Biology subjects

Zhao, Y.

Publications and source records attributed to Zhao, Y..

At least 379 records · Page 21Linked to original sources

bioRxiv2024

Silaffins-Driven Genetic Engineering of Diatom Cell Walls: Insight into Biosilica Morphology and Nanomaterial Design

Diatoms synthesize silica cell walls (frustules) with genetically encoded morphologies, ranging from nanopatterns to micropatterns, that far exceed current synthetic chemistry. Silaffins, a family of phosphoproteins undergoing complex post-translational modifications, have been isolated from frustules and shown to facilitate and regulate biosilica formation in vitro with long-chain polyamines. However, their particular role in frustule morphogenesis and functionality remains unclear. In this study, functions of two representative silaffins, TpSil1 and TpSil3, were investigated in the model organism Thalassiosira pseudonana using gene overexpression and CRISPR/Cas9-mediated knockout approaches. Due to high sequence homology, TpSil2 was concurrently disrupted in TpSil1 knockout strains, while the homozygous knockout of TpSil3 proved to be lethal. Quantitative morphological analysis revealed distinct yet complementary roles: TpSil3 governs both microscale overall size and mesoscale features, including macropore (fultoportula) density and mesopore (cribrum pore) pattern, whereas TpSil1/2 exclusively contribute to macropore morphogenesis and mesopore density. Overexpression of silaffins increased silica deposition, while knockouts exhibited reduced silicification but enhanced cell growth and photosynthetic efficiency. Furthermore, these genetic modifications significantly influenced the physicochemical and optical properties of bulk frustules, potentially enhancing the hemostatic, catalytic and photonic performances, thereby positioning them as versatile candidates for a wide range of biotechnological and industrial applications. Collectively, our findings elucidate the distinct roles of TpSil1/2 and TpSil3 in diatom physiology and frustule morphology, highlighting a promising pathway for engineering nanostructured silica materials with tailored properties through synthetic biology.

synthetic biology↗

bioRxiv2024

Radiation damage behavior of soft matter in ultrafast cryo-electron microscopy (cryo-UEM)

Whether time-modulated pulsed-electron imaging in ultrafast electron microscopy (UEM) can mitigate the electron radiation damage that occurs to samples, is still controversial. The effectiveness of such mitigation effect and relevant potential application in cryo-EM remain to be explored. Herein, we built an ultrafast cryo-EM (cryo-UEM) device based on an ultrafast laser system. Using such equipment and the saturated aliphatic hydrocarbon compounds (C44H90), the fading curves of diffraction intensity and corresponding critical electron doses (Ne) of the samples were carefully measured under different imaging modes, temperatures, imaging dose rates and pulsed repetition rates. Our experimental results show that, the fading curves and Ne values of the C44H90 crystals are uncorrelated with the imaging electron dose rates and do not show dependence on the dose-rate effect. As the temperature decreased, the Ne values of the sample increased, indicating the cryoprotective effect on radiation damage to the samples. Surprisingly, at a constant temperature, the fading curves and Ne values of the sample in multi-electrons-packet and near-single-electron-packet pulsed modes are all approximately the same as those in conventional continuous electron-beam mode, even when the results are obtained at different pulsed repetition rates. These results show that the time-modulated pulsed electron beam does not seem to mitigate the electron radiation damage that occurs on samples. Our findings offer new insights and experimental basis for the radiation damage behavior of samples under electron beams, and provide guidance and inspiration for elucidating the fundamental principles of radiation damage.

biophysics↗

bioRxiv2024

A CRISPR-associated Rossman-fold Ring Nuclease and Adenosine Deaminase Fusion Allosterically Converts ATP to ITP

The recently identified CARF (CRISPR-associated Rossman-fold) family of proteins play a critical role in prokaryotic defense, mediating cOA (cyclic oligoadenylate)-stimulated ancillary immune responses in the type III CRISPR-Cas systems. Whereas most previously characterized CARF proteins contain nucleic acids or protein degradation effectors, a subset comprises CARF-fused adenosine deaminase (ADA) (Cad1) and has a yet to be determined function. Here we present biochemical and structural analyses of a ring nuclease Cad1, revealing its unexpected role in deaminating adenosine-5'-triphosphate to inosine-5'-triphosphate in a cOA-dependent manner. Despite an overall structural similarity to canonical ADA enzymes, the ADA domain of Cad1 possesses unique structural features underlying its specificity for ATP. Supported by mutational analysis, we demonstrate an allosteric link between the cOA-binding CARF and the ADA domain, suggesting that Cad1 is a cOA-stimulated effector that influences cellular metabolic processes. HighlightsO_LITaqCad1 converts ATP to ITP in a magnesium- and cOA-dependent manner C_LIO_LICryo-EM structures reveal TaqCad1 forms a homohexamer C_LIO_LICryo-EM structures reveal how cA4 is degraded and modulates the ADA active site C_LI

cell biology↗

bioRxiv2024

Bridging the Dimensional Gap from Planar Spatial Transcriptomics to 3D Cell Atlases

Spatial transcriptomics (ST) has revolutionized our understanding of tissue architecture, yet constructing comprehensive three-dimensional (3D) cell atlases remains challenging due to technical limitations and high cost. Current approaches typically capture only sparsely sampled two-dimensional sections, leaving substantial gaps that limit our understanding of continuous organ organization. Here, we present SpatialZ, a computational framework that bridges these gaps by generating virtual slices between experimentally measured sections, enabling the construction of dense 3D cell atlases from planar ST data. SpatialZ is designed to operate at single-cell resolution and function independently of gene coverage limitations inherent to specific spatial technologies. Comprehensive validation using real 3D ST and independent serial sectioning datasets demonstrates that SpatialZ accurately reconstructs virtual slices while preserving cell identities, gene expression patterns, and spatial relationships. Leveraging the BRAIN Initiative Cell Census Network data, we constructed a 3D hemisphere atlas comprising over 38 million cells, a scale not feasible experimentally. This dense atlas enables unprecedented capabilities, including in silico sectioning at arbitrary angles, explorations of gene expression across both 3D volumes and surfaces, and 3D mapping of query tissue sections. While currently validated for spatial transcriptomics, the underlying principles of SpatialZ could potentially be adapted for spatial proteomics, spatial metabolomics, and even spatial multi-omics. Validated through internal and external testing, our computationally generated atlas maintains biological accuracy, providing unprecedented resolution of spatial molecular landscapes and demonstrating the potential of computational approaches in advancing 3D ST.

bioinformatics↗

bioRxiv2024

NicheTrans: Spatial-aware Cross-omics Translation

Spatial omics technologies have revolutionized our studies on tissue architecture and cellular interactions at single-cell resolution. While spatial multi-omics approaches offer unprecedented insights into complex biological systems, their widespread adoption is hindered by technical challenges, specialized requirements, and limited accessibility. To address these limitations, we present NicheTrans, the first spatially-aware cross-omics translation method and a flexible Transformer-based multi-modal deep learning framework. Unlike existing single-cell (non-spatial) translation methods, NicheTrans uniquely incorporates both cellular microenvironment information and flexible integration of multi-modal data, such as morphology and prior knowledge. We validated NicheTrans across diverse biological cases: Parkinsons Disease (PD), Alzheimers Disease (AD), breast cancer, and lymph nodes. Our approach demonstrated superior performance compared to existing single-cell methods, highlighting the crucial role of spatial and multi-modal information in cross-omics translation. Through NicheTrans, we uncovered spatial multi-omics domains that were not detectable through single-omics analysis alone. Model interpretation revealed key molecular relationships, including gene programs associated with dopamine metabolism and amyloid {beta}-associated cell states. Additionally, using translated protein markers as spatial landmarks, we quantified the spatial organization of key glial cell subtypes in the AD brain. NicheTrans represents a powerful tool for generating comprehensive spatial multi-omics insights from more accessible single-omics measurements, making multi-omics analysis more feasible for the broader research community.

bioinformatics↗

bioRxiv2024

Deep Ensemble Model for Quantitative Optical Property and Chromophore Concentration Images of Biological Tissues

The ability to quantify widefield tissue optical properties (OPs, i.e., absorption and scattering) has major implications on the characterization of various physiological and disease processes. However, conventional image processing methods for tissue optical properties are either limited to qualitative analysis, or have tradeoffs in speed and accuracy. The key to quantification of optical properties is the extraction of amplitude maps from reflectance images under sinusoidal illumination of different spatial frequencies. Conventional three-phase demodulation (TPD) method has been demonstrated for the mapping of OPs, but it requires as many as 14 measurement images for accurate OP extraction, which leads to limited throughput and hinders practical translation. Although single-phase demodulation (SPD) method has been proposed to map OPs with a single measurement image, it is typically subject to image artifacts and decreased measurement accuracy. To tackle those challenges, here we develop a deep ensemble model (DEM) that can map tissue optical properties with high accuracy in a single snapshot, increasing the measurement speed by 14x compared to conventional TPD method. The proposed method was validated with measurements on an array of optical phantoms, ex vivo tissues, and in vivo tissues. The errors for OP extraction were 0.83 {+/-} 5.0% for absorption and 0.40 {+/-} 1.9% for reduced scattering, dramatically lower than that of the state-of-the-art SPD method (2.5 {+/-} 15% for absorption and -1.2 {+/-} 11% for reduced scattering). It was further demonstrated that while trained with data from a single wavelength, the DEM can be directly applied to other wavelengths and effectively obtain optical property and chromophore concentration images of biological tissues. Together, these results highlight the potential of DEM to enable new capabilities for quantitative monitoring of tissue physiological and disease processes.

biophysics↗

bioRxiv2024

Volitional Regulation and Transferable Patterns of Midbrain Oscillations

Dopaminergic brain areas are crucial for cognition and their dysregulation is linked to neuropsychiatric disorders typically treated with pharmacological interventions. These treatments often have side effects and variable effectiveness, underscoring the need for alternatives. We introduce the first demonstration of neurofeedback using local field potentials (LFP) from the ventral tegmental area (VTA). This approach leverages the real-time temporal resolution of LFP and ability to target deep brain. In our study, non-human primates learned to regulate VTA beta power using a customized normalized metric to stably quantify VTA LFP signal modulation. The subjects demonstrated flexible and specific control with different strategies for specific frequency bands, revealing new insights into the plasticity of VTA neurons contributing to oscillatory activity that is functionally relevant to many aspects of cognition. Excitingly, the subjects showed transferable patterns, a key criterion for clinical applications beyond training settings. This work provides a foundation for neurofeedback-based treatments, which may be a promising alternative to conventional approaches and open new avenues for understanding and managing neuropsychiatric disorders. Significance statementThis study demonstrates, for the first time, that neurofeedback using local field potentials (LFP) from the ventral tegmental area (VTA) is feasible in non-human primates. By leveraging the temporal resolution and ability to target deep brain regions, this approach provides a novel way to modulate brain activity linked to dopamine-related functions. The findings reveal that subjects can flexibly control VTA LFP signals and transfer learned strategies to new settings, offering potential for developing neurofeedback-based treatments. This research opens new avenues for managing neuropsychiatric disorders, presenting an alternative to traditional pharmacological interventions that often have side effects and limited effectiveness. The study highlights the plasticity of VTA neurons and their relevance to cognition and mood regulation.

neuroscience↗

bioRxiv2024

Structural and functional insights into activation and regulation of the dynein-dynactin-NuMA complex

During cell division, NuMA orchestrates the focusing of microtubule minus-ends in spindle poles and cortical force generation on astral microtubules by interacting with dynein motors, microtubules, and other cellular factors. Here, we used in vitro reconstitution, cryo-electron microscopy, and live cell imaging to understand the mechanism and regulation of NuMA. We determined the structure of the processive dynein/dynactin/NuMA complex (DDN) and showed that the NuMA N-terminus drives dynein motility in vitro and facilitates dynein-mediated transport in live cells. The C-terminus of NuMA directly binds to and suppresses the dynamics of the microtubule minus-end. Full-length NuMA is autoinhibited for its interactions with dynein and microtubules, but mitotically phosphorylated NuMA activates dynein in vitro and interphase cells. Together with dynein, activated full-length NuMA focuses microtubule minus-ends into aster-like structures. These results provide critical insights into the activation of NuMA and dynein for their mitotic functions.

biophysics↗

bioRxiv2024

BPMs regulate Arabidopsis seedling development by promoting auxin-independent degradation of the Aux/IAA protein IAA10

After germination, seedlings undergo etiolated development (skotomorphogenesis), enabling them to grow towards the soil surface. In Arabidopsis, etiolated seedlings exhibit rapid hypocotyl elongation, apical hook formation and closed cotyledons to protect the meristem. In this study, we found that high-order mutants in the BPM gene family displayed defects in seedling development, characterized by a shorter hypocotyl, early apical hook opening, and opened cotyledons in the dark. BPM1, BPM2, BPM4, and BPM5 exhibit distinct expression patterns and subcellular localization in etiolated seedlings. In a hypocotyl segment assay the bpm mutants showed defects in auxin response indicating impaired auxin signaling in the hypocotyl. Expression of the auxin reporter DR5:GFP was also altered in the bpm1,4,5 mutant in various tissues compared to the wild type. Furthermore, we showed that BPM1 and IAA10 interact in yeast two-hybrid, BiFC, and Co-IP assays. Experiments in protoplasts indicated that BPM1 promotes ubiquitylation and degradation of IAA10, and the level of IAA10 protein is greater in the bpm1,4,5 mutant. In addition, IAA10 over-expression resulted in phenotypes similar to the bpm mutants. These results indicate that the BPMs target the Aux/IAA proteins for ubiquitylation and degradation. Overall, our findings shed light on the key roles of the BPMs in auxin signaling during seedling development.

plant biology↗

bioRxiv2024

The dual ubiquitin binding mode of SPRTN secures rapid spatiotemporal proteolysis of DNA-protein crosslinks

DNA-protein crosslinks (DPCs) are endogenous and chemotherapy-induced genotoxic DNA lesions and, if not repaired, lead to embryonic lethality, neurodegeneration, premature ageing, and cancer. DPCs are heavily polyubiquitinated, and the SPRTN protease and 26S proteasome emerged as two central enzymes for DPC proteolysis. The proteasome recognises its substrates by their ubiquitination status. How SPRTN protease, an essential enzyme for DPC proteolysis, achieves specificity for DPCs still needs to be discovered. We found that the N-terminal SPRTN catalytic region (SprT) possesses a ubiquitin-binding domain named the Ubiquitin interface of SprT Domain (USD). Using multiple biochemical, biophysical, and structural approaches, we reveal that USD binds ubiquitin chains. SPRTN binding to ubiquitin chains via USD leads to [~] 67-fold higher activation of SPRTN proteolysis towards polyubiquitinated DPCs than the unmodified DPCs. This study reveals the ubiquitination of DPCs is the key signal for SPRTNs substrate specificity and rapid proteolysis.

biochemistry↗

bioRxiv2024

Natural killer cell dysfunction drives keloid pathogenesis

Keloids, characterized by excessive scar tissue resulting from abnormal wound healing, are primarily driven by hyperproliferative fibroblasts and overproduction of extracellular matrix. Although human natural killer (NK) cells are known for their role in inhibiting uncontrolled cell growth through cytokines like IFN-{gamma} and TNF-, they also produce amphiregulin (AREG), which paradoxically promotes cell proliferation and survival. The involvement of NK cells in keloid pathogenesis, however, remains largely unexplored. This study uncovers remarkable functional changes in NK cells within both lesional skin and the blood of keloid patients. In the skin, NK cell-produced IFN-{gamma} plays a pivotal role in limiting keloid progression by inducing fibroblast apoptosis and curbing excessive extracellular matrix production, while NK cell-derived AREG actively opposes these protective effects. Notably, TGF-{beta}-driven fibroblasts in keloid lesions further dampen NK cell IFN-{gamma} production, revealing a complex and dynamic cellular interplay. In the bloodstream of keloid patients, a distinct NK cell subset emerges, marked by elevated interferon-stimulated genes (ISGs) and diminished IFN-{gamma} production, which correlates with increased plasma IFN-{beta} levels. This elevated IFN-{beta} serves as a key initiating factor, driving NK cell exhaustion through impaired mitochondrial function and metabolic disruption. These findings highlight a critical mechanism underlying the functional abnormalities of keloid-associated NK cells and emphasize the influence of both local and systemic factors in shaping NK cell responses in keloid pathogenesis. Graphical abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=132 SRC="FIGDIR/small/624588v1_ufig1.gif" ALT="Figure 1"> View larger version (40K): org.highwire.dtl.DTLVardef@7501f4org.highwire.dtl.DTLVardef@1d34de5org.highwire.dtl.DTLVardef@19f1c47org.highwire.dtl.DTLVardef@f98882_HPS_FORMAT_FIGEXP M_FIG C_FIG HighlightsO_LIIn the skin, NK cell-derived IFN-{gamma} is crucial for suppressing keloid fibroblast proliferation and extracellular matrix production. C_LIO_LIAREG from NK cells counteracts the inhibitory effects of IFN-{gamma} on keloid progression. C_LIO_LIIn the blood, elevated IFN-{beta} in keloid patients drives the emergence of ISG+NK cells with reduced IFN-{gamma} production. C_LIO_LIIFN-{beta} induces mitochondrial dysfunction and functional exhaustion in NK cells. C_LI

immunology↗

bioRxiv2024

Comparison of genomic assembly and annotation based on two clones of avian pathogenic Escherichia coli

Methods for assembly and annotation of whole genomic sequences were compared for six strains of avian pathogenic Escherichia coli (APEC). Two vertically transferred E. coli clones, represented by three isolates all belonging to pulse field genome electrophoresis (PFGE) type 65- sequence type (ST)95 and three isolates belonging to PFGE type 47- ST131, were selected for Illumina short read sequencing. There was no significant difference between SPAdes and CLC Genomic Workbench for benchmark parameters to assemble the short reads. The six strains were also sequenced by long read sequencing (Nanopore) and these reads were hybrid assembled with the short reads. Unicycler provided a lower number of contigs and higher N50 compared to Flye. No significant differences between total length of genomes were obtained from the four assemblers. At least 2.1 and 0.9% of coding gene sequences (CDSs) annotated with RAST and PROKKA, respectively were wrongly annotated. The errors were most often associated to CDS of shorter length (< 150 nt) with functions such as transposases, mobile genetic elements or being hypothetical. The investigation points out the importance of controlling automatic annotations and suggest further work to improve annotations in strains not belonging to the K12 or B lineages.

bioinformatics↗

bioRxiv2024

Ultra-small Particle Size Nanobubbles Water Drinking Exerts Antifatigue Effects in Chronic Exercise Fatigue Rats via Improving Antioxidative Activities

Chronic exercise fatigue is a prevalent issue that can lead to a decline in physical performance, metabolic dysregulation, and oxidative damage, posing significant health risks. This study aims to investigate the effects of Ultra-small Particle Size Nanobubbles Water (UsNBsW) which was Koishio technology produced water on physiological and biochemical parameters in a rat model of chronic exercise fatigue. We established a chronic fatigue model, and evaluated various physiological and biochemical indicators. Our key findings revealed that Koishio water significantly improved rats swimming performance, reduced blood lactate levels, and enhanced antioxidant capacity, as evidenced by increased hemoglobin content, decreased lactate dehydrogenase and creatine kinase activities (p<0.01). Additionally, liver glycogen levels were notably elevated (p<0.05). Futhermore, the antioxidant markers, including superoxide dismutase (SOD) and glutathione peroxidase (GSH-Px) activities were significantly higher (p<0.01), while malondialdehyde (MDA) and 8-hydroxydeoxyguanosine (8-OH-dG) levels were markedly reduced (p<0.01). Our results provide new clues for the promising role of Koishio water in mitigating chronic exercise fatigue and supply a reliable reference for future research into nutritional interventions and recovery strategies in exercise physiology.

physiology↗

bioRxiv2024

Single cell analysis of Barrett's esophagus and carcinoma reveals cell types conferring risk via genetic predisposition

Inherited genetic variants contribute to Barretts esophagus (BE) and esophageal adenocarcinoma (EAC) but it is unknown which cell types are involved in this process. We performed single cell RNA-sequencing of BE, EAC and paired normal tissues and integrated data of a genome-wide association study to determine cell type-specific genetic risk and cellular processes that contribute to BE and EAC. The analysis revealed that EAC development is driven to a greater extent by local cellular processes than BE development. One cell type of BE origin (BE-EAC) and cellular processes that control the differentiation of columnar cells are of particular relevance for EAC development. Further, specific subtypes of fibroblasts and endothelial cells contribute to BE and EAC development, while dendritic cells and CD4+ memory T cells contribute exclusively to BE development. The diagnostic use of markers characterizing the identified cell types and cellular processes should be explored in future for EAC prediction.

cancer biology↗

bioRxiv2024

ILSM: A package for analyzing the interconnection structure of tripartite interaction networks

O_LIIn natural or human-disturbed ecosystems, ecological networks often comprise multiple interaction types, which have been often represented by multipartite ecological networks. One important aspect of their network architecture is how different interaction types or subnetworks are interconnected by shared species, here defined as the interconnection structure. Previous studies have proposed various indices of shared species to characterize macro-scale interconnection patterns and micro-scale centrality, but the meso-scale interconnection structures (here defined as interconnection motifs) remain largely unexplored. Furthermore, there is no package available in the R programming language for conducting analyses of various interconnection structures. C_LIO_LIWithin a tripartite network with two interaction subnetworks, we define the forms of interconnection motifs and unique roles within these motifs. Then we introduce the R package ILSM for analyzing interconnection patterns, interconnection centrality, and interconnection motif. Specifically, we derive mathematical expressions for the frequencies of interconnection motifs and species roles within motifs. C_LIO_LIWe describe the main functions in the package and demonstrate their uses with an example pollinator-plant-herbivore network. C_LIO_LIILSM will help ecologists understand how different types of interactions are interconnected together by shared species using interconnection pattern, centrality and motif. C_LI

ecology↗

bioRxiv2024

Upregulating ANKHD1 in PS19 mice reduces Tau phosphorylation and mitigates Tau-toxicity-induced cognitive deficits

Abnormal accumulation of Tau protein in the brain disrupts normal cellular function and leads to neuronal death linked with many neurodegenerative disorders such as Alzheimers disease. An attractive approach to mitigate Tau-induced neurodegeneration is to enhance the clearance of toxic Tau aggregates. We previously showed that upregulation of the fly gene mask protects against FUS- and Tau-induced photoreceptor degeneration in fly disease models. Here we have generated a transgenic mouse line carrying Cre-inducible ANKHD1, the mouse homolog of mask, to determine whether the protective role of mask is conserved from flies to mammals. Utilizing the TauP301S-PS19 mouse model for Tau-related dementia, we observed that ANKHD1 significantly reduced hyperphosphorylated human Tau in 6-month-old mice. Additionally, there was a notable trend towards reduced gliosis levels in these mice, suggesting a protective role of ANKHD1 against TauP301S-linked degeneration. Further analysis of 9-month- old mice revealed a similar trend of effects. Moreover, we found that ANKHD1 also suppresses the cognitive defect of 9-month-old PS19 female mice in novel object recognition (NOR) behavioral assay. Unlike previous therapeutic strategies that primarily focus on inhibiting Tau phosphorylation or directly clearing aggregates, this study highlights the novel role of ANKHD1 in promoting autophagy as a means to mitigate Tau pathology. This novel mechanism not only underscores ANKHD1s potential as a unique therapeutic target for tauopathies but also provides new insights into autophagy-based interventions for neurodegenerative diseases.

neuroscience↗

bioRxiv2024

Distinct Binding Modes of Inter-Spike Cross-Linking Suggest a Supplementary Mechanism for SARS-CoV-2 Antibody Neutralization

The emergence of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and its Omicron subvariants drastically amplifies transmissibility, infectivity, and immune escape, mainly due to their resistance to most neutralizing antibodies. Thus, exploring the mechanisms underlying antibody evasion is crucial. Although the full-length native form of antibody, immunoglobulin G (IgG), offers valuable insights into the neutralization, structural investigations primarily focus on the fragment of antigen-binding (Fab). Here, we employ single-particle cryo-electron microscopy (cryo-EM) to characterize a W328-6H2 antibody, in its native IgG form complexed with severe acute respiratory syndrome (SARS), severe acute respiratory syndrome coronavirus 2 wild-type (WT) and Omicron variant BA.1 spike protein (S). Three high-resolution structures reveal that the full-length IgG forms a centered head-to-head dimer of trimer when bound fully stoichiometrically with both SARS and WT S, while adopting a distinct offset configuration with Omicron BA.1 S. Combined with functional assays, our results suggest that, beyond the binding affinity between the RBD epitope and Fab, the higher-order architectures of S trimer and full-length IgG play an additional role in neutralization, elucidate a potential molecular basis of Omicron immune escape, and expand our understanding on antibody evasion mechanisms.

immunology↗

bioRxiv2024

Distilling structural representations into protein sequence models

Protein language models, like the popular ESM2, are widely used tools for extracting evolution-based protein representations and have achieved significant success on downstream biological tasks. Representations based on sequence and structure models, however, show significant performance differences depending on the downstream task. A major open problem is to obtain representations that best capture both the evolutionary and structural properties of proteins in general. Here we introduce Implicit Structure Model (ISM), a sequence-only input model with structurally-enriched representations that outperforms state-of-the-art sequence models on several well-studied benchmarks including mutation stability assessment and structure prediction. Our key innovations are a microenvironment-based autoencoder for generating structure tokens and a self-supervised training objective that distills these tokens into ESM2s pre-trained model. We have made ISMs structure-enriched weights easily available: integrating ISM into any application using ESM2 requires changing only a single line of code. Our code is available at https://github.com/jozhang97/ISM.

bioinformatics↗