Histone acetyltransferase Kat2a regulates ferroptosis via enhancing Tfrc and Hmox1 expression in diabetic cardiomyopathy.

Diabetic cardiomyopathy (DCM) is a prevalent myocardial microvascular complication of the myocardium with a complex pathogenesis. Investigating the pathogenesis of DCM can significantly contribute to enhancing its prevention and treatment strategies. Our study revealed an upregulation of lysine acetyltransferase 2 A (Kat2a) expression in DCM, accompanied by a decrease in N6-methyladenosine (m6A) modified Kat2a mRNA levels. Our study revealed an upregulation of lysine acetyltransferase 2 A (Kat2a) expression in DCM, accompanied by a decrease in N6-methyladenosine (m6A) modified Kat2a mRNA levels. Functionally, inhibition of Kat2a effectively ameliorated high glucose-induced cardiomyocyte injury both in vitro and in vivo by suppressing ferroptosis. Mechanistically, Demethylase alkB homolog 5 (Alkbh5) was found to reduce m6A methylation levels on Kat2a mRNA, leading to its upregulation. YTH domain family 2 (Ythdf2) played a crucial role as an m6A reader protein mediating the degradation of Kat2a mRNA. Furthermore, Kat2a promoted ferroptosis by increasing Tfrc and Hmox1 expression via enhancing the enrichment of H3K27ac and H3K9ac on their promoter regions. In conclusion, our findings unveil a novel role for the Kat2a-ferroptosis axis in DCM pathogenesis, providing valuable insights for potential clinical interventions.

Single-cell RNA sequencing data imputation using bi-level feature propagation.

Single-cell RNA sequencing (scRNA-seq) enables the exploration of cellular heterogeneity by analyzing gene expression profiles in complex tissues. However, scRNA-seq data often suffer from technical noise, dropout events and sparsity, hindering downstream analyses. Although existing works attempt to mitigate these issues by utilizing graph structures for data denoising, they involve the risk of propagating noise and fall short of fully leveraging the inherent data relationships, relying mainly on one of cell-cell or gene-gene associations and graphs constructed by initial noisy data. To this end, this study presents single-cell bilevel feature propagation (scBFP), two-step graph-based feature propagation method. It initially imputes zero values using non-zero values, ensuring that the imputation process does not affect the non-zero values due to dropout. Subsequently, it denoises the entire dataset by leveraging gene-gene and cell-cell relationships in the respective steps. Extensive experimental results on scRNA-seq data demonstrate the effectiveness of scBFP in various downstream tasks, uncovering valuable biological insights.

Two sequence- and two structure-based ML models have learned different aspects of protein biochemistry.

Deep learning models are seeing increased use as methods to predict mutational effects or allowed mutations in proteins. The models commonly used for these purposes include large language models (LLMs) and 3D Convolutional Neural Networks (CNNs). These two model types have very different architectures and are commonly trained on different representations of proteins. LLMs make use of the transformer architecture and are trained purely on protein sequences whereas 3D CNNs are trained on voxelized representations of local protein structure. While comparable overall prediction accuracies have been reported for both types of models, it is not known to what extent these models make comparable specific predictions and/or generalize protein biochemistry in similar ways. Here, we perform a systematic comparison of two LLMs and two structure-based models (CNNs) and show that the different model types have distinct strengths and weaknesses. The overall prediction accuracies are largely uncorrelated between the sequence- and structure-based models. Overall, the two structure-based models are better at predicting buried aliphatic and hydrophobic residues whereas the two LLMs are better at predicting solvent-exposed polar and charged amino acids. Finally, we find that a combined model that takes the individual model predictions as input can leverage these individual model strengths and results in significantly improved overall prediction accuracy.

Attend Who is Weak: Pruning-assisted Medical Image Localization under Sophisticated and Implicit Imbalances.

Deep neural networks (DNNs) have rapidly become a de facto choice for medical image understanding tasks. However, DNNs are notoriously fragile to the class imbalance in image classification. We further point out that such imbalance fragility can be amplified when it comes to more sophisticated tasks such as pathology localization, as imbalances in such problems can have highly complex and often implicit forms of presence. For example, different pathology can have different sizes or colors (w.r.t.the background), different underlying demographic distributions, and in general different difficulty levels to recognize, even in a meticulously curated balanced distribution of training data. In this paper, we propose to use pruning to automatically and adaptively identify hard-to-learn (HTL) training samples, and improve pathology localization by attending them explicitly, during training in supervised, semi-supervised, and weakly-supervised settings. Our main inspiration is drawn from the recent finding that deep classification models have difficult-to-memorize samples and those may be effectively exposed through network pruning [15] - and we extend such observation beyond classification for the first time. We also present an interesting demographic analysis which illustrates HTLs ability to capture complex demographic imbalances. Our extensive experiments on the Skin Lesion Localization task in multiple training settings by paying additional attention to HTLs show significant improvement of localization performance by ~2-3%.

Two sequence- and two structure-based ML models have learned different aspects of protein biochemistry.

Deep learning models are seeing increased use as methods to predict mutational effects or allowed mutations in proteins. The models commonly used for these purposes include large language models (LLMs) and 3D Convolutional Neural Networks (CNNs). These two model types have very different architectures and are commonly trained on different representations of proteins. LLMs make use of the transformer architecture and are trained purely on protein sequences whereas 3D CNNs are trained on voxelized representations of local protein structure. While comparable overall prediction accuracies have been reported for both types of models, it is not known to what extent these models make comparable specific predictions and/or generalize protein biochemistry in similar ways. Here, we perform a systematic comparison of two LLMs and two structure-based models (CNNs) and show that the different model types have distinct strengths and weaknesses. The overall prediction accuracies are largely uncorrelated between the sequence- and structure-based models. Overall, the two structure-based models are better at predicting buried aliphatic and hydrophobic residues whereas the two LLMs are better at predicting solvent-exposed polar and charged amino acids. Finally, we find that a combined model that takes the individual model predictions as input can leverage these individual model strengths and results in significantly improved overall prediction accuracy.

Moral transgression modulates fairness considerations in the ultimatum game: Evidence from ERP and EEG data.

People tend to dislike and punish unfair behaviors in social interactions, and this disposition may be moderated by the characteristics of their interaction partner. We used a modified ultimatum game (UG) to investigate players' responses to fair and unfair offers from proposers described as having performed either a moral transgression or a neutral behavior, and recorded an electroencephalogram. The participants' behavior in the UG suggests that people quickly demand more fairness from proposers who have committed moral transgressions rather than neutral behavior. Event-related potentials (ERPs) revealed a significant effect of offer type and of proposer type on P300 activity. The prestimulus α-oscillation power in the neutral behavior condition was significantly lower than that in the moral transgression condition. The post-stimulus ß-event-related synchronization (ß-ERS) was larger for the moral transgression condition than the neutral behavior condition in response to the least fair offers, and larger for neutral behavior than the moral transgression condition in response to the fairest offers. In summary, ß-ERS was influenced by both proposer type and offer type, which revealed different neural responses to the offer from either a morally transgressive or a neutral behavior proposer.

Bag of Tricks for Training Deeper Graph Neural Networks: A Comprehensive Benchmark Study.

Training deep graph neural networks (GNNs) is notoriously hard. Besides the standard plights in training deep architectures such as vanishing gradients and overfitting, it also uniquely suffers from over-smoothing, information squashing, and so on, which limits their potential power for encoding the high-order neighbor structure in large-scale graphs. Although numerous efforts are proposed to address these limitations, such as various forms of skip connections, graph normalization, and random dropping, it is difficult to disentangle the advantages brought by a deep GNN architecture from those "tricks" necessary to train such an architecture. Moreover, the lack of a standardized benchmark with fair and consistent experimental settings poses an almost insurmountable obstacle to gauge the effectiveness of new mechanisms. In view of those, we present the first fair and reproducible benchmark dedicated to assessing the "tricks" of training deep GNNs. We categorize existing approaches, investigate their hyperparameter sensitivity, and unify the basic configuration. Comprehensive evaluations are then conducted on tens of representative graph datasets including the recent large-scale Open Graph Benchmark, with diverse deep GNN backbones. We demonstrate that an organic combo of initial connection, identity mapping, group and batch normalization attains the new state-of-the-art results for deep GNNs on large datasets. Codes are available: https://github.com/VITA-Group/Deep_GCN_Benchmarking.

Moral transgression modulates empathy for pain: Evidence from ERP and EEG data.

Empathy for others' pain plays a critical role in human social interactions; however, the influence of moral transgression remains unclear. We examined the effect of moral transgression on the behavioral and underlying neural processes of empathy for others' pain. Participants performed a pain-empathy task separately in a moral transgression condition and a neutral behavior condition, while an electroencephalogram was recorded. Event-related potential (ERP) results showed that empathic response, as reflected in the late positive component, was smaller when participants performed the task in the moral transgression condition than in the neutral behavior condition. Time-frequency results also showed decreased empathic effect on the beta event-related desynchronization response in the moral transgression as compared to the neutral behavior condition. However, empathic response as reflected in the N2 component was comparable between the moral conditions. These findings demonstrate a moral transgression effect on both cognitive evaluations and sensorimotor processes of empathy for others' pain. Furthermore, spontaneous alpha-oscillation power recorded prior to the onset of empathy-inducing stimuli was significantly higher in the moral transgression condition than in the neutral behavior condition. Consequently, differences in sustained attention may be the physiological foundation of the impact of moral transgression of the observed person on the cognitive and sensorimotor processes of empathy for pain.

Bringing Your Own View: Graph Contrastive Learning without Prefabricated Data Augmentations.

Self-supervision is recently surging at its new frontier of graph learning. It facilitates graph representations beneficial to downstream tasks; but its success could hinge on domain knowledge for handcraft or the often expensive trials and errors. Even its state-of-the-art representative, graph contrastive learning (GraphCL), is not completely free of those needs as GraphCL uses a prefabricated prior reflected by the ad-hoc manual selection of graph data augmentations. Our work aims at advancing GraphCL by answering the following questions: How to represent the space of graph augmented views? What principle can be relied upon to learn a prior in that space? And what framework can be constructed to learn the prior in tandem with contrastive learning? Accordingly, we have extended the prefabricated discrete prior in the augmentation set, to a learnable continuous prior in the parameter space of graph generators, assuming that graph priors per se, similar to the concept of image manifolds, can be learned by data generation. Furthermore, to form contrastive views without collapsing to trivial solutions due to the prior learnability, we have leveraged both principles of information minimization (InfoMin) and information bottleneck (InfoBN) to regularize the learned priors. Eventually, contrastive learning, InfoMin, and InfoBN are incorporated organically into one framework of bi-level optimization. Our principled and automated approach has proven to be competitive against the state-of-the-art graph self-supervision methods, including GraphCL, on benchmarks of small graphs; and shown even better generalizability on large-scale graphs, without resorting to human expertise or downstream validation. Our code is publicly released at https://github.com/Shen-Lab/GraphCL_Automated.

Investigation on the influence of AC electric filed and KCl on the structure and gel properties of Konjac glucomannan.

The influence of alternating current (AC) electric field and KCl on the structure and gel properties of Konjac Glucomannan (KGM) were studied in this work by high-performance gel permeation chromatography (HPGPC), acid-base titration, solid-state nuclear magnetic resonance (NMR), X-ray diffraction (XRD), simultaneous differential scanning calorimetry/thermo gravimetric analyzer (DSC/TGA) and a rheometer. HPGPC showed KGM was degraded by AC electric field and Acid-base titration showed that under the action of AC electric field and KCl KGM removed part of acetyl groups, which were consistent with the analysis of NMR. XRD and temperature sweep measurements respectively showed that the electrotreatment time and KCl concentration had important effects on the gel formation and its three-dimensional network. Simultaneous DSC/TGA and temperature sweep measurements both demonstrated the gel had good thermal stability.

Augmentations in Hypergraph Contrastive Learning: Fabricated and Generative.

This paper targets at improving the generalizability of hypergraph neural networks in the low-label regime, through applying the contrastive learning approach from images/graphs (we refer to it as HyperGCL). We focus on the following question: How to construct contrastive views for hypergraphs via augmentations? We provide the solutions in two folds. First, guided by domain knowledge, we fabricate two schemes to augment hyperedges with higher-order relations encoded, and adopt three vertex augmentation strategies from graph-structured data. Second, in search of more effective views in a data-driven manner, we for the first time propose a hypergraph generative model to generate augmented views, and then an end-to-end differentiable pipeline to jointly learn hypergraph augmentations and model parameters. Our technical innovations are reflected in designing both fabricated and generative augmentations of hypergraphs. The experimental findings include: (i) Among fabricated augmentations in HyperGCL, augmenting hyperedges provides the most numerical gains, implying that higher-order information in structures is usually more downstream-relevant; (ii) Generative augmentations do better in preserving higher-order information to further benefit generalizability; (iii) HyperGCL also boosts robustness and fairness in hypergraph representation learning. Codes are released at https://github.com/weitianxin/HyperGCL.

Cooperate or aggress? An opponent's tendency to cooperate modulates the neural dynamics of interpersonal cooperation.

Humans are social animals and need to cooperate to survive. However, individuals are not cooperative in every social interaction, and their cooperation may depend on social context. The present study used a social dilemma game to investigate whether an opponent's tendency to be cooperative over time influenced a player's behavior and neural response to outcomes in the game. University students ("players") thought they were playing against other students ("opponents") in the Chicken Game but were actually playing against a programmed computer. Participants were randomly assigned to play with an opponent who tended to be competitive (cooperative 20% of the time) or who tended to be cooperative (cooperative 80% of the time). The results showed that early in the game, participants in both groups adopted a "tit-for-tat" strategy. However, as the game progressed and the opponent's behavioral tendency became more noticeable, players in the competitive-opponent group became generally more cooperative to limit their losses. ERPs analyses indicated that players had a higher P300 and larger theta power in response to the opponent's aggression but not to the opponent's cooperation when their opponent showed a tendency to be cooperative vs. competitive. The results suggest that people adjust their cooperative behavior based on their opponent's behavior in social interaction, and aggression captures more attention than cooperation in this process.

Ventricular fibrillation storm after revascularization of chronic total occlusion of the left anterior descending artery: is this reperfusion arrhythmia?

Electrical storm is a life-threatening emergency condition defined as three or more episodes of ventricular tachycardia or ventricular fibrillation (VF) within 24 hours requiring anti-tachycardia therapy, electrical cardioversion, or defibrillation. However, studies of the incidence of electrical storm after chronic total occlusion-percutaneous coronary intervention (CTO-PCI) are limited,7 and post-procedural VF after revascularization of CTO has not been described. The purpose of this article was to present a case of post-operative VF electrical storm after revascularization of CTO of the left anterior descending (LAD) artery to determine whether the electrical storm was caused by reperfusion arrhythmia or compromise of either branch vessels or the collateral circulation during intervention.

When Does Self-Supervision Help Graph Convolutional Networks?

Self-supervision as an emerging technique has been employed to train convolutional neural networks (CNNs) for more transferrable, generalizable, and robust representation learning of images. Its introduction to graph convolutional networks (GCNs) operating on graph data is however rarely explored. In this study, we report the first systematic exploration and assessment of incorporating self-supervision into GCNs. We first elaborate three mechanisms to incorporate self-supervision into GCNs, analyze the limitations of pretraining & finetuning and self-training, and proceed to focus on multi-task learning. Moreover, we propose to investigate three novel self-supervised learning tasks for GCNs with theoretical rationales and numerical comparisons. Lastly, we further integrate multi-task self-supervision into graph adversarial training. Our results show that, with properly designed task forms and incorporation mechanisms, self-supervision benefits GCNs in gaining more generalizability and robustness. Our codes are available at https://github.com/Shen-Lab/SS-GCNs.

How High-Risk Comorbidities Co-Occur in Readmitted Patients With Hip Fracture: Big Data Visual Analytical Approach.

BACKGROUND: When older adult patients with hip fracture (HFx) have unplanned hospital readmissions within 30 days of discharge, it doubles their 1-year mortality, resulting in substantial personal and financial burdens. Although such unplanned readmissions are predominantly caused by reasons not related to HFx surgery, few studies have focused on how pre-existing high-risk comorbidities co-occur within and across subgroups of patients with HFx. OBJECTIVE: This study aims to use a combination of supervised and unsupervised visual analytical methods to (1) obtain an integrated understanding of comorbidity risk, comorbidity co-occurrence, and patient subgroups, and (2) enable a team of clinical and methodological stakeholders to infer the processes that precipitate unplanned hospital readmission, with the goal of designing targeted interventions. METHODS: We extracted a training data set consisting of 16,886 patients (8443 readmitted patients with HFx and 8443 matched controls) and a replication data set consisting of 16,222 patients (8111 readmitted patients with HFx and 8111 matched controls) from the 2010 and 2009 Medicare database, respectively. The analyses consisted of a supervised combinatorial analysis to identify and replicate combinations of comorbidities that conferred significant risk for readmission, an unsupervised bipartite network analysis to identify and replicate how high-risk comorbidity combinations co-occur across readmitted patients with HFx, and an integrated visualization and analysis of comorbidity risk, comorbidity co-occurrence, and patient subgroups to enable clinician stakeholders to infer the processes that precipitate readmission in patient subgroups and to propose targeted interventions. RESULTS: The analyses helped to identify (1) 11 comorbidity combinations that conferred significantly higher risk (ranging from P<.001 to P=.01) for a 30-day readmission, (2) 7 biclusters of patients and comorbidities with a significant bicluster modularity (P<.001; Medicare=0.440; random mean 0.383 [0.002]), indicating strong heterogeneity in the comorbidity profiles of readmitted patients, and (3) inter- and intracluster risk associations, which enabled clinician stakeholders to infer the processes involved in the exacerbation of specific combinations of comorbidities leading to readmission in patient subgroups. CONCLUSIONS: The integrated analysis of risk, co-occurrence, and patient subgroups enabled the inference of processes that precipitate readmission, leading to a comorbidity exacerbation risk model for readmission after HFx. These results have direct implications for (1) the management of comorbidities targeted at high-risk subgroups of patients with the goal of pre-emptively reducing their risk of readmission and (2) the development of more accurate risk prediction models that incorporate information about patient subgroups.

Efficient identification of multiple pathways: RNA-Seq analysis of livers from 56Fe ion irradiated mice.

BACKGROUND: mRNA interaction with other mRNAs and other signaling molecules determine different biological pathways and functions. Gene co-expression network analysis methods have been widely used to identify correlation patterns between genes in various biological contexts (e.g., cancer, mouse genetics, yeast genetics). A challenge remains to identify an optimal partition of the networks where the individual modules (clusters) are neither too small to make any general inferences, nor too large to be biologically interpretable. Clustering thresholds for identification of modules are not systematically determined and depend on user-settable parameters requiring optimization. The absence of systematic threshold determination may result in suboptimal module identification and a large number of unassigned features. RESULTS: In this study, we propose a new pipeline to perform gene co-expression network analysis. The proposed pipeline employs WGCNA, a software widely used to perform different aspects of gene co-expression network analysis, and Modularity Maximization algorithm, to analyze novel RNA-Seq data to understand the effects of low-dose 56Fe ion irradiation on the formation of hepatocellular carcinoma in mice. The network results, along with experimental validation, show that using WGCNA combined with Modularity Maximization, provides a more biologically interpretable network in our dataset, than that obtainable using WGCNA alone. The proposed pipeline showed better performance than the existing clustering algorithm in WGCNA, and identified a module that was biologically validated by a mitochondrial complex I assay. CONCLUSIONS: We present a pipeline that can reduce the problem of parameter selection that occurs with the existing algorithm in WGCNA, for applicable RNA-Seq datasets. This may assist in the future discovery of novel mRNA interactions, and elucidation of their potential downstream molecular effects.

Learning to Optimize in Swarms.

Learning to optimize has emerged as a powerful framework for various optimization and machine learning tasks. Current such "meta-optimizers" often learn in the space of continuous optimization algorithms that are point-based and uncertainty-unaware. To overcome the limitations, we propose a meta-optimizer that learns in the algorithmic space of both point-based and population-based optimization algorithms. The meta-optimizer targets at a meta-loss function consisting of both cumulative regret and entropy. Specifically, we learn and interpret the update formula through a population of LSTMs embedded with sample- and feature-level attentions. Meanwhile, we estimate the posterior directly over the global optimum and use an uncertainty measure to help guide the learning process. Empirical results over non-convex test functions and the protein-docking application demonstrate that this new meta-optimizer outperforms existing competitors. The codes are publicly available at: https://github.com/Shen-Lab/LOIS.

Enabling Comprehension of Patient Subgroups and Characteristics in Large Bipartite Networks: Implications for Precision Medicine.

A primary goal of precision medicine is to identify patient subgroups based on their characteristics (e.g., comorbidities or genes) with the goal of designing more targeted interventions. While network visualization methods such as Fruchterman-Reingold have been used to successfully identify such patient subgroups in small to medium sized data sets, they often fail to reveal comprehensible visual patterns in large and dense networks despite having significant clustering. We therefore developed an algorithm called ExplodeLayout, which exploits the existence of significant clusters in bipartite networks to automatically "explode" a traditional network layout with the goal of separating overlapping clusters, while at the same time preserving key network topological properties that are critical for the comprehension of patient subgroups. We demonstrate the utility of ExplodeLayout by visualizing a large dataset extracted from Medicare consisting of readmitted hip-fracture patients and their comorbidities, demonstrate its statistically significant improvement over a traditional layout algorithm, and discuss how the resulting network visualization enabled clinicians to infer mechanisms precipitating hospital readmission in specific patient subgroups.

Characterization of Salinity Tolerance of Transgenic Rice Lines Harboring HsCBL8 of Wild Barley (Hordeum spontanum) Line from Qinghai-Tibet Plateau.

Rice is more sensitive to salinity, particularly at its early vegetative and later productive stages. Wild plants growing in harsh environments such as wild barley from Qinghai-Tibet Plateau adapt to the adverse environment with allelic variations at the loci responsible for stressful environment, which could be used for rice genetic improvement. In this study, we overexpressed HsCBL8 encoding a calcium-sensor calcineurin B-like (CBL) protein in rice. The gene was isolated from XZ166, a wild-barley (Hordeum spontanum) line originated from Qinghai-Tibet Plateau. We found that XZ166 responded to high NaCl concentration (200 mM) with more HsCBL8 transcripts than CM72, a cultivated barley line known for salinity tolerance. XZ166 is significantly different from CM72 with nucleotide sequences at HsCBL8. The overexpression of HsCBL8 in rice resulted in significant improvement of water protection in vivo and plasma membrane, more proline accumulation, and a reduction of overall Na+ uptake but little change in K+ concentration in the plant tissues. Notably, HsCBL8 did not act on some genes downstream of the rice CBL family genes, suggesting an interesting interaction between HsCBL8 and unknown factors to be further investigated.

Reconstruction and topological characterization of the sigma factor regulatory network of Mycobacterium tuberculosis.

Accessory sigma factors, which reprogram RNA polymerase to transcribe specific gene sets, activate bacterial adaptive responses to noxious environments. Here we reconstruct the complete sigma factor regulatory network of the human pathogen Mycobacterium tuberculosis by an integrated approach. The approach combines identification of direct regulatory interactions between M. tuberculosis sigma factors in an E. coli model system, validation of selected links in M. tuberculosis, and extensive literature review. The resulting network comprises 41 direct interactions among all 13 sigma factors. Analysis of network topology reveals (i) a three-tiered hierarchy initiating at master regulators, (ii) high connectivity and (iii) distinct communities containing multiple sigma factors. These topological features are likely associated with multi-layer signal processing and specialized stress responses involving multiple sigma factors. Moreover, the identification of overrepresented network motifs, such as autoregulation and coregulation of sigma and anti-sigma factor pairs, provides structural information that is relevant for studies of network dynamics.