Artificial intelligence to unlock real-world evidence in clinical oncology: A primer on recent advances.

PURPOSE: Real world evidence is crucial to understanding the diffusion of new oncologic therapies, monitoring cancer outcomes, and detecting unexpected toxicities. In practice, real world evidence is challenging to collect rapidly and comprehensively, often requiring expensive and time-consuming manual case-finding and annotation of clinical text. In this Review, we summarise recent developments in the use of artificial intelligence to collect and analyze real world evidence in oncology. METHODS: We performed a narrative review of the major current trends and recent literature in artificial intelligence applications in oncology. RESULTS: Artificial intelligence (AI) approaches are increasingly used to efficiently phenotype patients and tumors at large scale. These tools also may provide novel biological insights and improve risk prediction through multimodal integration of radiographic, pathological, and genomic datasets. Custom language processing pipelines and large language models hold great promise for clinical prediction and phenotyping. CONCLUSIONS: Despite rapid advances, continued progress in computation, generalizability, interpretability, and reliability as well as prospective validation are needed to integrate AI approaches into routine clinical care and real-time monitoring of novel therapies.

Zuogui Pills alleviate cyclophosphamide-induced ovarian aging by reducing oxidative stress and restoring the stemness of oogonial stem cells through the Nrf2/HO-1 signaling pathway.

ETHNOPHARMACOLOGICAL RELEVANCE: Zuogui Pill (ZGP) is a traditional herbal formula of Chinese Medicine with a long history of use in alleviating ovarian aging. AIM OF THE STUDY: To examine the impact of ZGP on oxidative stress and the stemness of oogonial stem cells (OSCs) in cyclophosphamide (CTX)-induced ovarian aging, as well as its molecular mechanisms involving the nuclear factor erythroid 2-related factor 2 (Nrf2, NFE2L2)/heme oxygenase-1 (HO-1, Hmox1) pathway. MATERIALS AND METHODS: Female Sprague-Dawley (SD) rats were randomly divided into seven groups: control, model (CTX), estradiol valerate (EV, 0.103 mg/kg), ZGP-L (low dose Zuogui Pill, 1.851 g/kg), ZGP-H (high dose Zuogui Pill, 3.702 g/kg), ML385 (30 mg/kg), and ML385+ZGP-L. After CTX modeling, the EV, ZGP-L, ZGP-H, and ML385+ZGP-L groups were treated by gavage for 8 weeks, while the ML385 and ML385+ZGP-L groups were administered the Nrf2 antagonist ML385 twice a week. OSCs were isolated after modeling and then treated with drug serum containing 10% ZGP or 10 µM ML385. The general conditions of the rats, including body weight, ovarian weight/body weight ratio, and estrous cycle, were observed. Ovarian ultrastructure, follicle and corpus luteum counts were assessed via hematoxylin and eosin (H&E) staining. Serum hormone levels were measured using enzyme-linked immunosorbent assay (ELISA). Nrf2/HO-1 pathway, stem cell, germ cell, and cell cycle biomarkers were analyzed by qPCR and Western blot. Cell viability was assessed by cell counting kit-8 (CCK-8) assay. Oxidative stress biomarkers were evaluated using flow cytometry and assay kits. Immunofluorescence was employed to detect and locate OSCs in the ovary, quantify the average fluorescence intensity, and identify OSCs. RESULTS: After ZGP treatment, rats with CTX-induced ovarian aging exhibited improved general condition, increased body weight, higher total ovarian weight to body weight ratio, and a restoration of the estrous cycle similar to the control group. Serum levels of estradiol (E2) and follicle stimulating hormone (FSH), two sex hormones, were also improved. Ovarian ultrastructure and follicle count at all stages showed improvement. Moreover, the viability and proliferation capacity of OSCs were enhanced following ZGP intervention. The Nrf2/HO-1 pathway was found to be down-regulated in CTX-induced aging ovarian OSCs. However, ZGP reversed this effect by activating the expression of Nrf2, HO-1, and NAD(P)H oxidoreductase 1 (NQO1), increasing the activity of antioxidant enzymes superoxide dismutase (SOD) and glutathione peroxidase (GSH-PX), and reducing the accumulation of malonaldehyde (MDA) and reactive oxygen species (ROS), thus restoring resistance to oxidative stress. Additionally, ZGP improved the cell cycle of OSCs, up-regulated the expression of Cyclin D1 and Cyclin E1, restored cell stemness, promoted proliferation, enhanced the expression of cell stemness markers octamer-binding transcription factor 4 (Oct4) and mouse VASA homolog (MVH), and down-regulated the expression of P21, thereby inhibiting apoptosis. The therapeutic effects of ZGP against oxidative stress and restoration of cell stemness were attenuated following inhibition of the Nrf2 signaling pathway using ML385. CONCLUSIONS: ZGP protected against CTX-induced ovarian aging by restoring normal ovarian function, alleviating oxidative stress in aging OSCs, promoting OSCs proliferation, and restoring their stemness in rats, possibly through regulating the Nrf2/HO-1 pathway.

INSURE: An Information Theory iNspired diSentanglement and pURification modEl for Domain Generalization.

Domain Generalization (DG) aims to learn a generalizable model on the unseen target domain by only training on the multiple observed source domains. Although a variety of DG methods have focused on extracting domain-invariant features, the domain-specific class-relevant features have attracted attention and been argued to benefit generalization to the unseen target domain. To take into account the class-relevant domain-specific information, in this paper we propose an Information theory iNspired diSentanglement and pURification modEl (INSURE) to explicitly disentangle the latent features to obtain sufficient and compact (necessary) class-relevant feature for generalization to the unseen domain. Specifically, we first propose an information theory inspired loss function to ensure the disentangled class-relevant features contain sufficient class label information and the other disentangled auxiliary feature has sufficient domain information. We further propose a paired purification loss function to let the auxiliary feature discard all the class-relevant information and thus the class-relevant feature will contain sufficient and compact (necessary) class-relevant information. Moreover, instead of using multiple encoders, we propose to use a learnable binary mask as our disentangler to make the disentanglement more efficient and make the disentangled features complementary to each other. We conduct extensive experiments on five widely used DG benchmark datasets including PACS, VLCS, OfficeHome, TerraIncognita, and DomainNet. The proposed INSURE achieves state-of-the-art performance. We also empirically show that domain-specific class-relevant features are beneficial for domain generalization. The code is available at https://github.com/yuxi120407/INSURE.

Vitamin K2 ameliorates osteoarthritis by suppressing ferroptosis and extracellular matrix degradation through activation GPX4's dual functions.

Vitamin K2 (VK2) is an effective compound for anti-ferroptosis and anti-osteoporosis, and Semen sojae praeparatum (Dandouchi in Chinese) is the main source of VK2. Chondrocyte ferroptosis and extracellular matrix (ECM) degradation playing a role in the pathogenesis of osteoarthritis (OA). Glutathione peroxidase 4 (GPX4) is the intersection of two mechanisms in regulating OA progression. But no studies have elucidated the therapeutic effects and mechanisms of VK2 on OA. This study utilized an in vivo rat OA model created via anterior cruciate ligament transection (ACLT) and an in vitro chondrocyte oxidative damage model induced by TBHP to investigate the protective effects and mechanisms of action of VK2 in OA. Knee joint pain in mice was evaluated using the Von Frey test. Micro-CT and Safranin O-Fast Green staining were employed to observe the extent of damage to the tibial cartilage and subchondral bone, while immunohistochemistry and PCR were used to examine GPX4 levels in joint cartilage. The effects of VK2 on rat chondrocyte viability were assessed using CCK-8 and flow cytometry assays, and chondrocyte morphology was observed with toluidine blue and alcian blue staining. The impact of VK2 on intracellular ferroptosis-related markers was observed using fluorescent staining and flow cytometry. Protein expression changes were detected by immunofluorescence and Western blot analysis. Furthermore, specific protein inhibitors were applied to confirm the dual-regulatory effects of VK2 on GPX4. VK2 can increase bone mass and cartilage thickness in the subchondral bone of the tibia, and reduce pain and the OARSI score induced by OA. Immunohistochemistry results indicate that VK2 exerts its anti-OA effects by regulating GPX4 to delay ECM degradation. VK2 can inhibit the activation of the MAPK/NFκB signaling pathway caused by reduced expression of intracellular GPX4, thereby decreasing ECM degradation. Additionally, VK2 can reverse the inhibitory effect of RSL3 on GPX4, increase intracellular GSH content and the GSH/GSSG ratio, reduce MDA content, and rescue chondrocyte ferroptosis. The protective mechanism of VK2 may involve its dual-target regulation of GPX4, reducing chondrocyte ferroptosis and inhibiting the MAPK/NFκB signaling pathway to decelerate the degradation of the chondrocyte extracellular matrix.

Exploring Robust Features for Improving Adversarial Robustness.

While deep neural networks (DNNs) have revolutionized many fields, their fragility to carefully designed adversarial attacks impedes the usage of DNNs in safety-critical applications. In this article, we strive to explore the robust features that are not affected by the adversarial perturbations, that is, invariant to the clean image and its adversarial examples (AEs), to improve the model's adversarial robustness. Specifically, we propose a feature disentanglement model to segregate the robust features from nonrobust features and domain-specific features. The extensive experiments on five widely used datasets with different attacks demonstrate that robust features obtained from our model improve the model's adversarial robustness compared to the state-of-the-art approaches. Moreover, the trained domain discriminator is able to identify the domain-specific features from the clean images and AEs almost perfectly. This enables AE detection without incurring additional computational costs. With that, we can also specify different classifiers for clean images and AEs, thereby avoiding any drop in clean image accuracy.

Correlative single-cell hard X-ray computed tomography and X-ray fluorescence imaging.

X-ray computed tomography (XCT) and X-ray fluorescence (XRF) imaging are two non-invasive imaging techniques to study cellular structures and chemical element distributions, respectively. However, correlative X-ray computed tomography and fluorescence imaging for the same cell have yet to be routinely realized due to challenges in sample preparation and X-ray radiation damage. Here we report an integrated experimental and computational workflow for achieving correlative multi-modality X-ray imaging of a single cell. The method consists of the preparation of radiation-resistant single-cell samples using live-cell imaging-assisted chemical fixation and freeze-drying procedures, targeting and labeling cells for correlative XCT and XRF measurement, and computational reconstruction of the correlative and multi-modality images. With XCT, cellular structures including the overall structure and intracellular organelles are visualized, while XRF imaging reveals the distribution of multiple chemical elements within the same cell. Our correlative method demonstrates the feasibility and broad applicability of using X-rays to understand cellular structures and the roles of chemical elements and related proteins in signaling and other biological processes.

Allosteric regulation in SARS-CoV-2 spike protein.

Allosteric regulation is common in protein-protein interactions and is thus promising in drug design. Previous experimental and simulation work supported the presence of allosteric regulation in the SARS-CoV-2 spike protein. Here the route of allosteric regulation in SARS-CoV-2 spike protein is examined by all-atom explicit solvent molecular dynamics simulations, contrastive machine learning, and the Ohm approach. It was found that peptide binding to the polybasic cleavage sites, especially the one at the first subunit of the trimeric spike protein, activates the fluctuation of the spike protein's backbone, which eventually propagates to the receptor-binding domain on the third subunit that binds to ACE2. Remarkably, the allosteric regulation routes starting from the polybasic cleavage sites share a high fraction (39-67%) of the critical amino acids with the routes starting from the nitrogen-terminal domains, suggesting the presence of an allosteric regulation network in the spike protein. Our study paves the way for the rational design of allosteric antibody inhibitors.

Biochanin A abrogates osteoclastogenesis in type 2 diabetic osteoporosis via regulating ROS/MAPK signaling pathway based on integrating molecular docking and experimental validation.

BACKGROUND: There are accumulating type 2 diabetes patients who have osteoporosis simultaneously. More effective therapeutic strategies should be discovered. Biochanin A (BCA) has been indicated that can play a role in improving metabolic disorders of type 2 diabetes and preventing osteoporosis. But whether BCA can treat type 2 diabetic osteoporosis has not been studied. PURPOSE: To investigate if the BCA can protect against type 2 diabetic osteoporosis and clarify the mechanism. METHODS: Micro-CT and histology assays were performed to detect the trabecular bone and analyze the bone histomorphology effect of BCA. CCK-8 assay was performed to detect the toxicity of BCA. TRAcP staining, immunofluorescence and hydroxyapatite resorption assay were used to observe osteoclasts differentiation and resorptive activity. Molecular docking provided evidence about BCA regulating the MAPK axis via prediction by the algorithm. QRT-PCR and Western Blotting were utilized to detect the expression of osteoclastogenesis-related markers and MAPK signaling pathway. RESULTS: Accumulation of bone volume after BCA treatment could be found based on the 3D reconstruction. Besides, there were fewer osteoclasts in db/db mice treated with BCA than db/db mice treated with saline. In vitro, we found that BCA hadn't toxicity in osteoclasts precursor, but also inhibited differentiation of osteoclasts. Further, we found that BCA suppresses osteoclastogenesis via ROS/MAPK signaling pathway. CONCLUSION: BCA can prevent type 2 diabetic osteoporosis by restricting osteoclast differentiation via ROS/MAPK signaling pathway.

Corylifol A suppresses osteoclastogenesis and alleviates ovariectomy-induced bone loss via attenuating ROS production and impairing mitochondrial function.

Osteoporosis is a systemic disease characterized by an imbalance in bone homeostasis, where osteoblasts fail to fully compensate for the bone resorption induced by osteoclasts. Corylifol A, a flavonoid extracted from Fructus psoraleae, has been identified as a potential treatment for this condition. Predictions from network pharmacology and molecular docking studies suggest that Corylifol A exhibits strong binding affinity with NFATc1, Nrf2, PI3K, and AKT1. Empirical evidence from in vivo experiments indicates that Corylifol A significantly mitigates systemic bone loss induced by ovariectomy by suppressing both the generation and activation of osteoclasts. In vitro studies further showed that Corylifol A inhibited the activation of PI3K-AKT and MAPK pathways and calcium channels induced by RANKL in a time gradient manner, and specifically inhibited the phosphorylation of PI3K, AKT, GSK3 ß, ERK, CaMKII, CaMKIV, and Calmodulin. It also diminishes ROS production through Nrf2 activation, leading to a decrease in the expression of key regulators such as NFATcl, C-Fos, Acp5, Mmp9, and CTSK that are involved in osteoclastogenesis. Notably, our RNA-seq analysis suggests that Corylifol A primarily impacts mitochondrial energy metabolism by suppressing oxidative phosphorylation. Collectively, these findings demonstrate that Corylifol A is a novel inhibitor of osteoclastogenesis, offering potential therapeutic applications for diseases associated with excessive bone resorption.

Combination of curcumin and catalase protects against chondrocyte injury and knee osteoarthritis progression by suppressing oxidative stress.

Knee Osteoarthritis (KOA) is an age-related progressive degenerative joint disease, which is featured with pain, joint deformity, and disability. Accumulating evidence indicated oxidative stress plays a crucial role in the occurrence and development of KOA. Curcumin is a polyphenolic compound with significant antioxidant activity among various diseases while catalase (CAT) is an enzyme degrading hydrogen peroxide in treating oxidative diseases. We previously showed that the expression of CAT was low in cartilage. However, the combination of curcumin and CAT in KOA is still elusive. In this study, we demonstrated that the combination of curcumin and CAT has the potential to inhibit the IL1ß-induced chondrocyte apoptosis without cytotoxicity in vitro. Mechanistically, we found that the synergistic application curcumin and CAT not only promotes curcumin's regulation of the NRF2/HO-1 signaling pathway to enhance antioxidant enzyme expression to remove superoxide radicals, but also CAT can further remove downstream hydrogen peroxide which enhances the ability to scavenge reactive oxygen species (ROS). In vivo, studies revealed that combination of curcumin and catalase could better inhibit oxidative stress-induced chondrocyte injury by promoting the expression of ROS scavenging enzymes. In sum, the combination of curcumin and catalase can be used to treat KOA. Thus, combination of curcumin and catalase may act as a novel therapeutic agent to manage KOA and our research gives a rationale for their combined use in the therapeutic of KOA.

Cardamonin protects against iron overload induced arthritis by attenuating ROS production and NLRP3 inflammasome activation via the SIRT1/p38MAPK signaling pathway.

Iron homeostasis plays an essential role in joint health, while iron overload can cause damage and death of cartilage cells. Cardamonin (CAR) is a substance found in the fruit of the chasteberry plant and has anti-inflammatory and anti-tumor activities. We first administered iron dextran (500 mg/kg) intraperitoneally to establish an iron overload mouse model and surgically induced osteoarthritis. The extent of OA and iron deposition were assessed using Micro-ct, Safranin-O/fast green staining, H&E staining, and Prussian Blue 10 weeks later. We administered primary chondrocytes with Ferric Ammonium Citrate (FAC) to evaluate the chondrocyte changes. Chondrocytes were identified in vitro by toluidine blue staining, and chondrocyte viability was evaluated by CCK-8. The rate of apoptosis was determined by Annexin V-FITC/PI assay. The mechanism of action of CAR was verified by adding the SIRT1 inhibitor EX527, and the expression of SIRT1 and MAPK signaling pathways was detected by Western blot. Iron overload also promoted chondrocyte apoptosis, a process that was reversed by CAR. In addition, CAR reduced NLRP3 inflammasome production via the SIRT1-MAPK pathway, and the SIRT1 inhibitor EX527 inhibited the treatment of OA by CAR.CAR inhibited cartilage degeneration induced by iron overload both in vivo and in vitro. Besides, our study showed that iron overload not only inhibited type II collagen expression but also induced MMP expression by catalyzing the generation of NLRP3 inflammasome. Our results suggest that CAR can treat KOA by promoting SIRT1 expression and inhibiting p38MAPK pathway expression to reduce the production of NLRP3 inflammasome vesicles.

Isoimperatorin attenuates bone loss by inhibiting the binding of RANKL to RANK.

Osteoporosis, an immune disease characterized by bone mass loss and microstructure destruction, is often seen in postmenopausal women. Isoimperatorin (ISO), a bioactive, natural furanocoumarin isolated from many traditional Chinese herbal medicines, has therapeutic effects against various diseases; however, its effect on bone homeostasis remains unclear. In this study, we investigated the effect of ISO on the differentiation and activation of osteoclast and its molecular mechanism in vitro, and evaluated the effect of ISO on bone metabolism by ovariectomized (OVX) rat model. In vitro experiments showed that ISO affected RANKL-induced MAPK, NFAT, NFATc1 trafficking and expression, osteoclast F-actin banding, osteoclast-characteristic gene expression, ROS inhibitory activity, and calcium oscillations, NF-κB signaling pathway. In vivo experiments showed that oral administration of ISO effectively reduced bone loss caused by ovariectomy and retained bone mass.Collectively, ISO inhibits RANK/RANKL binding, thereby reducing the activity of NFATc1, calcium, and ROS and inhibiting osteoclast generation. In addition, ISO protects bone mass by slowing osteoclast production and downregulating NFATc1 gene and protein expression in the bone tissue microenvironment and inhibits OVX-induced bone loss in vivo.

Isoliensinine suppresses bone loss by targeted inhibition of RANKL-RANK binding.

BACKGROUND: Osteoporosis, a systemic metabolic bone disease, is often caused by the disruption of dynamic equilibrium between osteoclasts and osteoblasts. Overactive bone resorption, in which osteoclasts play a major role, is one of the most common and major causes of osteoporosis. Less costly and more effective drug treatments for this disease are needed. Based on the combination of molecular docking techniques and in vitro cell assays, this study aimed to explore the mechanism by which Isoliensinine (ILS) protects the bone loss by inhibiting osteoclast differentiation. METHODS: A virtual docking model based on molecular docking technology was used to investigate the interactions between ILS and the Receptor Activator of Nuclear Kappa-B (RANK)/Receptor Activator of Nuclear Kappa-B Ligand (RANKL).In this study, we determined the effective dose of action of ILS to inhibit osteoclast differentiation in vitro and, using bone resorption experiments, RT-CPR and Western Blot investigated the effects of ILS on bone resorption function and normal expression of osteoclast-associated genes and proteins, and validated potential mechanistic pathways. In vivo experiments revealed that ILS could inhibit bone loss through Micro-CT results. Finally, the molecular interaction between ILS and RANK/RANKL was investigated using biomolecular interaction experiments to verify the correctness and accuracy of the computational results. RESULTS: ILS binds to RANK and RANKL proteins, respectively, through virtual molecular docking. The Surface Plasmon Resonance (SPR) experiment results revealed that phosphorylated JNK, ERK, P38, and P65 expression was significantly downregulated when ILS were targeted to inhibit RANKL/RANK binding. At the same time, the expression of IKB-a was significantly increased under the stimulation of ILS, which rescued the degradation of IKB-a. ILS can significantly inhibit the levels of Reactive Oxygen Species (ROS) and Ca2 + concentration in vitro. Finally, the results of Micro-CT showed that ILS can significantly inhibit bone loss in vivo, indicating that ILS has a potential role in the treatment of osteoporosis. CONCLUSION: ILS inhibits osteoclast differentiation and bone loss by preventing the normal binding of RANKL/RANK, affecting downstream signaling pathways, including MAPK.NF-KB, ROS, Ca2+, genes, and proteins.

Dehydromiltirone inhibits osteoclast differentiation in RAW264.7 and bone marrow macrophages by modulating MAPK and NF-κB activity.

Background: Osteoporosis is a type of systematic metabolic bone disease caused by the decrease in osteogenic activity or excessive resorption of bone with the relative enhancement of osteoclast function. As osteoporosis seriously affects the quality of patients' life, effective drugs are needed to treat this disease. Based on the combination of network pharmacology and cellular studies, this study aimed to investigate the probable mechanism of Dehydromiltirone (DHT) in the treatment of osteoporosis. Method: The targets of DHT in osteoporosis were searched using the PharmGKB, OMIM, and Genecard platforms. The PPI core targets, and the GO and KEGG enrichment analysis results were obtained using Cytoscape software, and the David and Metascape databases, respectively. The network pharmacology results were also verified via in vitro cellular experiments. Results: Through network pharmacology and docking analysis, we found DHT was involved in peptide tyrosine phosphorylation, cell surface receptor tyrosine kinase signaling pathways, and MAPK signaling pathways. According to the molecular docking results, the binding of DHT to MAPK14 was more stable than other proteins, which suggests that DHT may affect osteoclast formation through the MAPK signaling pathway. Moreover, DHT was found to inhibit the expression of osteoclast-associated genes, including NFATc1, CTSK, c-Fos, Acp5, and MMP9; as well as the phosphorylation of P38, ERK, and JNK of the MAPK signaling pathway; and the degradation of IκB-α of NF-κB signaling pathway. Conclusion: DHT exhibited an anti-osteoclastogenesis effect by reducing the expression of related genes, ultimately inhibiting bone resorption in vitro.

A tool for predicting overall survival in patients with Ewing sarcoma: a multicenter retrospective study.

OBJECTIVE: The aim of this study was to establish and validate a clinical prediction model for assessing the risk of metastasis and patient survival in Ewing's sarcoma (ES). METHODS: Patients diagnosed with ES from the Surveillance, Epidemiology and End Results (SEER) database for the period 2010-2016 were extracted, and the data after exclusion of vacant terms was used as the training set (n=767). Prediction models predicting patients' overall survival (OS) at 1 and 3 years were created by cox regression analysis and visualized using Nomogram and web calculator. Multicenter data from four medical institutions were used as the validation set (n=51), and the model consistency was verified using calibration plots, and receiver operating characteristic (ROC) verified the predictive ability of the model. Finally, a clinical decision curve was used to demonstrate the clinical utility of the model. RESULTS: The results of multivariate cox regression showed that age, , bone metastasis, tumor size, and chemotherapy were independent prognostic factors of ES patients. Internal and external validation results: calibration plots showed that the model had a good agreement for patient survival at 1 and 3 years; ROC showed that it possessed a good predictive ability and clinical decision curve proved that it possessed good clinical utility. CONCLUSIONS: The tool built in this paper to predict 1- and 3-year survival in ES patients ( https://drwenleli0910.shinyapps.io/EwingApp/ ) has a good identification and predictive power.

Do we have to pursue complete reduction after PVA in osteoporotic vertebral compression fractures: a finite element analysis.

BACKGROUND: Consensus regarding the optimal amount of bone cement and vertebral height in the treatment of osteoporotic vertebral compression fractures (OVCFs) is lacking. Our purpose was to explore the optimal amount of bone cement and vertebral height in OVCF after percutaneous vertebral augmentation (PVA). METHODS: A three-dimensional finite element model of the L1-L3 segments was constructed from CT scans of aging osteoporosis patients. Four different postoperative vertebral height models were simulated according to Genant semiquantitative grades 0, 1, 2, and 3. The volume of bone cement filling ranged from 3 ml to 6 ml. These models evaluated the von Mises stress of injured vertebral bodies, adjacent vertebral bodies and intervertebral discs under flexion, extension, left flexion, and right flexion after PVA. RESULTS: When the bone cement content was held constant, as the height of the vertebral body decreased, the stress of the L2 vertebral body decreased during left flexion and right flexion, but the stress of the L2 vertebral body increased and decreased during flexion and extension. As the height of the vertebral body decreased, the stress of the L1-L2 intervertebral disc increased. There was no significant change in the stress of other adjacent vertebrae or intervertebral discs. When the Genant grade was 0, 1, or 2 (3 ml and 4 ml), the stress of the overall vertebral body was closest to normal. CONCLUSIONS: When the height of the vertebral body is restored to the same height, a bone cement filling volume of 3 ml to 6 ml is suitable and will not produce a significant change in the stress of the vertebral body or adjacent vertebral body. As vertebral body height was lost, it may promote the degeneration of the intervertebral disc above the injury vertebrae after PVA. It is appropriate for the height of the vertebral body to return to Genant grade 0 or Genant grade 1 after surgery. When the height of the vertebral body has Genant grade 2 status, it was best to use 3 ml to 4 ml of bone cement filling. Therefore, when treating OVCFs, clinicians do not need to pursue complete reduction of the vertebral body. It is also important to verify the biomechanics results in clinical studies.

End-to-end orientation estimation from 2D cryo-EM images.

Cryo-electron microscopy (cryo-EM) is a Nobel Prize-winning technique for determining high-resolution 3D structures of biological macromolecules. A 3D structure is reconstructed from hundreds of thousands of noisy 2D projection images. However, existing 3D reconstruction methods are still time-consuming, and one of the major computational bottlenecks is recovering the unknown orientation of the particle in each 2D image. The dominant methods typically exploit an expensive global search on each image to estimate the missing orientations. Here, a novel end-to-end supervised learning method is introduced to directly recover the missing orientations from 2D cryo-EM images. A neural network is used to approximate the mapping from images to orientations. A robust loss function is proposed for optimizing the parameters of the network, which can handle both asymmetric and symmetric 3D structures. Experiments on synthetic data sets with various symmetry types confirm that the neural network is capable of recovering orientations from 2D cryo-EM images, and the results on a real cryo-EM data set further demonstrate its potential under more challenging imaging conditions.

Point Adversarial Self-Mining: A Simple Method for Facial Expression Recognition.

In this article, we propose a simple yet effective approach, called point adversarial self mining (PASM), to improve the recognition accuracy in facial expression recognition (FER). Unlike previous works focusing on designing specific architectures or loss functions to solve this problem, PASM boosts the network capability by simulating human learning processes: providing updated learning materials and guidance from more capable teachers. Specifically, to generate new learning materials, PASM leverages a point adversarial attack method and a trained teacher network to locate the most informative position related to the target task, generating harder learning samples to refine the network. The searched position is highly adaptive since it considers both the statistical information of each sample and the teacher network capability. Other than being provided new learning materials, the student network also receives guidance from the teacher network. After the student network finishes training, the student network changes its role and acts as a teacher, generating new learning materials and providing stronger guidance to train a better student network. The adaptive learning materials generation and teacher/student update can be conducted more than one time, improving the network capability iteratively. Extensive experimental results validate the efficacy of our method over the existing state of the arts for FER.

Supervised Machine-Learning-Based Determination of Three-Dimensional Structure of Metallic Nanoparticles.

Tracking the structure of heterogeneous catalysts under operando conditions remains a challenge due to the paucity of experimental techniques that can provide atomic-level information for catalytic metal species. Here we report on the use of X-ray absorption near-edge structure (XANES) spectroscopy and supervised machine learning (SML) for refining the 3D geometry of metal catalysts. SML is used to unravel the hidden relationship between the XANES features and catalyst geometry. To train our SML method, we rely on ab initio XANES simulations. Our approach allows one to solve the structure of a metal catalyst from its experimental XANES, as demonstrated here by reconstructing the average size, shape, and morphology of well-defined platinum nanoparticles. This method is applicable to the determination of the nanoparticle structure in operando studies and can be generalized to other nanoscale systems. It also allows on-the-fly XANES analysis and is a promising approach for high-throughput and time-dependent studies.