Separate Domains Research Articles

Research on Isomorphic Task Transfer Algorithm Based on Knowledge Distillation in Multi-Agent Collaborative Systems.

In response to the increasing number of agents and changing task scenarios in multi-agent collaborative systems, existing collaborative strategies struggle to effectively adapt to new task scenarios. To address this challenge, this paper proposes a knowledge distillation method combined with a domain separation network (DSN-KD). This method leverages the well-performing policy network from a source task as the teacher model, utilizes a domain-separated neural network structure to correct the teacher model's outputs as supervision, and guides the learning of agents in new tasks. The proposed method does not require the pre-design or training of complex state-action mappings, thereby reducing the cost of transfer. Experimental results in scenarios such as UAV surveillance and UAV cooperative target occupation, robot cooperative box pushing, UAV cooperative target strike, and multi-agent cooperative resource recovery in a particle simulation environment demonstrate that the DSN-KD transfer method effectively enhances the learning speed of new task policies and improves the proximity of the policy model to the theoretically optimal policy in practical tasks.

Antimicrobial Peptides Increase Line Tension in Raft-Forming Lipid Membranes.

The formation of phase separated membrane domains is believed to be essential for the function of the cell. The precise composition and physical properties of lipid bilayer domains play crucial roles in regulating protein activity and governing cellular processes. Perturbation of the domain structure in human cells can be related to neurodegenerative diseases and cancer. Lipid rafts are also believed to be essential in bacteria, potentially serving as targets for antibiotics. An important question is how the membrane domain structure is affected by bioactive and therapeutic molecules, such as surface-active peptides, which target cellular membranes. Here we focus on antimicrobial peptides (AMPs), crucial components of the innate immune system, to gain insights into their interaction with model lipid membranes containing domains. Using small-angle neutron/X-ray scattering (SANS/SAXS), we show that the addition of several natural AMPs (indolicidin, LL-37, magainin II, and aurein 2.2) causes substantial growth and restructuring of the domains, which corresponds to increased line tension. Contrast variation SANS and SAXS results demonstrate that the peptide inserts evenly in both phases, and the increased line tension can be related to preferential and concentration dependent thinning of the unsaturated membrane phase. We speculate that the lateral restructuring caused by the AMPs may have important consequences in affecting physiological functions of real cells. This work thus shines important light onto the complex interactions and lateral (re)organization in lipid membranes, which is relevant for a molecular understanding of diseases and the action of antibiotics.

Transfer learning with cross-domain adaptation of vibro-acoustic signals for electric motor mechanical fault detection and diagnosis.

Contact-based vibration measurement techniques involving accelerometers are typically utilized for machine health monitoring and manufacturing end-of-line quality control. Acoustic signals, encompassing sound pressure and particle velocity, carry crucial information about the operational status of underlying equipment and can be utilized for detecting various machine faults. Utilizing one-dimensional Convolution Neural Networks (1D-CNN) based on deep learning for processing raw acoustic time signals has proven to be a valuable approach in detecting machinery faults, offering an effective alternative to using traditional accelerometer signals. However, the success of deep learning methods is heavily dependent on the quantity and availability of data for robust network training. When transitioning to new non-contact type sensor technology, manufacturers may face challenges due to insufficient data for training deep learning models. In such cases, leveraging existing historical accelerometer-based data becomes crucial. Transfer learning emerges as a promising solution, allowing deep learning models trained on a source domain (SD) to be applied to a separate but related target domain (TD). This paper hence introduces a domain adaptation methodology by implementing transfer learning, where the SD 1D-CNN network is trained on accelerometer signals, then the trained model weights are transferred to the TD 1D-CNN network that can process acoustic signals.

Differential effects of the N-terminal helix of FGF8b on the activity of a small-molecule FGFR inhibitor in cell culture and for the extracellular domain of FGFR3c in solution.

SSR128129E (SSR) is a unique small-molecule inhibitor of fibroblast growth factor receptors (FGFRs). SSR is a high-affinity allosteric binder that selectively blocks one of the two major FGFR-mediated pathways. The mechanisms of SSR activity were studied previously in much detail, allowing the identification of its binding site, located in the hydrophobic groove of the receptor D3 domain. The binding site overlaps with the position of an N-terminal helix, an element exclusive for the FGF8b growth factor, which could potentially convert SSR from an allosteric inhibitor into an orthosteric blocker for the particular FGFR/FGF8b system. In this regard, we report here on the structural and functional investigation of FGF8b/FGFR3c system and the effects imposed on it by SSR. We show that SSR is equally or more potent in inhibiting FGF8b-induced FGFR signaling compared to FGF2-induced activation. On the other hand, when studied in the context of separate extracellular domains of FGFR3c in solution with NMR spectroscopy, SSR is unable to displace the N-terminal helix of FGF8b from its binding site on FGFR3c and behaves as a weak orthosteric inhibitor. The substantial inconsistency between the results obtained with cell culture and for the individual water-soluble subdomains of the FGFR proteins points to the important role played by the cell membrane.

Efficient graph representation in graph neural networks for stress predictions in stiffened panels

Machine learning (ML) and deep learning (DL) techniques have gained significant attention as reduced order models (ROMs) to computationally expensive structural analysis methods, such as finite element analysis (FEA). Graph neural network (GNN) is a particular type of neural network which processes data that can be represented as graphs. This allows for efficient representation of complex geometries that can change during the conceptual design of a structure or a product. In this study, we propose a novel graph embedding for the efficient representation of 3D stiffened panels by considering separate plate domains as vertices. This approach is considered using Graph Sampling and Aggregation (GraphSAGE) to predict stress distributions in stiffened panels with varying geometries. A comparison between a finite element-vertex graph representation is conducted to demonstrate the effectiveness of the proposed approach. A comprehensive parametric study is performed to examine the effect of structural variables on stress predictions. A wide range of geometries is considered, material nonlinearity, a few boundary conditions, together with uniform and patch loading at various positions. The study involves straight and curved panels with uni- and bi-directional stiffeners. The proposed unit-vertex representation of the panel requires only about 2% of GPU memory and about 4% of training time in comparison to a finite element-vertex embedding. The GraphSAGE model with the proposed unit-vertex representation accurately captures stress distribution across all panels, achieving an average prediction accuracy of 92.3% for the maximum von Mises stress. Our results demonstrate the immense potential of graph neural networks with the proposed graph embedding as a robust reduced-order model for 3D structures.

Dotting Out AML by Targeting Fibrillarin.

Dysregulated biomolecular condensates, formed through multivalent interactions among proteins and nucleic acids, have been recently identified to drive tumorigenesis. In acute myeloid leukemia (AML), condensates driven by RNA-binding proteins alter transcriptional networks. Yang and colleagues performed a CRISPR screen and identified fibrillarin (FBL) as a new driver in AML leukemogenesis. FBL depletion caused cell cycle arrest and death in AML cells, with minimal impact on normal cells. FBL's phase separation domains are essential for pre-rRNA processing, influencing AML cell survival by regulating ribosome biogenesis and the translation of oncogenic proteins like MYC. Therapeutically, the chemotherapeutic agent CGX-635 targets FBL, inducing its aggregation, impairing pre-rRNA processing, and reducing AML cell survival. This highlights FBL's phase separation as a therapeutic vulnerability in AML. These findings suggest that targeting the phase separation properties of RNA-binding proteins could offer a novel and effective strategy for AML treatment. Further research into condensate dynamics in cancer and development of condensate-modulating drugs holds significant promise for future cancer therapies.

Mutational analyses of the interacting domains of Schizosaccharomyces pombe Byr2 with 14-3-3s

The Byr2 kinase of fission yeast Schizosaccharomyces pombe is recruited to the membrane with the assistance of Ras1. Byr2 is also negatively regulated by 14-3-3 proteins encoded by rad24 and rad25. We conducted domain and mutational analysis of Byr2 to determine which region is critical for its binding to 14-3-3 proteins. Rad24 and Rad25 bound to both the Ras interaction domain in the N-terminus and to the C-terminal catalytic domain of Byr2. When amino acid residues S87 and T94 of the Ras-interacting domain of Byr2 were mutated to alanine, Rad24 could no longer bind to Byr2. S402, S566, S650, and S654 mutations in the C-terminal domain of Byr2 also abolished its interaction with Rad24 and Rad25. More than three mutations in the C-terminal domain were required to abolish completely its interaction with 14-3-3 protein, suggesting that multiple residues are involved in this interaction. Expression of the N-terminal domain of Byr2 in wild-type cells lowered the mating ratio, because it likely blocked the interaction of Byr2 with Ste4 and Ras1, whereas expression of the catalytic domain of Byr2 increased the mating ratio as a result of freeing from intramolecular regulation by the N-terminal domain of Byr2. The S87A and T94A mutations of Byr2 increased the mating ratio and attenuated inhibition of Byr2 by Rad24; therefore, these two amino acids are critical for its regulation by Rad24. S566 of Byr2 is critical for activity of Byr2 but not for its interaction with 14-3-3 proteins. In this study, we show that 14-3-3 proteins interact with two separate domains in Byr2 as negative regulators.

Social support among gender diverse people: Are we measuring what we think we are?

Background To assess social support among gender diverse people, measures normed and validated with cisgender participants are used, most commonly the Multidimensional Scale of Social Support (MSPSS). Despite widespread use, the psychometric properties of the MSPSS have not been systematically investigated among gender diverse people. Thus, it is unclear whether use of the MSPSS is appropriate for this population. Method Using a mixed-method approach (e.g. deductive thematic analysis and confirmatory factor analyses), we explored the factor structure of the MSPSS among gender diverse participants (N = 428). The study was conducted in Canada between 2019 and 2020. Results After running confirmatory factor analyses on MSPSS items, the factor structure differed from prior studies of cisgender participants. First, support from a significant other did not correlate with support from family or friends, suggesting its inclusion in a global measure of social support may distort findings when all domains are combined. Second although social support from family and friends emerged as two separate domains, qualitative data suggested that the boundaries between these forms of support were unclear. A deductive thematic analysis further highlighted properties of the scale that may not align with gender diverse peoples’ experiences. Conclusions Overall, these findings raise questions about the interpretation of the MSPSS when used with gender diverse people and suggest that an adapted or newly developed measure is needed.

Mass Spectrometry-Based Protein Footprinting Defines the Binding Pocket of Crotonylated H3K14 in the PHD1 Domain of BAF45D within the BAF Chromatin Remodeling Complex.

The BRG-/BRM-associated factor (BAF) chromatin remodeling complex is a central actor in transcription. One mechanism by which BAF affects gene expression is via its various histone mark readers, including double plant homeodomains (DPF), located in the BAF45D subunit. DPF domains recognize lysine acetyl and acylations, including crotonylation, localized at promoters and enhancers. Despite a significant degree of conservation between DPF domains, attempts to crystallize BAF45D with a crotonylated histone 3 peptide (H3K14Cr) were unsuccessful. In addition, recent cryoEM and modeled structures failed to define the Req domain of BAF45D, which is responsible for reading lysine modifications. Thus, the precise mechanism of crotonyl group recognition and binding by BAF45D within the BAF complex remains unclear. We turned to protein footprinting mass spectrometry to map the binding interface between H3K14Cr and BAF45D. This technique is able to demarcate protein-binding interfaces by modifying surface-accessible residues and is not limited by protein size or composition. Experiments performed in the isolated DPF domain of BAF45D (BAF45DDPF)-delineated H3K14Cr peptide binding across the PHD1 and PHD2 pockets. We observed markedly similar effects on the BAF45D subunit when assessing H3K14Cr binding in the purified full BAF complex. The ATPase motor, BRM, also displayed H3K14Cr-protected peptides in two separate domains that were subsequently evaluated in direct binding assays. These data confirm the BAF45D-crotonylamide interaction within its obligate complex and are the first to demonstrate H3K14Cr direct binding to BRM.

Ligation of multiple protein domains using orthogonal inteins with non-native splice junctions.

Protein splicing is a self-catalyzed process in which an internal protein domain (the intein) is excised from its flanking sequences, linking them together with a canonical peptide bond. Trans-inteins are separated in two different precursor polypeptide chains that must assemble to catalytically self-excise and ligate the corresponding flanking exteins to join even when expressed separately either in vitro or in vivo. They are very interesting to construct full proteins from separate domains because their common small size favors chemical synthesis approaches. Therefore, trans-inteins have multiple applications such as protein modification and purification, structural characterization of protein domains or production of intein-based biosensors, among others. For many of these applications, when using more than one trans-intein, orthogonality between them is a critical issue to ensure the proper ligation of the exteins. Here, we confirm the orthogonality (lack of cross-reactivity) of four different trans- or split inteins, gp41-1, gp41-8, IMPDH-1 and NrdJ-1 both in vivo and in vitro, and built different constructs that allow for the sequential fusion of up to four protein fragments into one final spliced product. We have characterized the splicing efficiency of these constructs. All harbor non-native extein residues at the splice junction between the trans-intein and the neighboring exteins, except for the essential Ser + 1. Our results show that it is possible to ligate four different protein domains using inteins gp41-1, IMPDH-1 and NrdJ-1 with non-native extein residues to obtain a final four-domain spliced product with a not negligible yield that keeps its native sequence.

Promising properties of cytochrome P450 BM3 reconstituted from separate domains by split intein

Recombinant cytochrome P450 monooxygenases possess significant potential as biocatalysts, and efforts to improve heme content, electron coupling efficiency, and catalytic activity and stability are ongoing. Domain swapping between heme and reductase domains, whether natural or engineered, has thus received increasing attention. Here, we successfully achieved split intein-mediated reconstitution (IMR) of the heme and reductase domains of P450 BM3 both in vitro and in vivo. Intriguingly, the reconstituted enzymes displayed promising properties for practical use. IMR BM3 exhibited a higher heme content (>50 %) and a greater tendency for oligomerization compared to the wild-type enzyme. Moreover, these reconstituted enzymes exhibited a distinct increase in activity ranging from 165 % to 430 % even under the same heme concentrations. The reproducibility of our results strongly suggests that the proposed reconstitution approach could pave a new path for enhancing the catalytic efficiency of related enzymes.

Intestinal lysozyme engagement of Salmonella Typhimurium stimulates the release of barrier-impairing InvE and Lpp1

Lysozyme is a β-1,4-glycosidase that hydrolyzes the polysaccharide backbone of bacterial cell walls. With an additional bactericidal function mediated by a separate protein domain, lysozyme is considered a uniquely important antimicrobial molecule contributing to the host's innate immune response to infection. Elevated lysozyme production is found in various inflammatory conditions while patients with genetic risks for inflammatory bowel diseases demonstrate abnormal lysozyme expression, granule packaging, and secretion in Paneth cells. However, it remains unclear how a gain- or loss-of-function in host lysozyme may impact the host inflammatory responses to pathogenic infection. We challenged Lyz1-/- and ectopic Lyz1-expressing (Villin-Lyz1TG) mice with S. Typhimurium and then comprehensively assessed the inflammatory disease progression. We conducted proteomics analysis to identify molecules derived from human lysozyme-mediated processing of live Salmonella. We examined the barrier-impairing effects of these identified molecules in human intestinal epithelial cell monolayer and enteroids. Lyz1-/- mice are protected from infection in terms of morbidity, mortality, and barrier integrity, whereas Villin-Lyz1TG mice demonstrate exacerbated infection and inflammation. The growth and invasion of Salmonella invitro are not affected by human or chicken lysozyme, whereas lysozyme encountering of live Salmonella stimulates the release of barrier-disrupting factors, InvE-sipC and Lpp1, which directly or indirectly impair the tight junctions. The direct engagement of host intestinal lysozyme with an enteric pathogen such as Salmonella promotes the release of virulence factors that are barrier-impairing and pro-inflammatory. Controlling lysozyme function may help alleviate the inflammatory progression.

ZLR: a fast online authenticated encryption scheme achieving full security

AbstractOnline authenticated encryption has been considered of practical relevance in light-weight environments due to low latency and constant memory usage. In this paper, we propose a new tweakable block cipher-based online authenticated encryption scheme, dubbed , and its domain separation variant, dubbed . and follow the Encrypt-Mix-Encrypt paradigm. However, in contrast to existing schemes using the same paradigm such as and , and enjoy n-bit security by using larger internal states with an efficient -like hashing algorithm. In this way, 2n-bit blocks are processed with only a single primitive call for hashing and two primitive calls for encryption and decryption, when they are based on an n-bit tweakable block cipher using n-bit (resp. 2n-bit) tweaks for (resp. ). Furthermore, they support pipelined computation as well as online nonce-misuse resistance. To the best of our knowledge, and are the first pipelineable tweakable block cipher-based online authenticated encryption schemes of rate-2/3 that provide n-bit security with online nonce-misuse resistance.

Biochemical properties of chromatin domains define genome compartmentalization.

Chromatin three-dimensional (3D) organization inside the cell nucleus determines the separation of euchromatin and heterochromatin domains. Their segregation results in the definition of active and inactive chromatin compartments, whereby the local concentration of associated proteins, RNA and DNA results in the formation of distinct subnuclear structures. Thus, chromatin domains spatially confined in a specific 3D nuclear compartment are expected to share similar epigenetic features and biochemical properties, in terms of accessibility and solubility. Based on this rationale, we developed the 4f-SAMMY-seq to map euchromatin and heterochromatin based on their accessibility and solubility, starting from as little as 10000 cells. Adopting a tailored bioinformatic data analysis approach we reconstruct also their 3D segregation in active and inactive chromatin compartments and sub-compartments, thus recapitulating the characteristic properties of distinct chromatin states. A key novelty of the new method is the capability to map both the linear segmentation of open and closed chromatin domains, as well as their compartmentalization in one single experiment.

PHF2 regulates genome topology and DNA replication in neural stem cells via cohesin.

Cohesin plays a crucial role in the organization of topologically-associated domains (TADs), which influence gene expression and DNA replication timing. Whether epigenetic regulators may affect TADs via cohesin to mediate DNA replication remains elusive. Here, we discover that the histone demethylase PHF2 associates with RAD21, a core subunit of cohesin, to regulate DNA replication in mouse neural stem cells (NSC). PHF2 loss impairs DNA replication due to the activation of dormant replication origins in NSC. Notably, the PHF2/RAD21 co-bound genomic regions are characterized by CTCF enrichment and epigenomic features that resemble efficient, active replication origins, and can act as boundaries to separate adjacent domains. Accordingly, PHF2 loss weakens TADs and chromatin loops at the co-bound loci due to reduced RAD21 occupancy. The observed topological and DNA replication defects in PHF2 KO NSC support a cohesin-dependent mechanism. Furthermore, we demonstrate that the PHF2/RAD21 complex exerts little effect on gene regulation, and that PHF2's histone-demethylase activity is dispensable for normal DNA replication and proliferation of NSC. We propose that PHF2 may serve as a topological accessory to cohesin for cohesin localization to TADs and chromatin loops, where cohesin represses dormant replication origins directly or indirectly, to sustain DNA replication in NSC.

Compressible Flow Through Flat Valve Seals at Microscale Displacements

Abstract Existing analytical flow models for predicting flow rates at microscale seal displacements are limited to two separate domains. The first assumes a small channel length to height aspect ratio at relatively large seal displacements. The second assumes a large channel length to height aspect ratio at relatively small seal displacements. A piecewise analytical model for compressible flow is developed here to enable predicting flow rates in valves with fluid pathways of any aspect ratio. The new model is validated by numerical studies and experiment. The results are applicable to flat valve seals having a cylindrical seal boss feature with fluid passage length to height aspect ratios ranging from 3.3 to 800. The new model is particularly useful for the design of microvalves and macroscale valves with small actuator displacements.

Facial Action Unit detection based on multi-task learning strategy for unlabeled facial images in the wild

Facial Action Unit (AU) detection often relies on highly-cost accurate labeling or inaccurate pseudo labeling techniques in recent years. How to introduce large amounts of unlabeled facial images in the wild into supervised AU detection frameworks has become a challenging problem. Additionally, nearly every type of AUs has the problem of unbalanced positive and negative samples. Inspired by other multi-task learning frameworks, we first propose a multi-task learning strategy boosting AU detection in the wild through jointing facial landmark detection and AU domain separation and reconstruction. Our introduced dual domains facial landmark detection framework can solve the lack of accurate facial landmark coordinates during the AU domain separation and reconstruction training process, while the parameters of homostructural facial extraction modules from these two similar facial tasks are shared. Moreover, we propose a pixel-level feature alignment scheme to maintain the consistency of features obtained from two separation and reconstruction processes. Furthermore, a weighted asymmetric loss is proposed to change the contribution of positive and negative samples of each type of AUs to model parameters updating. Experimental results on three widely used benchmarks demonstrate our superiority to most state-of-the-art methods for AU detection.

‘Mr Rules’: Keir Starmer and the juridification of politics

AbstractKeir Starmer’s moniker of ‘Mr Rules’ captures his deep investment in a rules-based form of politics that seeks to uphold established standards of probity and competency in public office. Rather than a mere tactic of opposition politics, we argue that it is symptomatic of the juridification of politics. By this we mean the ceding of the terrain of politics to the seemingly superior and separate domains of law and administration. Drawing upon and extending existing analyses of depoliticisation and unpolitics, the juridification of politics marks the abandonment of consciously values-based politics in favour of a reliance upon legal and quasi-legal (i.e. rules, norms, conventions, procedures) means to address substantive matters of public policy. Crucially, we locate this trend as a consequence of the neoliberal way of politics in which the task of governing in a post-ideological age is reduced to administration. This is significant, we conclude, because such an approach is incapable of responding to the intersecting crises confronting national and international politics.

Distributed Heterogeneous Transfer Learning

Transfer learning has proved to be effective for building predictive models even in complex conditions with a low amount of available labeled data, by constructing a predictive model for a target domain also using the knowledge coming from a separate domain, called source domain. However, several existing transfer learning methods assume identical feature spaces between the source and the target domains. This assumption limits the possible real-world applications of such methods, since two separate, although related, domains could be described by totally different feature spaces. Heterogeneous transfer learning methods aim to overcome this limitation, but they usually i) make other assumptions on the features, such as requiring the same number of features, ii) are not generally able to distribute the workload over multiple computational nodes, iii) cannot work in the Positive-Unlabeled (PU) learning setting, which we also considered in this study, or iv) their applicability is limited to specific application domains, i.e., they are not general-purpose methods.In this manuscript, we present a novel distributed heterogeneous transfer learning method, implemented in Apache Spark, that overcomes all the above-mentioned limitations. Specifically, it is able to work also in the PU learning setting by resorting to a clustering-based approach, and can align totally heterogeneous feature spaces, without exploiting peculiarities of specific application domains. Moreover, our distributed approach allows us to process large source and target datasets.Our experimental evaluation was performed in three different application domains that can benefit from transfer learning approaches, namely the reconstruction of the human gene regulatory network, the prediction of cerebral stroke in hospital patients, and the prediction of customer energy consumption in power grids. The results show that the proposed approach is able to outperform 4 state-of-the-art heterogeneous transfer learning approaches and 3 baselines, and exhibits ideal performances in terms of scalability.

Phase separation-competent FBL promotes early pre-rRNA processing and translation in acute myeloid leukaemia.

RNA-binding proteins (RBPs) are pivotal in acute myeloid leukaemia (AML), a lethal disease. Although specific phase separation-competent RBPs are recognized in AML, the effect of their condensate formation on AML leukaemogenesis, and the therapeutic potential of inhibition of phase separation are underexplored. In our in vivo CRISPR RBP screen, fibrillarin (FBL) emerges as a crucial nucleolar protein that regulates AML cell survival, primarily through its phase separation domains rather than methyltransferase or acetylation domains. These phase separation domains, with specific features, coordinately drive nucleoli formation and early processing of pre-rRNA (including efflux, cleavage and methylation), eventually enhancing the translation of oncogenes such as MYC. Targeting the phase separation capability of FBL with CGX-635 leads to elimination of AML cells, suggesting an additional mechanism of action for CGX-635 that complements its established therapeutic effects. We highlight the potential of PS modulation of critical proteins as a possible therapeutic strategy for AML.