Xiang Zhang

Bio: Xiang Zhang is an academic researcher from Baylor College of Medicine. The author has contributed to research in topics: Medicine & Computer science. The author has an hindex of 154, co-authored 1733 publications receiving 117576 citations. Previous affiliations of Xiang Zhang include University of California, Berkeley & University of Texas MD Anderson Cancer Center.

Costs associated with failure to respond to treatment among patients with rheumatoid arthritis initiating TNFi therapy: a retrospective claims analysis

Abstract: Tumor necrosis factor inhibitors (TNFi) are common second-line treatments for rheumatoid arthritis (RA) This study was designed to compare the real-world clinical and economic outcomes between patients with RA who responded to TNFi therapy and those who did not For this retrospective cohort analysis we used medical and pharmacy claims from members of 14 large US commercial health plans represented in the HealthCore Integrated Research Database Adult patients (aged ≥18 years) diagnosed with RA and initiating TNFi therapy (index date) between 1 January 2007 and 30 April 2014 were included in the study Treatment response was assessed using a previously developed and validated claims-based algorithm Patients classified as treatment responders in the 12 months postindex were matched 1:1 to nonresponders on important baseline characteristics, including sex, age, index TNFi agent, and comorbidities The matched cohorts were then compared on their all-cause and RA-related healthcare resource use, and costs were assessed from a payer perspective during the first, second, and third years postindex using parametric tests, regressions, and a nonparametric bootstrap A total of 7797 patients met the study inclusion criteria, among whom 2337 (30%) were classified as treatment responders The responders had significantly lower all-cause hospitalizations, emergency department visits, and physical/occupational therapy visits than matched nonresponders during the first-year postindex Mean total all-cause medical costs were $5737 higher for matched nonresponders, largely driven by outpatient visits and hospitalizations Mean all-cause pharmacy costs (excluding costs of biologics) were $354 higher for matched nonresponders Mean RA-related pharmacy costs (conventional synthetic and biologic drugs), however, were $8579 higher in the responder cohort, driven by higher adherence to their index TNFi agent (p < 001 for all comparisons) A similar pattern of cost differentiation was observed over years 2 and 3 of follow-up In this real-world study we found that, compared with matched nonresponders, patients who responded to TNFi treatments had lower all-cause medical, pharmacy, and total costs (excluding biologics) up to 3 years from initiation of TNFi therapy These cost differences between the two cohorts provide a considerable offset to the cost of RA medications and should encourage close monitoring of treatment response to minimize disease progression with appropriate therapy choices

Replication stress response defects are associated with response to immune checkpoint blockade in nonhypermutated cancers.

Abstract: Treatment with immune checkpoint blockade (ICB) has resulted in durable responses for a subset of patients with cancer, with predictive biomarkers for ICB response originally identified largely in ...

Genome-wide identification, phylogenetic relationships, and expression analysis of the carotenoid cleavage oxygenase gene family in pepper.

Abstract: Carotenoid cleavage oxygenases (CCOs) are a family of dioxygenases, which specifically catalyze the cleavage of conjugated double bonds in carotenoids and apocarotenoids in plants. In this study, genome-wide analysis of CCO genes in pepper plants was performed using bioinformatic methods. At least 11 members of the CCO gene family were identified in the pepper genome. Phylogenetic analysis showed that pepper and tomato CCO genes could be divided into two groups (CCDs and NCEDs). The CCD group included five sub-groups (CCD1, CCD4, CCD7, CCD8, and CCD-like). These results indicate that there is a close genetic relationship between the two species. Sequence analysis using the online tool, Multiple Expectation Maximization for Motif Elicitation (MEME), showed that the CCO proteins comprise multiple conserved motifs, with 20 to 41 amino acids. In addition, multiple cis-acting elements in the promoter of CCO genes were identified using the online tool PlantCARE, and were found to be involved in light responsiveness, plant hormone regulation, and biotic and abiotic stresses, suggesting potential roles of these proteins under different conditions. RNA-seq analysis revealed that the CCO genes exhibit distinct patterns of expression in the roots, stems, leaves, and fruit. These findings suggest that the CCO genes have important roles in the vegetative and reproductive development of pepper plants.

Localized plasmonic structured illumination microscopy with gaps in spatial frequencies

Abstract: Localized plasmonic structured illumination microscopy (LPSIM) is a super-resolution fluorescent microscopy method to image samples at a high speed with a wide field of view and low phototoxicity. Here we propose a methodology to extend the resolution capability of LPSIM by shifting spatial frequencies farther away from the diffraction-limited cutoff frequency with a plasmonic nano-array. We analyze the performance and accuracy of image reconstruction by using simulations of standard structured illumination microscopy (SIM) and blind-LPSIM. LPSIM experiments were also performed by using various LPSIM substrates and different microscope objectives. The experiments and simulations show that by shifting spatial frequencies farther away, resolution improvement can be extended up to 5 times beyond the diffraction limit with minimal deformation and artifacts in the reconstructed image.

Deep Residual Learning for Image Recognition

Abstract: Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers—8× deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation.

Very Deep Convolutional Networks for Large-Scale Image Recognition

Abstract: In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

Deep Residual Learning for Image Recognition

Abstract: Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers---8x deeper than VGG nets but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation.

Going deeper with convolutions

Abstract: We propose a deep convolutional neural network architecture codenamed Inception that achieves the new state of the art for classification and detection in the ImageNet Large-Scale Visual Recognition Challenge 2014 (ILSVRC14). The main hallmark of this architecture is the improved utilization of the computing resources inside the network. By a carefully crafted design, we increased the depth and width of the network while keeping the computational budget constant. To optimize quality, the architectural decisions were based on the Hebbian principle and the intuition of multi-scale processing. One particular incarnation used in our submission for ILSVRC14 is called GoogLeNet, a 22 layers deep network, the quality of which is assessed in the context of classification and detection.