Search

BackgroundCell metabolic reprogramming is a hallmark of tumor prognosis, and fatty acid metabolism (FAM) plays a crucial role in the tumor microenvironment (TME). However, the relationship between FAM, TME, and prognosis of acute myeloid leukemia (AML) patients remains elusive.MethodsWe extracted the single-cell RNA sequencing (scRNA-Seq) and bulk transcriptome data of AML patients from the TCGA and GEO databases and assessed the relationship between FAM, TME, and AML patient prognosis. We also performed functional enrichment (FE) assay to evaluate the significance of FAM in anti-AML immunosurveillance.ResultsOur scRNA-Seq analysis revealed that the leukemic stem cell (LSC)-enriched population exhibited elevated levels of FAM-related genes. Using these FAM-related genes, we developed a prognostic model that accurately estimated AML patient outcome. FE analysis showed that FAM was strongly related to alterations of TME-based immunosurveillance in AML patients. More importantly, we demonstrated that FAM inhibition via pharmaceutical targeting of PLA2G4A, a highly expressed FAM gene in AML patients with poor prognosis, enhanced the NK cell-mediated immunosurveillance in leukemia cells.ConclusionsLeukemic stem cell (LSC)-enriched population exhibited elevated levels of FAM-related genes. We have successfully established the FAM formula that predicts AML patient prognosis and alterations in the TME-based immunosurveillance. We also found that PLA2G4A was a highly expressed FAM gene in AML patients with poor prognoses. Pharmaceutical targeting of PLA2G4A increased the expression of NKG2DL in leukemia cells in vitro and thus enhanced the NK cell-mediated immunosurveillance.

BackgroundCDK12 inactivation leading to increased neoantigen burden has been hypothesized to sensitize tumors to immune checkpoint inhibition. Pan-cancer data regarding the frequency of CDK12 alterations are limited. We aimed to characterize CDK12 alterations across all cancer types through real-world clinical-grade sequencing.MethodsThis was a single-center retrospective analysis of 4994 cancer patients who underwent tissue or blood genomic profiling, including CDK12 assessment, conducted as part of routine care from December 2012 to January 2020. Prevalence, clinical characteristics, and treatment outcomes of patients with tumors with pathogenic CDK12 alterations were described.ResultsIn all, 39 (0.78%, n=39/4994) patients had pathogenic CDK12 alterations. Among CDK12-altered tumors, the most common organ site was prostate (n=9, 23.1%) followed by colorectal (n=5, 12.8%). Adenocarcinoma was the most common histology (n=26, 66.7%). Median follow-up from time of diagnosis was 4.02years. Median overall survival from time of metastasis was 4.43years (95% CI: 3.11-5.74). Ten patients with CDK12-altered tumors received at least one immune checkpoint inhibitor-containing regimen. The majority of patients (n=6/10, 60%) experienced an objective response. Progression-free survival for patients who had metastatic disease and received a checkpoint inhibitor-containing regimen was 1.16years (95% CI: 0.32-2.00).ConclusionCDK12 alterations are rare events across hematologic and solid tumor malignancies. They represent a clinically distinct molecular cancer subtype which may have increased responsiveness to checkpoint inhibition. Prospective studies are warranted to investigate checkpoint inhibition in CDK12-altered tumors.

Supplemental material, sj-docx-2-tan-10.1177_17562864221092566 for Association between cigarette smoking and Parkinson’s disease: a neuroimaging study by Chao Wang, Cheng Zhou, Tao Guo, Peiyu Huang, Xiaojun Xu and Minming Zhang in Therapeutic Advances in Neurological Disorders

Machine learning (ML) robustness and domain generalization are fundamentally correlated: they essentially concern data distribution shifts under adversarial and natural settings, respectively. On one hand, recent studies show that more robust (adversarially trained) models are more generalizable. On the other hand, there is a lack of theoretical understanding of their fundamental connections. In this paper, we explore the relationship between regularization and domain transferability considering different factors such as norm regularization and data augmentations (DA). We propose a general theoretical framework proving that factors involving the model function class regularization are sufficient conditions for relative domain transferability. Our analysis implies that ``robustness" is neither necessary nor sufficient for transferability; rather, regularization is a more fundamental perspective for understanding domain transferability. We then discuss popular DA protocols (including adversarial training) and show when they can be viewed as the function class regularization under certain conditions and therefore improve generalization. We conduct extensive experiments to verify our theoretical findings and show several counterexamples where robustness and generalization are negatively correlated on different datasets., Comment: ICML2022

Recent studies show that training deep neural networks (DNNs) with Lipschitz constraints are able to enhance adversarial robustness and other model properties such as stability. In this paper, we propose a layer-wise orthogonal training method (LOT) to effectively train 1-Lipschitz convolution layers via parametrizing an orthogonal matrix with an unconstrained matrix. We then efficiently compute the inverse square root of a convolution kernel by transforming the input domain to the Fourier frequency domain. On the other hand, as existing works show that semi-supervised training helps improve empirical robustness, we aim to bridge the gap and prove that semi-supervised learning also improves the certified robustness of Lipschitz-bounded models. We conduct comprehensive evaluations for LOT under different settings. We show that LOT significantly outperforms baselines regarding deterministic l2 certified robustness, and scales to deeper neural networks. Under the supervised scenario, we improve the state-of-the-art certified robustness for all architectures (e.g. from 59.04% to 63.50% on CIFAR-10 and from 32.57% to 34.59% on CIFAR-100 at radius rho = 36/255 for 40-layer networks). With semi-supervised learning over unlabelled data, we are able to improve state-of-the-art certified robustness on CIFAR-10 at rho = 108/255 from 36.04% to 42.39%. In addition, LOT consistently outperforms baselines on different model architectures with only 1/3 evaluation time., Comment: NeurIPS 2022

Supplemental material, sj-docx-2-tan-10.1177_17562864221092566 for Association between cigarette smoking and Parkinson’s disease: a neuroimaging study by Chao Wang, Cheng Zhou, Tao Guo, Peiyu Huang, Xiaojun Xu and Minming Zhang in Therapeutic Advances in Neurological Disorders

Supplementary materials

Additional file 1: Fig. S1. Variation of upconversion emission intensity as a function of Yb3+ dopant concentration when fixed the concentration of Ho3+ (0.1 mol%). Fig. S2. The UV–vis–NIR absorption spectra of SrF2:Yb3+/Ho3+ (12/0.1 mol%) NCs. Fig. S3. The dependence of luminescence intensity at (a) 1012 nm and (c) 2020 nm of SrF2:Yb3+/Ho3+ NCs on the temperature under 980 nm excitation. Arrhenius equation is used to fit the luminescence intensity dependent on temperature at (b) 1012 nm and (d) 2020 nm.

Supplemental material, sj-docx-1-tan-10.1177_17562864221092566 for Association between cigarette smoking and Parkinson’s disease: a neuroimaging study by Chao Wang, Cheng Zhou, Tao Guo, Peiyu Huang, Xiaojun Xu and Minming Zhang in Therapeutic Advances in Neurological Disorders

Federated Learning (FL) has become a practical and popular paradigm in machine learning. However, currently, there is no systematic solution that covers diverse use cases. Practitioners often face the challenge of how to select a matching FL framework for their use case. In this work, we present UniFed, the first unified benchmark for standardized evaluation of the existing open-source FL frameworks. With 15 evaluation scenarios, we present both qualitative and quantitative evaluation results of nine existing popular open-sourced FL frameworks, from the perspectives of functionality, usability, and system performance. We also provide suggestions on framework selection based on the benchmark conclusions and point out future improvement directions., Comment: Code: https://github.com/AI-secure/FLBenchmark-toolkit Website: https://unifedbenchmark.github.io/

Supplemental material, sj-docx-1-tan-10.1177_17562864221092566 for Association between cigarette smoking and Parkinson’s disease: a neuroimaging study by Chao Wang, Cheng Zhou, Tao Guo, Peiyu Huang, Xiaojun Xu and Minming Zhang in Therapeutic Advances in Neurological Disorders

Objection: The study aimed to explain the effects and mechanisms of long noncoding RNA (lncRNA) MIAT in development of colon cancer. Methods: The adjacent and cancer tissues which were collected from 30 cases colon cancer patients were evaluated pathology by HE staining and lncRNA MIAT expression by ISH assay. In the vitro study, SW620 cells were divided into NC, siRNA-control and siMIAT groups. Measuring cell proliferation, apoptosis, cell cycle, invasion and migration by CCK-8, flow cytometry, transwell and wound healing, and evaluating the relative proteins expressions by WB assay. Results: Compared with adjacent normal tissues, the lncRNA MIAT was significantly up-regulation in colon cancer tissues (P < 0.001). By cell experiment, with lncRNA MIAT knockdown, the cell proliferation was significantly depressed (P < 0.001) via significantly improving cell apoptosis and keeping the cell in the G1 phase (P < 0.001, respectively); the invasion SW620 cell number and wound healing rate were significantly suppressed (P < 0.001, respectively). The relative proteins (CDK2, Cyclin D3 and MMP-9) expressions of siMIAT group were significantly down-regulation compared with those of NC group (P < 0.001, respectively). Conclusion: Our study found that lncRNA MIAT expression is closely associated with colon cancer and may be one of key roles in progression and metastasis in colon cancer.

Harnessing the color tuning capability of upconversion nanoparticles (UCNPs) is of great significance in the field of advanced bioimaging and color display. Here, we report the tunable size and upconversion luminescence (UCL) multicolor in CaF2:Yb3+/Ho3+/Ce3+ UCNPs, which were synthesized by a facile hydrothermal method. It was found that the size of these UCNPs could be controlled (from 600 to 30 nm) by varying the concentration of Ce3+ ions. Under the excitation of a 980 nm continuous-wave (CW) laser, the UCL color of these UCNPs can be tuned from green to red as the doped Ce3+ ions gradually increase from 0 to 10 mol% and the red-to-green (R/G) ratio is enhanced remarkably. It is suggested that the cross-relaxation (CR) processes between Ho3+ and Ce3+ ions contribute to the tunable multicolor and enhancement of the R/G ratio. The mechanism of these processes is well supported by the time-resolved decay and near infrared (NIR) emission measurements.

Additional file 1: Table S1. Number (%) of participants in each survey location in Guangdong province during 2007 and 2016. Fig. S1. Scatter plot of 10-fold cross-validation of interpolated daily temperatures, relative humidity and sunlight. Fig. S2. The non-linear relationships between daily mean, daily minimum, daily maximum temperature and FPG in total population and different T2DM subgroups. Fig. S3. The non-linear relationships between daily mean and FPG in total population and different T2DM subgroups at k = 3,4,5. Fig. S4. The density distribution of FPG in total population and different T2DM subgroups. Table S2. Statistical description of FPG and daily mean ambient temperature in total population and different T2DM subgroups. Table S3. Number (%), Mean and SE of FPG and daily mean temperature according to month. Fig. S5. The relationships between lag1 to lag6 of ambient temperature and FPG in total population and different T2DM subgroups. Table S4. The interaction analyses about the association between temperature and FPG based on effect modifier (T2DM status). Fig. S6. The relationships between ambient temperature and FPG in total population and different T2DM subgroups additionally adjusted sunshine and precipitation. Fig. S7. The relationships between ambient temperature and FPG in total population and different T2DM subgroups in survey of 2010 and 2015. Fig. S8. The relationships between ambient temperature and FPG in normal fasting glucose and newly detected-T2DM subgroups for the diagnostic standard of T2DM based on FPG and 2-h plasma glucose in 2013 survey.

Gradient estimation and vector space projection have been studied as two distinct topics. We aim to bridge the gap between the two by investigating how to efficiently estimate gradient based on a projected low-dimensional space. We first provide lower and upper bounds for gradient estimation under both linear and nonlinear projections, and outline checkable sufficient conditions under which one is better than the other. Moreover, we analyze the query complexity for the projection-based gradient estimation and present a sufficient condition for query-efficient estimators. Built upon our theoretic analysis, we propose a novel query-efficient Nonlinear Gradient Projection-based Boundary Blackbox Attack (NonLinear-BA). We conduct extensive experiments on four image datasets: ImageNet, CelebA, CIFAR-10, and MNIST, and show the superiority of the proposed methods compared with the state-of-the-art baselines. In particular, we show that the projection-based boundary blackbox attacks are able to achieve much smaller magnitude of perturbations with 100% attack success rate based on efficient queries. Both linear and nonlinear projections demonstrate their advantages under different conditions. We also evaluate NonLinear-BA against the commercial online API MEGVII Face++, and demonstrate the high blackbox attack performance both quantitatively and qualitatively. The code is publicly available at https://github.com/AI-secure/NonLinear-BA., Comment: Accepted by AISTATS 2021; 9 pages excluding references and appendices

Recent studies show that deep neural networks (DNN) are vulnerable to adversarial examples, which aim to mislead DNNs by adding perturbations with small magnitude. To defend against such attacks, both empirical and theoretical defense approaches have been extensively studied for a single ML model. In this work, we aim to analyze and provide the certified robustness for ensemble ML models, together with the sufficient and necessary conditions of robustness for different ensemble protocols. Although ensemble models are shown more robust than a single model empirically; surprisingly, we find that in terms of the certified robustness the standard ensemble models only achieve marginal improvement compared to a single model. Thus, to explore the conditions that guarantee to provide certifiably robust ensemble ML models, we first prove that diversified gradient and large confidence margin are sufficient and necessary conditions for certifiably robust ensemble models under the model-smoothness assumption. We then provide the bounded model-smoothness analysis based on the proposed Ensemble-before-Smoothing strategy. We also prove that an ensemble model can always achieve higher certified robustness than a single base model under mild conditions. Inspired by the theoretical findings, we propose the lightweight Diversity Regularized Training (DRT) to train certifiably robust ensemble ML models. Extensive experiments show that our DRT enhanced ensembles can consistently achieve higher certified robustness than existing single and ensemble ML models, demonstrating the state-of-the-art certified L2-robustness on MNIST, CIFAR-10, and ImageNet datasets., Comment: ICLR 2022. 51 pages, 10 pages for main text. Forum and code: https://openreview.net/forum?id=tUa4REjGjTf

The UC emission spectra, UCL intensity dependence on the pump power density and decay curves of UC emissions of the CaF2 UCNCs under multi-wavelength excitation.

Supplemental material, sj-pdf-1-pic-10.1177_09544062211021445 for Investigation on trim control of semi-planing amphibious cargo truck using experimental and numerical approaches by Chengliang Sun, Xiaojun Xu and Tengan Zou in Proceedings of the Institution of Mechanical Engineers, Part C: Journal of Mechanical Engineering Science

Supplemental material, sj-pdf-1-pic-10.1177_09544062211021445 for Investigation on trim control of semi-planing amphibious cargo truck using experimental and numerical approaches by Chengliang Sun, Xiaojun Xu and Tengan Zou in Proceedings of the Institution of Mechanical Engineers, Part C: Journal of Mechanical Engineering Science

Parkinson's disease (PD) is commonly characterized by asymmetrical motor impairment. This study aimed to clarify the iron distributions in PD patients with significant motor asymmetry and their longitudinal alterations. This study included 123 PD patients and 121 normal controls. Thirty-eight PD patients were revisited. PD patients with significant motor asymmetry were identified by using an objective criterion. Inter-group, inter-hemisphere and inter-visit differences of regional tissue susceptibility were analyzed. Iron accumulation in dominantly and non-dominantly affected substantia nigra (SN) were observed in PD patients with motor asymmetry compared with normal controls (p < 0.005, Bonferroni corrected). Iron accumulation in the dominantly affected SN was significantly higher than that in the non-dominantly affected SN (p < 0.01, Bonferroni corrected). After follow-up, time effect on the iron content in SN was observed, directing to decrease in PD patients with motor asymmetry without hemispherical difference (p < 0.05). In conclusion, asymmetrical iron accumulation in SN was associated with the motor asymmetry in PD at baseline, while along the disease evolution iron content in SN became longitudinally decreased. All these findings provide new evidence for PD pathogenesis that the abnormal iron metabolism in SN is complicated and not always unidirectional.

Recent studies have shown that deep neural networks (DNNs) are vulnerable to adversarial attacks, including evasion and backdoor (poisoning) attacks. On the defense side, there have been intensive efforts on improving both empirical and provable robustness against evasion attacks; however, the provable robustness against backdoor attacks still remains largely unexplored. In this paper, we focus on certifying the machine learning model robustness against general threat models, especially backdoor attacks. We first provide a unified framework via randomized smoothing techniques and show how it can be instantiated to certify the robustness against both evasion and backdoor attacks. We then propose the first robust training process, RAB, to smooth the trained model and certify its robustness against backdoor attacks. We prove the robustness bound for machine learning models trained with RAB and prove that our robustness bound is tight. In addition, we theoretically show that it is possible to train the robust smoothed models efficiently for simple models such as K-nearest neighbor classifiers, and we propose an exact smooth-training algorithm that eliminates the need to sample from a noise distribution for such models. Empirically, we conduct comprehensive experiments for different machine learning (ML) models such as DNNs, support vector machines, and K-NN models on MNIST, CIFAR-10, and ImageNette datasets and provide the first benchmark for certified robustness against backdoor attacks. In addition, we evaluate K-NN models on a spambase tabular dataset to demonstrate the advantages of the proposed exact algorithm. Both the theoretic analysis and the comprehensive evaluation on diverse ML models and datasets shed light on further robust learning strategies against general training time attacks., Comment: IEEE Symposium on Security and Privacy 2023

Urbanicity, the impact of living in urban areas, is among the greatest environmental challenges for mental health. While urbanicity might be distinct in different sociocultural conditions and geographic locations, there are likely to exist common features shared in different areas of the globe. Understanding these common and specific relations of urbanicity with human brain and behavior will enable to assess the impact of urbanicity on mental disorders, especially in childhood and adolescence, where prevention and early interventions are likely to be most effective.We constructed from satellite-based remote sensing data a factor for urbanicity that was highly correlated with population density ground data. This factor, ‘UrbanSat’ was utilized in the Chinese CHIMGEN sample (N=831) and the longitudinal European IMAGEN cohort (N=810) to investigate if exposure to urbanicity during childhood and adolescence is associated with differences in brain structure and function in young adults, and if these changes are linked to behavior.Urbanicity was found negatively correlated with medial prefrontal cortex volume and positively correlated with cerebellar vermis volume in young adults from both China and Europe. We found an increased correlation of urbanicity with functional network connectivity within- and between- brain networks in Chinese compared to European participants. Urbanicity was highly correlated with a measure of perceiving a situation from the perspective of others, as well as symptoms of depression in both datasets. These correlations were mediated by the structural and functional brain changes observed. Susceptibility to urbanicity was greatest in two developmental windows during mid-childhood and adolescence.Using innovative technology, we were able to probe the relationship between urban upbringing with brain change and behavior in different sociocultural conditions and geographic locations. Our findings help to identify shared and distinct determinants of adolescent brain development and mental health in different regions of the world, thus contributing to targeted prevention and early-intervention programs for young people in their unique environment. Our approach may be relevant for public health, policy and urban planning globally.

In Parkinson's disease (PD), iron accumulation in the substantia nigra (SN) exacerbates oxidative stress and α-synuclein aggregation, leading to neuronal death. However, the influence of iron-related nigral degeneration on the subcortical function and global network configuration in PD remains unknown. Ninety PD patients and 38 normal controls underwent clinical assessments and multimodality magnetic resonance imaging scans. Iron accumulation in the inferior SN and disrupted functional connectivity between the bilateral striatums were observed in PD, and negative correlation between them was found in the whole population. The binarized functional network exhibited enhanced global efficiency and reduced local efficiency while the weighted functional network exhibited reduction in both, and both changes were correlated with nigral iron accumulation in PD. Mediation analysis demonstrated that the functional connectivity between bilateral striatums was a mediator between the nigral iron accumulation and weighted functional network alterations. In conclusion, our findings reveal that iron-related nigral degeneration possibly influences the functional topology mediated by striatal dysfunction, which extends the scientific understanding of PD pathogenesis.

BackgroundDifferent oscillations of brain networks could carry different dimensions of brain integration. We aimed to investigate oscillation-specific nodal alterations in patients with Parkinson's disease (PD) across early stage to middle stage by using graph theory-based analysis.MethodsEighty-eight PD patients including 39 PD patients in the early stage (EPD) and 49 patients in the middle stage (MPD) and 36 controls were recruited in the present study. Graph theory-based network analyses from three oscillation frequencies (slow-5: 0.01-0.027 Hz; slow-4: 0.027-0.073 Hz; slow-3: 0.073-0.198 Hz) were analyzed. Nodal metrics (e.g. nodal degree centrality, betweenness centrality and nodal efficiency) were calculated.ResultsOur results showed that (1) a divergent effect of oscillation frequencies on nodal metrics, especially on nodal degree centrality and nodal efficiency, that the anteroventral neocortex and subcortex had high nodal metrics within low oscillation frequencies while the posterolateral neocortex had high values within the relative high oscillation frequency was observed, which visually showed that network was perturbed in PD; (2) PD patients in early stage relatively preserved nodal properties while MPD patients showed widespread abnormalities, which was consistently detected within all three oscillation frequencies; (3) the involvement of basal ganglia could be specifically observed within slow-5 oscillation frequency in MPD patients; (4) logistic regression and receiver operating characteristic curve analyses demonstrated that some of those oscillation-specific nodal alterations had the ability to well discriminate PD patients from controls or MPD from EPD patients at the individual level; (5) occipital disruption within high frequency (slow-3) made a significant influence on motor impairment which was dominated by akinesia and rigidity.ConclusionsCoupling various oscillations could provide potentially useful information for large-scale network and progressive oscillation-specific nodal alterations were observed in PD patients across early to middle stages.

In the adaptive optics system, the accuracy and robustness of wavefront sensor greatly affect the ability of aberration detection and closed-loop correction. In the condition of the nonuniformity of amplitude distributions or insufficient of beacon light energy, it will cause accuracy decrease of Hartmann wavefront sensing due to the lack of sub-aperture light. Meanwhile, the real-time performance of the wavefront sensing-free adaptive system based on far-field spot inversion cannot meet the practical requirements. The method of the wavefront inversion based on deep learning is to directly obtain aberrations by inputting the far-field light intensity image, which can be used as an effective supplement to the adaptive optical system. Through numerical simulation, this paper proved that the deep residual neural network could directly predict the Zernike coefficient of the wavefront phase through the far-field spot. And experimental demonstrated the corrected residual RMS between input and reconstructed wavefront phase was 0.08 waves, the average computation time was less than 2 ms by GPU acceleration. This method can predict the Zernike coefficient of incident wavefront distortion more accurately, and has a good aberration correction capability, suitable for measuring and correcting the main components of wavefront distortion in traditional adaptive optics method, or providing a good initial wavefront estimation for optimized adaptive optics.

Based on the vehicle front crash finite element analysis, it shows that there is a large acceleration, so it needs further optimization. In order to improve the performance of vehicle collision, eight parts were selected which have large impact for the result, its thickness as design variables to the right of the B-pillar acceleration peak of optimization goal; 17 sample points were selected by Latin hypercube sampling method. Many structure parameters are optimized using sequential quadratic program (SQP) based on the surrogate model. The results show that the improved RSM has high accuracy; the right B-pillar acceleration reduced approximately 22.8%, reached the expected objective and was more conducive to the occupant safety.

In this paper, we show that the recent integration of statistical models with deep recurrent neural networks provides a new way of formulating volatility (the degree of variation of time series) models that have been widely used in time series analysis and prediction in finance. The model comprises a pair of complementary stochastic recurrent neural networks: the generative network models the joint distribution of the stochastic volatility process; the inference network approximates the conditional distribution of the latent variables given the observables. Our focus here is on the formulation of temporal dynamics of volatility over time under a stochastic recurrent neural network framework. Experiments on real-world stock price datasets demonstrate that the proposed model generates a better volatility estimation and prediction that outperforms mainstream methods, e.g., deterministic models such as GARCH and its variants, and stochastic models namely the MCMC-based stochvol as well as the Gaussian-process-based, on average negative log-likelihood.