The BUB3-BUB1 Complex Promotes Telomere DNA Replication.

Telomeres and telomere-binding proteins form complex secondary nucleoprotein structures that are critical for genome integrity but can present serious challenges during telomere DNA replication. It remains unclear how telomere replication stress is resolved during S phase. Here, we show that the BUB3-BUB1 complex, a component in spindle assembly checkpoint, binds to telomeres during S phase and promotes telomere DNA replication. Loss of the BUB3-BUB1 complex results in telomere replication defects, including fragile and shortened telomeres. We also demonstrate that the telomere-binding ability of BUB3 and kinase activity of BUB1 are indispensable to BUB3-BUB1 function at telomeres. TRF2 targets BUB1-BUB3 to telomeres, and BUB1 can directly phosphorylate TRF1 and promote TRF1 recruitment of BLM helicase to overcome replication stress. Our findings have uncovered previously unknown roles for the BUB3-BUB1 complex in S phase and shed light on how proteins from diverse pathways function coordinately to ensure proper telomere replication and maintenance.

Multi-omics Mendelian randomization integrating GWAS, eQTL and pQTL data revealed GSTM4 as a potential drug target for migraine.

INTRODUCTION: Migraine, as a complex neurological disease, brings heavy burden to patients and society. Despite the availability of established therapies, existing medications have limited efficacy. Thus, we aimed to find the drug targets that improve the prognosis of migraine. METHOD: We used Mendelian Randomization (MR) and Summary-data-based MR (SMR) analyses to study possible drug targets of migraine by summary statistics from FinnGen cohorts (nCase = 44,616, nControl = 367,565), with further replication in UK Biobank (nCase = 26,052, nControl = 487,214). Genetic instruments were obtained from eQTLGen and UKB-PPP to verify the drug targets at the gene expression and protein levels. The additional analyses including Bayesian co-localization, the heterogeneity in dependent instruments(HEIDI), Linkage Disequilibrium Score(LDSC), bidirectional MR, multivariate MR(MVMR), heterogeneity test, horizontal pleiotropy test, and Steiger filtering were implemented to consolidate the findings further. Lastly, drug prediction analysis and phenome-wide association study(PheWAS) were employed to imply the possibility of drug targets for future clinical applications. RESULT: The MR analysis of eQTL data showed that four drug targets (PROCR, GSTM4, SLC4A1, and TNFRSF10A) were significantly associated with migraine risk in both the FinnGen and UK Biobank cohorts. However, only GSTM4 exhibited consistent effect directions across the two outcomes(Discovery cohort: OR(95%CI) = 0.94(0.93-0.96); p = 2.70e - 10; Replication cohort: OR(95%CI) = 0.93(0.91-0.94); p = 4.21e - 17). Furthermore, GSTM4 passed the SMR at p < 0.05 and HEIDI test at p > 0.05 at both the gene expression and protein levels. The protein-level MR analysis revealed a strong correlation between genetically predicted GSTM4 with a lower incidence of migraine and its subtypes(Overall migraine: OR(95%CI) = 0.91(0.87-0.95); p = 6.98e-05; Migraine with aura(MA): OR(95%CI) = 0.90(0.85-0.96); p = 2.54e-03; Migraine without aura(MO): OR(95%CI) = 0.90(0.83-0.96); p = 2.87e-03), indicating a strong co-localization relationship (PPH4 = 0.86). Further analyses provided additional validation for the possibility of GSTM4 as a migraine treatment target. CONCLUSION: This study identifies GSTM4 as a potential druggable gene and promising therapeutic target for migraine.

A simple method for restoring the femoral head center in hip arthroplasty: a 3-dimensional analysis in the Chinese population.

BACKGROUND: Various authors have successfully demonstrated that the distance from the greater trochanter to the femoral head center (GTFHC) and the distance from the lesser trochanter to the femoral head center (LTFHC) can be used as parameters to determine the recovery of the femoral head center (FHC) during hip arthroplasty. It is necessary to undertake an anatomical study concerning the correlations between the greater trochanter (GT), the lesser trochanter (LT), and the FHC using data obtained from the 3D-CT reconstruction method. METHODS: The study comprised 293 patients (151 males and 142 females), with an average age of 65.06 years. The femoral head diameter(FHD), the linear distance from FHC to GT (GTFHC), and the linear distance from FHC to LT(LTFHC) were all measured and recorded data. The correlation between FHD with LTFHC and GTFHC was assessed using Pearson correlation coefficients, and the ratio of LTFHC and GTFHC to FHD was calculated from this ratio. All measured parameters were compared between the left and right sides and the sexes of the participants. RESULTS: The average ratios of GTFHC/FHD and LTFHC/FHD were 0.99 and 0.95, respectively .96% of the LTFHC had absolute lateral differences of < 4 mm . 92% of the GTFHC had absolute lateral differences of < 4 mm. CONCLUSION: LTFHC and GTFHC are reliable reference parameters for preoperative planning and reconstruction of FHC of hip arthroplasty. The ratio displayed in this research may yield insight into a practical and straightforward method for orthopedic surgeons to perform hip arthroplasty in patients with femoral neck fractures. Ratios from studies based on the same race may be desirable for future work.

Interaction Between Sympk and Oct4 Promotes Mouse Embryonic Stem Cell Proliferation.

The scaffold protein Symplekin (Sympk) is involved in cytoplasmic RNA polyadenylation, transcriptional modulation, and the regulation of epithelial differentiation and proliferation via tight junctions. It is highly expressed in embryonic stem cells (ESCs), in which its role remains unknown. In this study, we found Sympk overexpression in mouse ESCs significantly increased colony formation, and Sympk deletion via CRISPR/Cas9 decreased colony formation. Sympk promoted ESC growth and its overexpression sustained ESC pluripotency, as assessed by teratoma and chimeric mouse formation. Genomic stability was preserved in these cells after long-term passage. The domain of unknown function 3453 (DUF3453) in Sympk was required for its interaction with the key pluripotent factor Oct4, and its depletion led to impaired colony formation. Sympk activated proliferation-related genes and suppressed differentiation-related genes. Our results indicate that Sympk interacts with Oct4 to promote self-renewal and pluripotency in ESCs and preserves genome integrity; accordingly, it has potential value for stem cell therapies. Stem Cells 2019;37:743-753.

Ccndbp1 is a new positive regulator of skeletal myogenesis.

Skeletal myogenesis is a multistep process in which basic helix-loop-helix (bHLH) transcription factors, such as MyoD (also known as MyoD1), bind to E-boxes and activate downstream genes. Ccndbp1 is a HLH protein that lacks a DNA-binding region, and its function in skeletal myogenesis is currently unknown. We generated Ccndbp1-null mice by using CRISPR-Cas9. Notably, in Ccndbp1-null mice, the cross sectional area of the skeletal tibialis anterior muscle was smaller, and muscle regeneration ability and grip strength were impaired, compared with those of wild type. This phenotype resembled that of myofiber hypotrophy in some human myopathies or amyoplasia. Ccndbp1 expression was upregulated during C2C12 myogenesis. Ccndbp1 overexpression promoted myogenesis, whereas knockdown of Ccndbp1 inhibited myogenic differentiation. Co-transfection of Ccndbp1 with MyoD and/or E47 (encoded by TCF3) significantly enhanced E-box-dependent transcription. Furthermore, Ccndbp1 physically associated with MyoD but not E47. These data suggest that Ccndbp1 regulates muscle differentiation by interacting with MyoD and enhancing its binding to target genes. Our study newly identifies Ccndbp1 as a positive modulator of skeletal myogenic differentiation in vivo and in vitro, providing new insights in order to decipher the complex network involved in skeletal myogenic development and related diseases.

Effects of OsCDPK1 on the Structure and Physicochemical Properties of Starch in Developing Rice Seeds.

Overexpression of a constitutively active truncated form of OsCDPK1 (OEtr) in rice produced smaller seeds, but a double-stranded RNA gene-silenced form of OsCDPK1 (Ri) yielded larger seeds, suggesting that OsCDPK1 plays a functional role in rice seed development. In the study presented here, we propose a model in which OsCDPK1 plays key roles in negatively controlling the grain size, amylose content, and endosperm appearance, and also affects the physicochemical properties of the starch. The dehulled transgenic OEtr grains were smaller than the dehulled wild-type grains, and the OEtr endosperm was opaque and had a low amylose content and numerous small loosely packed polyhedral starch granules. However, the OEtr grain sizes and endosperm appearances were not affected by temperature, which ranged from low (22 °C) to high (31 °C) during the grain-filling phase. In contrast, the transgenic Ri grains were larger, had higher amylose content, and had more transparent endosperms filled with tightly packed polyhedral starch granules. This demonstrates that OsCDPK1 plays a novel functional role in starch biosynthesis during seed development and affects the transparent appearance of the endosperm. These results improve our understanding of the molecular mechanisms through which the grain-filling process occurs in rice.

The identification of human aldo-keto reductase AKR7A2 as a novel cytoglobin-binding partner.

Cytoglobin (CYGB), a member of the globin family, is thought to protect cells from reactive oxygen and nitrogen species and deal with hypoxic conditions and oxidative stress. However, its molecular mechanisms of action are not clearly understood. Through immunoprecipitation combined with a two-dimensional electrophoresis-mass spectrometry assay, we identified a CYGB interactor: aldo-keto reductase family 7 member A2 (AKR7A2). The interaction was further confirmed using yeast two-hybrid and co-immunoprecipitation assays. Our results show that AKR7A2 physically interacts with CYGB.

A proteomic analysis of acute leukemia cells treated with 9-hydroxyoctadecadienoic acid.

BACKGROUND: 9s-hydroxy-octadecadienoic acid (9S-HOD), one of the natural products of linoleic acid oxygenation by 15-lipoxygenase (15-LOX), has been found to have anti-tumor properties in vitro and in vivo. The present study therefore investigated whether 9S-HOD affects acute leukemia HL-60 cells. METHODS: The cytotoxicity of 9S-HOD in HL-60 with or without the presence of fetal bovine serum (FBS) in the culture media was tested using cell viability assays and flow cytometry. To explore the mechanism of its anti-tumor activity by 9S-HOD, we used a proteomic analysis to identify HL-60 cells protein profiles, based on two-dimensional gel electrophoresis (2-DE) and mass spectrometry (MS) identification. RESULTS: 9S-HOD exerted cytotoxicity efficacy and induced apoptosis in HL-60 cells, and the cytotoxicity was largely attenuated by the presence of FBS in culture media. The proteomic results revealed that 9S-HOD remarkably altered the abundance of 23 proteins that were involved in mRNA metabolic process, protein binding, DNA replication and apoptosis. CONCLUSIONS: Our results indicated that 9S-HOD exerts cytotoxicity in HL-60 cells by affecting several pathways.

The causal association between type 2 diabetes and spinal stenosis: A Mendelian randomization analysis.

Spinal stenosis is a prevalent degenerative spinal disease and one of the main causes of pain and dysfunction in older adults. Substantial evidence indicates a potentially relevant association between type 2 diabetes mellitus (T2DM) and spinal stenosis. However, the causality between these 2 disorders remains unclear. Therefore, we intended to elucidate this relationship using Mendelian Randomization (MR) analysis in this study. Based on genome-wide association study (GWAS) data on T2DM and spinal stenosis, we performed a bidirectional 2-sample MR analysis to evaluate the causality of T2DM and spinal stenosis. We assessed heterogeneity using Cochran's Q statistic and horizontal pleiotropy using the MR-Egger-intercept. "Leave-one-out" analysis was performed to determine the reliability of causal relationships. In addition, we conducted multivariate MR to clarify the direct influence of T2DM on spinal stenosis after accounting for the effect of body mass index (BMI) on spinal stenosis. Our results indicated that Individuals with T2DM had a heightened risk of spinal stenosis (odds ratio [OR]: 1.050; 95% CI: 1.004-1.098, P = .031). Moreover, no reverse causality existed between T2DM and spinal stenosis. The results of the sensitivity analysis suggest that causality is steady and robust. Multivariate MR results demonstrated that the causality of T2DM on spinal stenosis was not related to BMI (OR, 1.047; 95% CI: 1.003-1.093; P = .032). MR analyses demonstrated a possible positive causal relationship between T2DM and spinal stenosis and that this causality was unrelated to BMI.

Minimally Invasive Surgical Therapies for Ureteral Polyps: A Systematic Review.

Background: Ureteral polyps are rare benign ureteral tumor. No guideline recommends that open or minimally invasive surgery is best for treating ureteral polyps. This article aims to provide a comprehensive review of the minimally invasive techniques currently available for treating ureteral polyps. Materials and Methods: We performed a comprehensive search of articles published in PubMed, using the keywords "ureteral" and "polyp," or "polyps." Results: A total of 275 studies were obtained from the literature search but 96 articles were excluded. Conclusions: Several minimally invasive approaches were developed with the advancement of medical technology, including endoscopic, laparoscopic, and robotic approaches; however, the best surgical technique was yet to be decided. Due to the advantages and disadvantages of these approaches, the best surgical approach should be tailored to each patient's needs and the surgeon's preferences and experience.

An integrated machine learning framework for developing and validating a diagnostic model of major depressive disorder based on interstitial cystitis-related genes.

BACKGROUND: Major depressive disorder (MDD) and interstitial cystitis (IC) are two highly debilitating conditions that often coexist with reciprocal effect, significantly exacerbating patients' suffering. However, the molecular underpinnings linking these disorders remain poorly understood. METHODS: Transcriptomic data from GEO datasets including those of MDD and IC patients was systematically analyzed to develop and validate our model. Following removal of batch effect, differentially expressed genes (DEGs) between respective disease and control groups were identified. Shared DEGs of the conditions then underwent functional enrichment analyses. Additionally, immune infiltration analysis was quantified through ssGSEA. A diagnostic model for MDD was constructed by exploring 113 combinations of 12 machine learning algorithms with 10-fold cross-validation on the training sets following by external validation on test sets. Finally, the "Enrichr" platform was utilized to identify potential drugs for MDD. RESULTS: Totally, 21 key genes closely associated with both MDD and IC were identified, predominantly involved in immune processes based on enrichment analyses. Immune infiltration analysis revealed distinct profiles of immune cell infiltration in MDD and IC compared to healthy controls. From these genes, a robust 11-gene (ABCD2, ATP8B4, TNNT1, AKR1C3, SLC26A8, S100A12, PTX3, FAM3B, ITGA2B, OLFM4, BCL7A) diagnostic signature was constructed, which exhibited superior performance over existing MDD diagnostic models both in training and testing cohorts. Additionally, epigallocatechin gallate and 10 other drugs emerged as potential targets for MDD. CONCLUSION: Our work developed a diagnostic model for MDD employing a combination of bioinformatic techniques and machine learning methods, focusing on shared genes between MDD and IC.

An Integrated Machine Learning Framework Identifies Prognostic Gene Pair Biomarkers Associated with Programmed Cell Death Modalities in Clear Cell Renal Cell Carcinoma.

BACKGROUND: Clear cell renal cell carcinoma (ccRCC) is a common and lethal urological malignancy for which there are no effective personalized therapeutic strategies. Programmed cell death (PCD) patterns have emerged as critical determinants of clinical prognosis and immunotherapy responses. However, the actual clinical relevance of PCD processes in ccRCC is still poorly understood. METHODS: We screened for PCD-related gene pairs through single-sample gene set enrichment analysis (ssGSEA), consensus cluster analysis, and univariate Cox regression analysis. A novel machine learning framework incorporating 12 algorithms and 113 unique combinations were used to develop the cell death-related gene pair score (CDRGPS). Additionally, a radiomic score (Rad_Score) derived from computed tomography (CT) image features was used to classify the CDRGPS status as high or low. Finally, we conclusively verified the function of PRSS23 in ccRCC. RESULTS: The CDRGPS was developed through an integrated machine learning approach that leveraged 113 algorithm combinations. CDRGPS represents an independent prognostic biomarker for overall survival and demonstrated consistent performance between training and external validation cohorts. Moreover, CDRGPS showed better prognostic accuracy compared to seven previously published cell death-related signatures. In addition, patients classified as high-risk by CDRGPS exhibited increased responsiveness to tyrosine kinase inhibitors (TKIs), mammalian Target of Rapamycin (mTOR) inhibitors, and immunotherapy. The Rad_Score demonstrated excellent discrimination for predicting high versus low CDRGPS status, with an area under the curve (AUC) value of 0.813 in the Cancer Imaging Archive (TCIA) database. PRSS23 was identified as a significant factor in the metastasis and immune response of ccRCC, thereby validating experimental in vitro results. CONCLUSIONS: CDRGPS is a robust and non-invasive tool that has the potential to improve clinical outcomes and enable personalized medicine in ccRCC patients.

A new perspective: deciphering the aberrance and clinical implication of disulfidptosis signatures in clear cell renal cell carcinoma.

Recent research has discovered disulfidptosis as a form of programmed cell death characterized by disulfide stress. However, its significance in clear cell renal cell carcinoma (ccRCC) remains unclear. To investigate this, data from The Cancer Genome Atlas were collected and used to identify ccRCC subgroups. Unsupervised clustering was employed to determine ccRCC heterogeneity. The mutation landscape and immune microenvironment of the subgroups were analyzed. The Disulfidptosis-Related Score was calculated using the LASSO-penalized Cox regression algorithm. The E-MATB-1980 cohort was used to validate the signature. The role of SLC7A11 in ccRCC metastasis was explored using western blotting and Transwell assays. Disulfidptosis-related genes are commonly downregulated in cancers and are linked to hypermethylation and copy number variation. The study revealed that ccRCC is divided into two sub-clusters: the disulfidptosis-desert sub-cluster, which is associated with a poor prognosis, a higher mutation frequency, and an immunosuppressive microenvironment. A 14-gene prognostic model was developed using differentially expressed genes and was validated in the E-MATB-1980 cohort. The low-risk group demonstrated longer overall and disease-free survival and responded better to targeted immunotherapy. Results from in vitro experiments identified SLC7A11 as a key participant in ccRCC metastasis.

Unveiling potential drug targets for hyperparathyroidism through genetic insights via Mendelian randomization and colocalization analyses.

Hyperparathyroidism (HPT) manifests as a complex condition with a substantial disease burden. While advances have been made in surgical interventions and non-surgical pharmacotherapy for the management of hyperparathyroidism, radical options to halt underlying disease progression remain lacking. Identifying putative genetic drivers and exploring novel drug targets that can impede HPT progression remain critical unmet needs. A Mendelian randomization (MR) analysis was performed to uncover putative therapeutic targets implicated in hyperparathyroidism pathology. Cis-expression quantitative trait loci (cis-eQTL) data serving as genetic instrumental variables were obtained from the eQTLGen Consortium and Genotype-Tissue Expression (GTEx) portal. Hyperparathyroidism summary statistics for single nucleotide polymorphism (SNP) associations were sourced from the FinnGen study (5590 cases; 361,988 controls). Colocalization analysis was performed to determine the probability of shared causal variants underlying SNP-hyperparathyroidism and SNP-eQTL links. Five drug targets (CMKLR1, FSTL1, IGSF11, PIK3C3 and SLC40A1) showed significant causation with hyperparathyroidism in both eQTLGen and GTEx cohorts by MR analysis. Specifically, phosphatidylinositol 3-kinase catalytic subunit type 3 (PIK3C3) and solute carrier family 40 member 1 (SLC40A1) showed strong evidence of colocalization with HPT. Multivariable MR and Phenome-Wide Association Study analyses indicated these two targets were not associated with other traits. Additionally, drug prediction analysis implies the potential of these two targets for future clinical applications. This study identifies PIK3C3 and SLC40A1 as potential genetically proxied druggable genes and promising therapeutic targets for hyperparathyroidism. Targeting PIK3C3 and SLC40A1 may offer effective novel pharmacotherapies for impeding hyperparathyroidism progression and reducing disease risk. These findings provide preliminary genetic insight into underlying drivers amenable to therapeutic manipulation, though further investigation is imperative to validate translational potential from preclinical models through clinical applications.

Leveraging cell death patterns to predict metastasis in prostate adenocarcinoma and targeting PTGDS for tumor suppression.

Metastasis is the major cause of treatment failure in patients with prostate adenocarcinoma (PRAD). Diverse programmed cell death (PCD) patterns play an important role in tumor metastasis and hold promise as predictive indicators for PRAD metastasis. Using the LASSO Cox regression method, we developed PCD score (PCDS) based on differentially expressed genes (DEGs) associated with PCD. Clinical correlation, external validation, functional enrichment analysis, mutation landscape analysis, tumor immune environment analysis, and immunotherapy analysis were conducted. The role of Prostaglandin D2 Synthase (PTGDS) in PRAD was examined through in vitro experiments, single-cell, and Mendelian randomization (MR) analysis. PCDS is elevated in patients with higher Gleason scores, higher T stage, biochemical recurrence (BCR), and higher prostate-specific antigen (PSA) levels. Individuals with higher PCDS are prone to metastasis, metastasis after BCR, BCR, and castration resistance. Moreover, PRAD patients with low PCDS responded positively to immunotherapy. Random forest analysis and Mendelian randomization analysis identified PTGDS as the top gene associated with PRAD metastasis and in vitro experiments revealed that PTGDS was considerably downregulated in PRAD cells against normal prostate cells. Furthermore, the overexpression of PTGDS was found to suppress the migration, invasion, proliferationof DU145 and LNCaP cells. To sum up, PCDS may be a useful biomarker for forecasting the possibility of metastasis, recurrence, castration resistance, and the efficacy of immunotherapy in PRAD patients. Additionally, PTGDS was identified as a viable therapeutic target for the management of PRAD.

Deep Learning-Based Surgical Treatment Recommendation and Nonsurgical Prognosis Status Classification for Scaphoid Fractures by Automated X-ray Image Recognition.

Biomedical information retrieval for diagnosis, treatment and prognosis has been studied for a long time. In particular, image recognition using deep learning has been shown to be very effective for cancers and diseases. In these fields, scaphoid fracture recognition is a hot topic because the appearance of scaphoid fractures is not easy to detect. Although there have been a number of recent studies on this topic, no studies focused their attention on surgical treatment recommendations and nonsurgical prognosis status classification. Indeed, a successful treatment recommendation will assist the doctor in selecting an effective treatment, and the prognosis status classification will help a radiologist recognize the image more efficiently. For these purposes, in this paper, we propose potential solutions through a comprehensive empirical study assessing the effectiveness of recent deep learning techniques on surgical treatment recommendation and nonsurgical prognosis status classification. In the proposed system, the scaphoid is firstly segmented from an unknown X-ray image. Next, for surgical treatment recommendation, the fractures are further filtered and recognized. According to the recognition result, the surgical treatment recommendation is generated. Finally, even without sufficient fracture information, the doctor can still make an effective decision to opt for surgery or not. Moreover, for nonsurgical patients, the current prognosis status of avascular necrosis, non-union and union can be classified. The related experimental results made using a real dataset reveal that the surgical treatment recommendation reached 80% and 86% in accuracy and AUC (Area Under the Curve), respectively, while the nonsurgical prognosis status classification reached 91% and 96%, respectively. Further, the methods using transfer learning and data augmentation can bring out obvious improvements, which, on average, reached 21.9%, 28.9% and 5.6%, 7.8% for surgical treatment recommendations and nonsurgical prognosis image classification, respectively. Based on the experimental results, the recommended methods in this paper are DenseNet169 and ResNet50 for surgical treatment recommendation and nonsurgical prognosis status classification, respectively. We believe that this paper can provide an important reference for future research on surgical treatment recommendation and nonsurgical prognosis classification for scaphoid fractures.

Prioritized Subnet Sampling for Resource-Adaptive Supernet Training.

A resource-adaptive supernet adjusts its subnets for inference to fit the dynamically available resources. In this paper, we propose prioritized subnet sampling to train a resource-adaptive supernet, termed PSS-Net. We maintain multiple subnet pools, each of which stores the information of substantial subnets with similar resource consumption. Considering a resource constraint, subnets conditioned on this resource constraint are sampled from a pre-defined subnet structure space and high-quality ones will be inserted into the corresponding subnet pool. Then, the sampling will gradually be prone to sampling subnets from the subnet pools. Moreover, the one with a better performance metric is assigned with higher priority to train our PSS-Net, if sampling is from a subnet pool. At the end of training, our PSS-Net retains the best subnet in each pool to entitle a fast switch of high-quality subnets for inference when the available resources vary. Experiments on ImageNet using MobileNet-V1/V2 and ResNet-50 show that our PSS-Net can well outperform state-of-the-art resource-adaptive supernets. Our project is publicly available at https://github.com/chenbong/PSS-Net.

Training Compact CNNs for Image Classification Using Dynamic-Coded Filter Fusion.

The mainstream approach for filter pruning is usually either to force a hard-coded importance estimation upon a computation-heavy pretrained model to select "important" filters, or to impose a hyperparameter-sensitive sparse constraint on the loss objective to regularize the network training. In this paper, we present a novel filter pruning method, dubbed dynamic-coded filter fusion (DCFF), to derive compact CNNs in a computation-economical and regularization-free manner for efficient image classification. Each filter in our DCFF is first given an inter-similarity distribution with a temperature parameter as a filter proxy, on top of which, a fresh Kullback-Leibler divergence based dynamic-coded criterion is proposed to evaluate the filter importance. In contrast to simply keeping high-score filters in other methods, we propose the concept of filter fusion, i.e., the weighted averages using the assigned proxies, as our preserved filters. We obtain a one-hot inter-similarity distribution as the temperature parameter approaches infinity. Thus, the relative importance of each filter can vary along with the training of the compact CNN, leading to dynamically changeable fused filters without both the dependency on the pretrained model and the introduction of sparse constraints. Extensive experiments on classification benchmarks demonstrate the superiority of our DCFF over the compared counterparts. For example, our DCFF derives a compact VGGNet-16 with only 72.77M FLOPs and 1.06M parameters while reaching top-1 accuracy of 93.47% on CIFAR-10. A compact ResNet-50 is obtained with 63.8% FLOPs and 58.6% parameter reductions, retaining 75.60% top-1 accuracy on ILSVRC-2012. Our code, narrower models and training logs are available at https://github.com/lmbxmu/DCFF.

Distinct dosage compensations of ploidy-sensitive and -insensitive X chromosome genes during development and in diseases.

The active X chromosome in mammals is upregulated to balance its dosage to autosomes during evolution. However, it is elusive why the known dosage compensation machinery showed uneven and small influence on X genes. Here, based on >20,000 transcriptomes, we identified two X gene groups (ploidy-sensitive [PSX] and ploidy-insensitive [PIX]), showing distinct but evolutionarily conserved dosage compensations (termed XAR). We demonstrated that XAR-PIX was downregulated whereas XAR-PSX upregulated at both RNA and protein levels across cancer types, in contrast with their trends during stem cell differentiation. XAR-PIX, but not XAR-PSX, was lower and correlated with autoantibodies and inflammation in patients of lupus, suggesting that insufficient dosage of PIX genes contribute to lupus pathogenesis. We further identified and experimentally validated two XAR regulators, TP53 and ATRX. Collectively, we provided insights into X dosage compensation in mammals and demonstrated different regulation of PSX and PIX and their pathophysiological roles in human diseases.

1xN Pattern for Pruning Convolutional Neural Networks.

Though network pruning receives popularity in reducing the complexity of convolutional neural networks (CNNs), it remains an open issue to concurrently maintain model accuracy as well as achieve significant speedups on general CPUs. In this paper, we propose a novel 1×N pruning pattern to break this limitation. In particular, consecutive N output kernels with the same input channel index are grouped into one block, which serves as a basic pruning granularity of our pruning pattern. Our 1×N pattern prunes these blocks considered unimportant. We also provide a workflow of filter rearrangement that first rearranges the weight matrix in the output channel dimension to derive more influential blocks for accuracy improvements and then applies similar rearrangement to the next-layer weights in the input channel dimension to ensure correct convolutional operations. Moreover, the output computation after our 1×N pruning can be realized via a parallelized block-wise vectorized operation, leading to significant speedups on general CPUs. The efficacy of our pruning pattern is proved with experiments on ILSVRC-2012. For example, given the pruning rate of 50% and N=4, our pattern obtains about 3.0% improvements over filter pruning in the top-1 accuracy of MobileNet-V2. Meanwhile, it obtains 56.04ms inference savings on Cortex-A7 CPU over weight pruning. Our project is made available at https://github.com/lmbxmu/1xN.