the regulation of metabolite transporters that facilitate and support metabolic demands. In this study, we found that equilibrative nucleoside transporter 3 (ENT3) expression is part of the innate immune response, and is rapidly upregulated upon bacterial and viral infection. The transcription of ENT3 is directly under the regulation of IFN-induced signaling, positioning this metabolite transporter as an Interferon-stimulated gene (ISG). Moreover, we unveil that several viruses, including SARS-CoV2, require ENT3 to facilitate their entry into the cytoplasm. The removal or suppression of ENT3 expression is sufficient to significantly decrease viral replication in vitro and in vivo.

immunology↗

bioRxiv2020

Dihydroceramide desaturase directs the switch between exosome production and autophagy for neuronal maintenance

Exosomes are important for cell-cell communication. Deficiencies in the human dihydroceramide desaturase gene, DEGS1, increase the dihydroceramide-to-ceramide ratio and causes hypomyelinating leukodystrophy. However, the disease mechanism remains unknown. Here, we developed an in vivo assay with spatially controlled expression of exosome markers in Drosophila eye imaginal discs and showed that the level and activity of the DEGS1 ortholog, ifc, correlated with exosome production. Knocking out ifc decreased the density of the exosome precursor intraluminal vesicles (ILVs) in the multivesicular endosomes and reduced the number of exosomes released. While ifc overexpression and autophagy inhibition both enhanced exosome production, combining the two had no additive effect. Moreover, DEGS1 activity was sufficient to drive ILV formation in vitro. Together, DEGS1/Ifc controls the dihydroceramide-to-ceramide ration and enhances exosome secretion by promoting ILV formation and preventing the autophagic degradation of MVEs. These findings provide a potential cause for the neuropathy associated with DEGS1-deficient mutations. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=115 SRC="FIGDIR/small/376046v3_ufig1.gif" ALT="Figure 1"> View larger version (43K): org.highwire.dtl.DTLVardef@c7479corg.highwire.dtl.DTLVardef@12e7f46org.highwire.dtl.DTLVardef@1a5e7f0org.highwire.dtl.DTLVardef@11e7f08_HPS_FORMAT_FIGEXP M_FIG C_FIG HighlightsO_LIAn in vivo system was developed for observing exosome production in Drosophila. C_LIO_LIDihydroceramide desaturase (DEGS1/Ifc) promotes exosome production by two means. C_LIO_LIIfc drives membrane invagination to promote the formation of intraluminal vesicles. C_LIO_LIIfc inhibits autophagic degradation of MVEs and increases exosome release. C_LI eTOC BlurbThe level and activity of dihydroceramide desaturase (human DEGS1 and Drosophila Ifc) correlate with exosome production. Wu et al. show that DEGS1 drives the formation of intraluminal vesicles in vivo and in vitro. Overexpressing Ifc inhibits autophagy and reduces the degradation of multivesicular endosomes, thus increases exosome release.

cell biology↗

bioRxiv2020

Accurate prediction of residue-residue contacts across homo-oligomeric protein interfaces through deep leaning

Protein-protein interactions play a fundamental role in all cellular processes. Therefore, determining the structure of protein-protein complexes is crucial to understand their molecular mechanisms and develop drugs targeting the protein-protein interactions. Recently, deep learning has led to a breakthrough in intraprotein contact prediction, achieving an unusual high accuracy in recent CASP structure prediction challenges. However, due to the limited number of known homologous protein-protein interactions and the challenge to generate joint multiple sequence alignments (MSA) of two interacting proteins, the advances in inter-protein contact prediction remain limited. Here, we have proposed a deep learning model to predict inter-protein residue-residue contacts across homo-oligomeric protein interfaces, named as DeepHomo, by integrating evolutionary coupling, sequence conservation, distance map, docking pattern, and physic-chemical information of monomers. DeepHomo was extensively tested on both experimentally determined structures and realistic CASP-CAPRI targets. It was shown that DeepHomo achieved a high accuracy of >60% for the top predicted contact and outperformed state-of-the-art direct-coupling analysis (DCA) and machine learning (ML)-based approaches. Integrating predicted contacts into protein docking with blindly predicted monomer structures also significantly improved the docking accuracy. The present study demonstrated the success of DeepHomo in inter-protein contact prediction. It is anticipated that DeepHomo will have a far-reaching implication in the inter-protein contact and structure prediction for protein-protein interactions.

bioinformatics↗

bioRxiv2020

Automatic de novo atomic-accuracy structure determination for cryo-EM maps using deep learning

Advances in microscopy instruments and image processing algorithms have led to an increasing number of cryo-EM maps. However, building accurate models for the EM maps at 3-5 [A] resolution remains a challenging and time-consuming process. With the rapid growth of deposited EM maps, there is an increasing gap between the maps and reconstructed/modeled 3-dimensional (3D) structures. Therefore, automatic reconstruction of atomic-accuracy full-atom structures from EM maps is pressingly needed. Here, we present a semi-automatic de novo structure determination method using a deep learning-based framework, named as DeepMM, which builds atomic-accuracy all-atom models from cryo-EM maps at near-atomic resolution. In our method, the main-chain and C positions as well as their amino acid and secondary structure types are predicted in the EM map using Densely Connected Convolutional Networks. DeepMM was extensively validated on 40 simulated maps at 5 [A] resolution and 30 experimental maps at 2.6-4.8 [A] resolution as well as an EMDB-wide data set of 2931 experimental maps at 2.6-4.9 [A] resolution, and compared with state-of-the-art algorithms including RosettaES, MAINMAST, and Phenix. Overall, our DeepMM algorithm obtained a significant improvement over existing methods in terms of both accuracy and coverage in building full-length protein structures on all test sets, demonstrating the efficacy and general applicability of DeepMM. Availabilityhttps://github.com/JiahuaHe/DeepMM Supplementary informationSupplementary data are available.

bioinformatics↗