Biology subjects

Wang, B.

Publications and source records attributed to Wang, B..

At least 415 records · Page 23Linked to original sources

bioRxiv2023

An ULK1/2-PXN mechanotransduction complex suppresses breast cancer cell migration

The remodeling and stiffening of the extracellular matrix (ECM) associated with breast cancers is a well-recognized modulator of disease progression. However, how changes in the mechanical properties of the ECM are converted into biochemical signals that direct tumor cell migration and metastasis remains poorly characterized. Here, we describe a new role for the autophagy-inducing serine/threonine kinases ULK1 and ULK2 in mechanotransduction. We demonstrate that ULK1/2 activity inhibits the assembly of actin stress fibers and focal adhesions (FAs), and as a consequence impedes cell contraction and migration. Mechanistically, we identify PXN/paxillin, a key component of the mechanotransducing machinery, as a direct binding partner and substrate of ULK1/2. ULK-mediated phosphorylation of PXN at S32 and S119 weakens homotypic interactions and liquid-liquid phase separation of PXN, impairing FA assembly, which in turn impedes the mechanotransduction of breast cancer cells. ULK1/2 and the well characterized PXN regulator, FAK/Src, have opposing functions on mechanotransduction and compete for phosphorylation of adjacent serine and tyrosine residues. Thus, our study reveals ULK1/2 as important regulators of PXN-dependent mechanotransduction. HighlightsO_LIULK1/2 interact with PXN and phosphorylate PXN at S32 and S119 in response to mechanical stimuli C_LIO_LIULK1/2-mediated phosphorylation of PXN regulates mechanotransduction and migration of breast cancer cells C_LIO_LIULK1/2 modulate the biomaterial properties of focal adhesions through PXN phosphorylation C_LIO_LIULK1/2 and FAK/Src act antagonistically in mechanotransduction through competitive phosphorylation of PXN C_LI

cell biology↗

bioRxiv2023

In vivo Modulation of Intraocular and Intracranial Pressures Causes Nonlinear and Non-monotonic Deformations of the Lamina Cribrosa and Scleral Canal

PurposeTo evaluate changes in monkey optic nerve head (ONH) morphology under acutely controlled intraocular pressure (IOP) and intracranial pressure (ICP). MethodsSeven ONHs from six monkeys were imaged via optical coherence tomography while IOP and ICP were maintained at one of 16 conditions. These conditions were defined by 4 levels for each pressure: low, baseline, high and very high. Images were processed to determine scleral canal area, aspect ratio, and planarity and anterior lamina cribrosa (ALC) shape index and curvature. Linear mixed effect models were utilized to investigate the effects of IOP, ICP and their interactions on ONH morphological features. The IOP-ICP interaction model was compared with one based on translaminar pressure difference (TLPD). ResultsWe observed complex, eye-specific, non-linear patterns of ONH morphological changes with changes in IOP and ICP. For all ONH morphological features, linear mixed effects models demonstrated significant interactions between IOP and ICP that were unaccounted for by TLPD. Interactions indicate that the effects of IOP and ICP depend on the other pressure. The IOP-ICP interaction model was a higher quality predictor of ONH features than a TLPD model. ConclusionsIn vivo modulation of IOP and ICP causes nonlinear and non-monotonic changes in monkey ONH morphology that depend on both pressures and is not accounted for by a simplistic TLPD. These results support and extend prior findings. Translational Relevance: A better understanding of ICPs influence on the effects of IOP can help inform the highly variable presentations of glaucoma and effective treatment strategies.

bioengineering↗

bioRxiv2023

INHIBITORY EFFECTS OF GLUCOCORTICOID ON IMMUNOCYTES

Glucocorticoids (GCs) are efficacious agents for reducing inflammation and suppressing immune responses, exerting various effects on immune cells through the intracellular glucocorticoid receptor (GR), and impacting both innate and adaptive immunity. In the context of COVID-19, glucocorticoids are often used to treat severe cases of patients by reducing inflammation, suppressing immune responses, and ameliorating the severity of COVID-19. However, the precise inhibitory effects on immune cells have yet to be comprehensively delineated. In this study, we extensively examined the inhibitory effects of treating Balb/c mice with dexamethasone (DEX) on lymphoid and myeloid cells. We observed that high doses of DEX treatment resulted in a reduction in the number of immunocytes and an attenuation of their activity. Particularly noteworthy, macrophages, DC cells, and monocytes were diminished by approximately 90% following high doses of DEX, while B cells experienced a reduction of about 70% and CD3 T cells were less affected. Furthermore, our findings demonstrated that DEX induces the inhibition of immune cells by engaging in high-affinity binding to GR. Consequently, we conclude that DEX treatments affect a broad range of immune cells, encompassing both lymphoid and myeloid cells, through depletion or the down-regulation of immune function, potentially acting via the GR signaling pathway. These findings may enhance the clinical applicability of DEX in achieving transient immune deficiency.

immunology↗

bioRxiv2023

Adaptive robustness through incoherent signaling mechanisms in a regenerative brain

Animal behavior emerges from collective dynamics of interconnected neurons, making it vulnerable to connectome damage. Paradoxically, many organisms maintain significant behavioral output after large-scale neural injury. Molecular underpinnings of this extreme robustness remain largely unknown. Here, we develop a quantitative behavioral analysis pipeline to measure previously uncharacterized long-lasting latent memory states in planarian flatworms during whole-brain regeneration. By combining >20,000 animal trials with neural population dynamic modeling, we show that long-range volumetric peptidergic signals allow the planarian to rapidly reestablish latent states and restore coarse behavior after large structural perturbations to the nervous system, while small-molecule neuromodulators gradually refine the precision. The different time and length scales of neuropeptide and small-molecule transmission generate incoherent patterns of neural activity which competitively regulate behavior and memory. Controlling behavior through opposing communication mechanisms creates a more robust system than either alone and may serve as a generic approach to construct robust neural networks.

neuroscience↗

bioRxiv2023

Decoupling the correlation between cytotoxic and exhausted T lymphocyte transcriptomic signatures enhances melanoma immunotherapy response prediction

BackgroundCytotoxic T lymphocytes (CTL) play a crucial role in anti-cancer immunity. Progression of CTL to terminal exhausted T lymphocytes (ETL) that overexpress inhibitory receptors can substantially decrease effector cytokines production and diminish cytolytic activity and terminal exhausted T cell cannot be reprogrammed by ICIs in tumor microenvironment (TME). However, while the activity levels of CTL and ETL are considered important determinants of immune checkpoint inhibitors (ICIs) response, it has been repeatedly observed that their predictive power of the latter is quite limited. Studying this conundrum on a large scale across the TCGA cohort, we find that ETL and CTL activity (estimated based on conventional gene signatures in the bulk tumor expression) is strongly positively correlated in most cancer types. We hypothesized that the limited predictive power of CTL activity might result from the high concordance of CTL and ETL activities, which mutually cancels out their individual antagonistic effects on ICI response. MethodsConsequently, we have set out to identify a set of genes whose expression identifies a subset of patients where the CTL and ETL correlation is diminished, such that the association between these CD8+ T cell states and ICIs response is enhanced. Resultsnalyzing TCGA melanoma bulk gene expression, we identified a set of genes whose over-expression markedly diminishes the CTL and ETL correlation, termed a decoupling signature (DS). Reassuringly, we first find that the correlation between ETL and CTL activities is indeed markedly lower across high scoring DS patients than that observed across low scoring DS patients in numerous independent melanoma ICIs cohorts. Second, indeed, this successful decoupling increases the power of CTL activity in predicting ICIs response in high DS scoring patients. We show that the resulting prediction accuracy is superior to other state-of-art ICI predictive transcriptomic signatures. ConclusionThe new decoupling score boosts the power of CTL activity in predicting ICIs response in melanoma from the tumor bulk expression. Its use enables a two-step stratification approach, where the response of high scoring DS patient can be predicted more accurately that with extant transcriptomic signatures. What is already known on this topicThe predictive power of CTL activity based on bulk tumor transcriptomics, despite being a widely studied important determinant of ICI treatment, is very limited. What this study addsThe efficacy of CTL activity in predicting ICI therapy response is significantly higher among patients with decoupled CTL and ETL activities. How this study might affect research, practice or policyWe identified a set of genes as the decoupling signature, whose upregulation markedly diminishes the correlation between CTL and ETL activities. Our decoupling signature enhances the power of CTL in predicting ICI treatment response, outperforming other extant expression-based signatures.

bioinformatics↗

bioRxiv2023

Competing differentiation gradients coordinate fruit morphogenesis

Morphogenesis requires the coordination of cellular behaviors along developmental axes1. In plants, gradients of growth and differentiation are typically established along a single longitudinal primordium axis to control organ shaping2. Here we combine quantitative live-imaging at cellular resolution with genetics, chemical treatments, and modeling to understand the formation of Arabidopsis thaliana female reproductive organ (gynoecium). We show that, contrary to other aerial organs, gynoecium shape is determined by two competing differentiation gradients positioned along two orthogonal axes. An early mediolateral gradient, dependent on meristematic activity in the medial domain, controls the valve morphogenesis while simultaneously restricting an auxin-dependent, longitudinal gradient to the style. This gradient competition serves to finetune the common developmental program governing organ morphogenesis to ensure the specialized function of the gynoecium3,4.

plant biology↗

bioRxiv2022

A toolkit for converting Gal4 into LexA and Flippase transgenes in Drosophila

Drosophila has been a powerful model system for biological studies due to the wide range of genetic tools established for it. Among these tools, Gal4 is the most abundant, offering unparalleled tissue- and developmental stage-specificity for gene manipulation. In comparison, other genetic reagents are far fewer in choices. Here we present a genetic toolkit for converting Gal4 strains into LexA and Flippase transgenes through simple genetic crosses and fluorescence screening. We demonstrate the proof-of-principle by converting ten Gal4 lines that exhibit diverse tissue specificities and examined the activity patterns of the converted LexA and Flippase lines. Gal4-to-LexA and Flp conversion is fast and convenient and should greatly expand the choices of LexA and Flp for binary expression and FRT-based mosaic analysis, respectively, in Drosophila.

genetics↗

bioRxiv2022

Genetic basis and selection of glyceollin induction in wild soybean

O_LIGlyceollins, a family of phytoalexin induced in legume species, play essential roles in responding to environmental stresses and in human health. However, little is known about the genetic basis and selection of glyceollin induction. C_LIO_LIWe employed a metabolite-based genome-wide association (mGWA) approach to identify candidate genes involved in glyceollin induction from genetically diverse and understudied wild soybeans subjected to soybean cyst nematode stress. C_LIO_LIEight SNPs on chromosomes 3, 9, 13, 15, and 20 showed significant association with glyceollin induction. Six genes close to one of the significant SNPs (ss715603454) on chromosome 9 fell into two clusters, and they encode enzymes in the glycosyltransferase class within the phenylpropanoid pathway. Transcription factors (TFs) genes, such as MYB and WRKY were also found within the linkage disequilibrium of the significant SNPs on chromosome 9. Epistasis and a strong selection signal were detected on the four significant SNPs on chromosome 9. C_LIO_LIGene clusters and transcription factors may play important roles in regulating glyceollin induction in wild soybeans. Additionally, as major evolutionary factors, epistatic interactions and selection may influence glyceollin variation in natural populations. C_LI

plant biology↗

bioRxiv2022

PXN/Paxillin Phase Separation Promotes Focal Adhesion Assembly and Integrin Signaling

Focal adhesions (FAs) are transmembrane protein assemblies mediating cell-matrix connection. Tools to manipulate the compositionally intricate and dynamic FAs are currently limited, rendering many fundamental hypotheses untestable. Although protein liquid-liquid phase separation (LLPS) has been tied to the organization and dynamics of FAs, the underlying mechanisms remain unclear. Here, we experimentally tune the LLPS of PXN/Paxillin, an essential scaffold protein of FAs, by utilizing light-inducible Cry2 system. In addition to nucleating FA components, light-triggered PXN LLPS potently activates integrin signaling and subsequently accelerates cell spreading. PXN favors homotypic interaction-driven LLPS in vitro. In cells, PXN condensates are associated with plasma membrane, and modulated by actomyosin contraction and client proteins of FAs. Interestingly, non-specific weak inter-molecular interactions, together with specific molecular interactions, underlie the multicomponent condensation of PXN, and are efficient to promote FA assembly and integrin signaling. Thus, our data establish an active role of PXN phase transition into a condensed membrane-associated compartment in promoting assembly/maturation of FAs.

cell biology↗

bioRxiv2022

Cellular integration with a subretinal honeycomb-shaped prosthesis

In patients blinded by geographic atrophy, subretinal photovoltaic implant with 100{micro}m pixels provided visual acuity closely matching the pixel pitch. However, such flat bipolar pixels cannot be scaled below 75{micro}m, limiting the attainable visual acuity. This limitation can be overcome by shaping the electric field with 3-dimensional electrodes. In particular, elevating the return electrode on top of honeycomb-shaped vertical walls surrounding each pixel extends the electric field vertically and decouples its penetration into tissue from the pixel width. This approach relies on migration of the retinal cells into the honeycomb wells. Here, we demonstrate that the majority of the inner retinal neurons migrate into 25{micro}m deep wells, leaving the third-order neurons, such as amacrine and ganglion cells, outside. This is important for selective stimulation of the second-order neurons to preserve the retinal signal processing in prosthetic vision. Comparable glial response to that with flat implants suggests that migration and separation of the retinal cells by the walls does not cause additional stress. Furthermore, retinal migration into the honeycombs does not negatively affect its electrical excitability.

neuroscience↗

bioRxiv2022

Disparate roles for C. elegans DNA translocase paralogs RAD-54.L and RAD-54.B in meiotic prophase germ cells

RAD54 family DNA translocases partner with RAD51 recombinases to ensure stable genome inheritance, exhibiting biochemical activities both in promoting recombinase removal and in stabilizing recombinase association with DNA. Understanding how such disparate activities of RAD54 paralogs align with their biological roles is an ongoing challenge. Here we investigate the in vivo functions of C. elegans RAD54 paralogs RAD-54.L and RAD-54.B during meiotic prophase, revealing distinct contributions to the dynamics of RAD-51 association with DNA and to the progression of meiotic double-strand break repair (DSBR). While RAD-54.L is essential for RAD-51 removal from meiotic DSBR sites to enable recombination progression, RAD-54.B is largely dispensable for meiotic DSBR. However, RAD-54.B is required to prevent hyperaccumulation of RAD-51 on unbroken DNA during a key meiotic sub-stage when protein kinase CHK-2 is active. Moreover, DSB-independent hyperaccumulation of RAD-51 foci in the absence of RAD-54.B is RAD-54.L-dependent, revealing a hidden activity of RAD-54.L in promoting promiscuous RAD-51 association that is antagonized by RAD-54.B. We propose a model wherein a division of labor among RAD-54 paralogs allows germ cells to ramp up their capacity for efficient homologous recombination that is crucial to successful meiosis while counteracting potentially deleterious effects of unproductive RAD-51 association with unbroken DNA.

genetics↗

bioRxiv2022

Fortuitous Somatic Mutations during Antibody Evolution Endow Broad Neutralization against SARS-CoV-2 Omicron Variants

Striking antibody evasion by emerging circulating SARS-CoV-2 variants drives the identification of broadly neutralizing antibodies (bNAbs). However, how a bNAb acquires increased neutralization breadth during antibody evolution is still elusive. Here, we identified a clonally-related antibody family from a convalescent individual. One of the members, XG005, exhibited potent and broad neutralizing activities against SARS-CoV-2 variants, while the other members showed significant reductions in neutralization breadth and potency, especially against the Omicron sublineages. Structural analysis visualizing the XG005-Omicron spike binding interface revealed how crucial somatic mutations endowed XG005 with greater neutralization potency and breadth. A single administration of XG005 with extended half-life, reduced antibody-dependent enhancement (ADE) effect, and increased antibody product quality, exhibited a high therapeutic efficacy in BA.2- and BA.5-challenged mice. Our results provided a natural example to show the importance of somatic hypermutation during antibody evolution for SARS-CoV-2 neutralization breadth and potency.

immunology↗

bioRxiv2022

Pleiotropy of autism-associated chromatin regulators

Gene ontology analyses of high confidence autism spectrum disorder (hcASD) risk genes have historically highlighted chromatin regulation and synaptic function as major contributors to pathobiology. Our recent functional work in vivo has additionally implicated microtubule biology and identified disrupted cellular proliferation as a convergent ASD phenotype. As many chromatin regulators, including ASD risk genes ADNP and CHD3, are known to directly regulate both tubulins and histones, we studied the five chromatin regulators most strongly associated with ASD (ADNP, CHD8, CHD2, POGZ, and SUV420H1/KMT5B) specifically with respect to microtubule biology. We observe that all five localize to microtubules of the mitotic spindle in vitro and in vivo. Further in-depth investigation of CHD2 provides evidence that patient-derived mutations lead to a range of microtubule-related phenotypes, including disrupted localization of the protein at the mitotic spindle, spindle defects, cell cycle stalling, DNA damage, and cell death. Lastly, we observe that ASD genetic risk is significantly enriched among microtubule-associated proteins, suggesting broader relevance. Together, these results provide further evidence that the role of tubulin biology and cellular proliferation in ASD warrant further investigation and highlight the pitfalls of relying solely on annotated gene functions in the search for pathological mechanisms.

developmental biology↗

bioRxiv2022

Species and condition dependent mutational spectrum in experimentally evolved biofilms of Bacilli

Laboratory experimental evolution provides a powerful tool for studying microbial adaptation to different environments. To understand the differences and similarities of the dynamic evolutionary landscapes of two model species from the Bacillus genus as they adapt to abiotic and biotic surfaces, we revived the archived population samples from our four previous experimental evolution studies and performed longitudinal whole-population genome sequencing. Surprisingly, higher number of mutations, higher genotypic diversity, and higher evolvability were detected in the biotic conditions with smaller population size. Different adaptation strategies were observed in different environments within each species, with more diversified mutational spectrum detected in biotic conditions. The insertion sequences of Bacillus thuringiensis are critical for its adaptation to the plastic bead-attached biofilm environment, but insertion sequence mobility was a general phenomenon in this species independent of the selection condition. Additionally, certain parallel evolution has been observed across species and environments, particularly when two species adapt to the same environment at the same time. Further, our results suggest that the population size might be an important driver of evolution. Together, these results provide the first comprehensive mutational landscape of two bacterial species biofilms that is adapted to an abiotic and biotic surface. ImportanceBiofilm formation is a vital factor for the survival and adaptation of bacteria in diverse environmental niches. Experimental evolution combined with the advancement of whole-population genome sequencing provides us a powerful tool to understand the genomic dynamic of evolutionary adaptation to different environments, such as during biofilm development. Previous studies described the genetic and phenotypic changes of selected clones from experimentally evolved Bacillus thuringiensis and Bacillus subtilis that were adapted under abiotic and biotic biofilm conditions. However, the full understanding of the dynamic evolutionary landscapes was lacking. Further, the differences and similarities of adaptive mechanisms in B. thuringiensis and B. subtilis were not identified. To overcome these limitations, we performed longitudinal whole-population genome sequencing to study the underlying genetic dynamics at high resolution. Our study provides the first comprehensive mutational landscape of two bacterial species biofilms that is adapted to an abiotic and biotic surface.

microbiology↗

bioRxiv2022

Evolutionary trajectory of organelle-derived nuclear DNAs in the Triticum/Aegilops complex species

Organelle-derived nuclear DNAs, nuclear plastid DNAs (NUPTs) and nuclear mitochondrial DNAs (NUMTs), have been identified in plants. Most, if not all, genes residing in NUPTs/NUMTs (NUPGs/NUMGs) are known to be inactivated and pseudogenized. However, the role of epigenetic control in silencing NUPGs/NUMGs and the dynamic evolution of NUPTs/NUMTs with respect to organismal phylogeny remain barely explored. Based on the available nuclear and organellar genomic resources of the Triticum/Aegilops complex species, we investigated the evolutionary fates of NUPTs/NUMTs in terms of their epigenetic silencing and their dynamic occurrence rates in the nuclear diploid genomes and allopolyploid subgenomes. NUPTs and NUMTs possessed similar genomic atlas, including preferential integration to the transposable element-rich intergenic regions and generating sequence variations in the nuclear genome. The global transcriptional silencing of NUPGs/NUMGs with disrupted and intact open reading frames can be mainly attributed to their repressive chromatin states, namely high levels of DNA methylation and low levels of active histone modifications. Phylogenomic analyses suggested that the species-specific and gradual accumulation of NUPTs/NUMTs accompanied the speciation processes. Moreover, based on further pan-genomic analyses, we found significant subgenomic asymmetry in the NUPT/NUMT occurrence, which accumulated during allopolyploid wheat evolution. Our findings provide novel insights into the dynamic evolutionary fates of organelle-derived nuclear DNA in plants.

genomics↗

bioRxiv2022

Potent Immunogenicity and Broad-Spectrum Protection Potential of Microneedle Array Patch-Based COVID-19 DNA Vaccine Candidates Encoding Dimeric RBD Chimera of SARS-CoV and SARS-CoV-2 Variants

Breakthrough infections by SARS-CoV-2 variants pose a global challenge to pandemic control, and the development of more effective vaccines of broadspectrum protection is needed. In this study, we constructed pVAX1-based plasmids encoding heterodimeric receptor-binding domain (RBD) chimera of SARS-CoV and SARS-CoV-2 Omicron BA.1 (RBDSARS/BA1), SARS-CoV and SARS-CoV-2 Beta (RBDSARS/Beta), or Omicron BA.1 and Beta (RBDBA1/Beta) in secreted form. When i.m. injected in mice, RBDSARS/BA1 and RBDSARS/Beta encoding plasmids (pAD1002 and pAD131, respectively) were by far more immunogenic than RBDBA1/Beta plasmid (pAD1003). Dissolvable microneedle array patches (MAP) laden with these DNA plasmids were fabricated. All 3 resulting MAP-based vaccine candidates, namely MAP-1002, MAP1003 and MAP-131, were comparable to i.m. inoculated plasmids with electroporation assistance in eliciting strong and durable IgG responses in BALB/c and C57BL/6 mice as well as rabbits, while MAP-1002 was comparatively the most immunogenic. More importantly, MAP-1002 significantly outperformed inactivated SARS-CoV-2 virus vaccine in inducing RBD-specific IFN-{gamma}+ T cells. Moreover, MAP-1002 antisera effectively neutralized pseudoviruses displaying spike proteins of SARS-CoV, prototype SARS-CoV-2 or Beta, Delta, Omicron BA1, BA2 and BA4/5 variants. Collectively, MAP-based DNA constructs encoding chimeric RBDs of SARS-CoV and SARS-CoV-2 variants, as represented by MAP-1002, are potential COVID-19 vaccine candidates worthy further translational study.

immunology↗

bioRxiv2022

ExposomeX: Integrative Exposomic Platform Expediates Discovery of "Exposure-Biology-Disease" Nexus

Exposome has become the hotspot of next-generation health studies. To date, there is no available effective platform to standardize the analysis of exposomic data. In this study, we aim to propose one new framework of exposomic analysis and build up one novel integrated platform "ExposomeX" to expediate the discovery of the "Exposure-Biology-Disease" nexus. We have developed 13 standardized modules to accomplish six major functions including statistical learning (E-STAT), exposome database search (E-DB), mass spectrometry data processing (E-MS), meta-analysis (E-META), biological link via pathway integration and protein-protein interaction (E-BIO) and data visualization (E-VIZ). Using ExposomeX, we can effectively analyze the multiple-dimensional exposomics data and investigate the "Exposure-Biology-Disease" nexus by exploring mediation and interaction effects, understanding statistical and biological mechanisms, strengthening prediction performance, and automatically conducting meta-analysis based on well-established literature databases. The performance of ExposomeX has been well validated by re-analyzing two previous multi-omics studies. Additionally, ExposomeX can efficiently help discover new associations, as well as relevant in-depth biological pathways via protein-protein interaction and gene ontology network analysis. In sum, we have proposed a novel framework for standardized exposomic analysis, which can be accessed using both R and online interactive platform (http://www.exposomex.cn/).

bioinformatics↗

bioRxiv2022

OpenFold: Retraining AlphaFold2 yields new insights into its learning mechanisms and capacity for generalization

AlphaFold2 revolutionized structural biology with the ability to predict protein structures with exceptionally high accuracy. Its implementation, however, lacks the code and data required to train new models. These are necessary to (i) tackle new tasks, like protein-ligand complex structure prediction, (ii) investigate the process by which the model learns, which remains poorly understood, and (iii) assess the models generalization capacity to unseen regions of fold space. Here we report OpenFold, a fast, memory-efficient, and trainable implementation of AlphaFold2. We train OpenFold from scratch, fully matching the accuracy of AlphaFold2. Having established parity, we assess OpenFolds capacity to generalize across fold space by retraining it using carefully designed datasets. We find that OpenFold is remarkably robust at generalizing despite extreme reductions in training set size and diversity, including near-complete elisions of classes of secondary structure elements. By analyzing intermediate structures produced by OpenFold during training, we also gain surprising insights into the manner in which the model learns to fold proteins, discovering that spatial dimensions are learned sequentially. Taken together, our studies demonstrate the power and utility of OpenFold, which we believe will prove to be a crucial new resource for the protein modeling community.

bioinformatics↗