Convex Calibrated Surrogates for the Multi-Label F-Measure.
Ontology highlight
ABSTRACT:
SUBMITTER: Zhang M
PROVIDER: S-EPMC8276679 | biostudies-literature |
REPOSITORIES: biostudies-literature
Similar Datasets
Project description:A common way to learn about a system's properties is to analyze temporal fluctuations in associated variables. However, conclusions based on fluctuations from a single entity can be misleading when used without proper reference to other comparable entities or when examined only on one timescale. Here we introduce a method that uses predictions from a fluctuation scaling law as a benchmark for the observed standard deviations. Differences from the benchmark (residuals) are aggregated across multiple timescales using Principal Component Analysis to reduce data dimensionality. The first component score is a calibrated measure of fluctuations-the reactivity RA of a given entity. We apply our method to activity records from the media industry using data from the Event Registry news aggregator-over 32M articles on selected topics published by over 8000 news outlets. Our approach distinguishes between different news outlet reporting styles: high reactivity points to activity fluctuations larger than expected, reflecting a bursty reporting style, whereas low reactivity suggests a relatively stable reporting style. Combining our method with the political bias detector Media Bias/Fact Check we quantify the relative reporting styles for different topics of mainly US media sources grouped by political orientation. The results suggest that news outlets with a liberal bias tended to be the least reactive while conservative news outlets were the most reactive.
| S-EPMC7691371 | biostudies-literature
Project description:A novel on-the-fly calibration method of optical tweezers is presented, which enables in situ control and measure of absolute temperature and viscosity at nanoscale dimensions. Such noncontact measurement and control at the nanoscale are challenging as the present techniques only provide off-line measurements that do not provide absolute values. Additionally, some of the present methods have a low spatial resolution. We simultaneously apply the high temporal sensitivity of position autocorrelation and equipartition theorem to precisely measure and control in situ temperature and the corresponding microrheological property around the focal volume of the trap at high spatial resolution. The femtosecond optical tweezers (FOTs) use a single-beam high repetition rate laser for optical trapping to result in finer temperature gradients in comparison to the continuous-wave laser tweezers. Such finer temperature gradients are due to the additional nonlinear optical (NLO) phenomena occurring only at the nanoscale focal plane of the FOTs. Because NLO processes are laser peak power-dependent, they promote an effective study of physical properties occurring only at the focal plane. Using FOTs at optically benign near-infrared wavelengths, we demonstrate microrheological control and measurement in water by adding a highly absorbing yet low fluorescent dye (IR780).
| S-EPMC6645231 | biostudies-literature
Project description:With the aim of supporting ecological analyses in butterflies, the third most species-rich superfamily of Lepidoptera, this paper presents the first time-calibrated phylogeny of all 496 extant butterfly species in Europe, including 18 very localised endemics for which no public DNA sequences had been available previously. It is based on a concatenated alignment of the mitochondrial gene COI and up to eleven nuclear gene fragments, using Bayesian inferences of phylogeny. To avoid analytical biases that could result from our region-focussed sampling, our European tree was grafted upon a global genus-level backbone butterfly phylogeny for analyses. In addition to a consensus tree, the posterior distribution of trees and the fully concatenated alignment are provided for future analyses. Altogether a complete phylogenetic framework of European butterflies for use by the ecological and evolutionary communities is presented.
| S-EPMC7289901 | biostudies-literature
Project description:A convex code is a binary code generated by the pattern of intersections of a collection of open convex sets in some Euclidean space. Convex codes are relevant to neuroscience as they arise from the activity of neurons that have convex receptive fields. In this paper, we use algebraic methods to determine if a code is convex. Specifically, we use the neural ideal of a code, which is a generalization of the Stanley-Reisner ideal. Using the neural ideal together with its standard generating set, the canonical form, we provide algebraic signatures of certain families of codes that are non-convex. We connect these signatures to the precise conditions on the arrangement of sets that prevent the codes from being convex. Finally, we also provide algebraic signatures for some families of codes that are convex, including the class of intersection-complete codes. These results allow us to detect convexity and non-convexity in a variety of situations, and point to some interesting open questions.
| S-EPMC6750060 | biostudies-literature
Project description:Monitoring protected areas (PAs) is essential for systematic evaluation of their effectiveness in terms of habitat protection, preservation and representativeness. This study illustrates how the use of species distribution models that combine remote sensing data and information about biodiversity surrogates can contribute to develop a systematic protocol for monitoring PAs. In particular, we assessed the effectiveness of the Natura 2000 (N2000) network, for conserving and preserving the representativeness of seven raptor species in a highly-dynamic landscape in northwest Spain between 2001 and 2014. We also evaluated the cost-effectiveness of the N2000 network by using the total area under protection as a proxy for conservation costs. Overall, the N2000 network was found to poorly represent the habitats of the raptor species. Despite the low representativeness, this network showed a high degree of effectiveness due to increased overall habitat availability for generalist and forest specialist species between 2001 and 2014. Nevertheless, additional protected areas should be established in the near future to increase their representativeness, and thus ensure the protection of open-habitat specialist species and their priority habitats. In addition, proactive conservation measures in natural and semi-natural ecosystems (in particular, montane heathlands) will be essential for long-term protection of Montagu's harrier (species listed in the Annex I of the Bird Directive), and thus complying with the current European Environmental Legislation. This study sheds light on how the development and application of new protected area indices based on the combined use of freely-available satellite data and species distribution models may contribute substantially to the cost-efficiency of the PA monitoring systems, and to the 'Fitness Check' process of EU Nature Directives.
| S-EPMC5524325 | biostudies-literature
Project description:BackgroundMagnetic Resonance Spectroscopy (MRS) can measure in vivo brain tissue metabolism that exhibits unique biochemical characteristics in brain tumors. For clinical application, an efficient and versatile quantification method of MRS would be an important tool for medical research, particularly for exploring the scientific problem of tumor monitoring. The objective of our study is to propose an automated MRS quantitative approach and assess the feasibility of this approach for glioma grading, prognosis and boundary detection.MethodsAn automated quantitative approach based on a convex envelope (AQoCE) is proposed in this paper, including preprocessing, convex-envelope based baseline fitting, bias correction, sectional baseline removal, and peak detection, in a total of 5 steps. Some metabolic ratios acquired by this quantification are selected for statistical analysis. An independent sample t-test and the Kruskal-Wallis test are used for distinguishing low-grade gliomas (LGG) and high-grade gliomas (HGG) and for detecting the tumor, peritumoral and contralateral areas, respectively. Seventy-eight cases of pre-operative brain gliomas with pathological reports are included in this study.ResultsCho/NAA, Cho/Cr and Lip-Lac/Cr (LL/Cr) calculated by AQoCE in the tumor area differ significantly between LGG and HGG, with p≤0.005. Using logistic regression combining Cho/NAA, Cho/Cr and LL/Cr to generate a ROC curve, AQoCE achieves a sensitivity of 92.9%, a specificity of 72.2%, and an area under ROC curve (AUC) of 0.860. Moreover, both Cho/NAA and Cho/Cr in the AQoCE approach show a significant difference (p≤0.019) between tumoral, peritumoral, and contralateral areas. The comparison between the results of AQoCE and Siemens MRS processing software are also discussed in this paper.ConclusionsThe AQoCE approach is an automated method of residual water removal and metabolite quantification. It can be applied to multi-voxel 1H-MRS for evaluating brain glioma grading and demonstrating characteristics of brain glioma metabolism. It can also detect infiltration in the peritumoral area. Under the limited clinical data used, AQoCE is significantly more versatile and efficient compared to the reference approach of Siemens.
| S-EPMC4569259 | biostudies-literature
Project description:Although deep learning (DL) has demonstrated impressive diagnostic performance for a variety of computational pathology tasks, this performance often markedly deteriorates on whole slide images (WSI) generated at external test sites. This phenomenon is due in part to domain shift, wherein differences in test-site pre-analytical variables (e.g., slide scanner, staining procedure) result in WSI with notably different visual presentations compared to training data. To ameliorate pre-analytic variances, approaches such as CycleGAN can be used to calibrate visual properties of images between sites, with the intent of improving DL classifier generalizability. In this work, we present a new approach termed Multi-Site Cross-Organ Calibration based Deep Learning (MuSClD) that employs WSIs of an off-target organ for calibration created at the same site as the on-target organ, based off the assumption that cross-organ slides are subjected to a common set of pre-analytical sources of variance. We demonstrate that by using an off-target organ from the test site to calibrate training data, the domain shift between training and testing data can be mitigated. Importantly, this strategy uniquely guards against potential data leakage introduced during calibration, wherein information only available in the testing data is imparted on the training data. We evaluate MuSClD in the context of the automated diagnosis of non-melanoma skin cancer (NMSC). Specifically, we evaluated MuSClD for identifying and distinguishing (a) basal cell carcinoma (BCC), (b) in-situ squamous cell carcinomas (SCC-In Situ), and (c) invasive squamous cell carcinomas (SCC-Invasive), using an Australian (training, n = 85) and a Swiss (held-out testing, n = 352) cohort. Our experiments reveal that MuSCID reduces the Wasserstein distances between sites in terms of color, contrast, and brightness metrics, without imparting noticeable artifacts to training data. The NMSC-subtyping performance is statistically improved as a result of MuSCID in terms of one-vs. rest AUC: BCC (0.92 vs 0.87, p = 0.01), SCC-In Situ (0.87 vs 0.73, p = 0.15) and SCC-Invasive (0.92 vs 0.82, p = 1e-5). Compared to baseline NMSC-subtyping with no calibration, the internal validation results of MuSClD (BCC (0.98), SCC-In Situ (0.92), and SCC-Invasive (0.97)) suggest that while domain shift indeed degrades classification performance, our on-target calibration using off-target tissue can safely compensate for pre-analytical variabilities, while improving the robustness of the model.
| S-EPMC9825103 | biostudies-literature
Project description:ObjectivesThe interleukin-6 receptor (IL-6R) blocker tocilizumab (TCZ) reduces inflammatory disease activity in rheumatoid arthritis (RA) but elevates lipid concentrations in some patients. We aimed to characterise the impact of IL-6R inhibition on established and novel risk factors in active RA.MethodsRandomised, multicentre, two-part, phase III trial (24-week double-blind, 80-week open-label), MEASURE, evaluated lipid and lipoprotein levels, high-density lipoprotein (HDL) particle composition, markers of coagulation, thrombosis and vascular function by pulse wave velocity (PWV) in 132 patients with RA who received TCZ or placebo.ResultsMedian total-cholesterol, low-density lipoprotein-cholesterol (LDL-C) and triglyceride levels increased in TCZ versus placebo recipients by week 12 (12.6% vs 1.7%, 28.1% vs 2.2%, 10.6% vs -1.9%, respectively; all p<0.01). There were no significant differences in mean small LDL, mean oxidised LDL or total HDL-C concentrations. However, HDL-associated serum amyloid A content decreased in TCZ recipients. TCZ also induced reductions (>30%) in secretory phospholipase A2-IIA, lipoprotein(a), fibrinogen and D-dimers and elevation of paraoxonase (all p<0.0001 vs placebo). The ApoB/ApoA1 ratio remained stable over time in both groups. PWV decreases were greater with placebo than TCZ at 12 weeks (adjusted mean difference 0.79 m/s (95% CI 0.22 to 1.35; p=0.0067)).ConclusionsThese data provide the first detailed evidence for the modulation of lipoprotein particles and other surrogates of vascular risk with IL-6R inhibition. When compared with placebo, TCZ induced elevations in LDL-C but altered HDL particles towards an anti-inflammatory composition and favourably modified most, but not all, measured vascular risk surrogates. The net effect of such changes for cardiovascular risk requires determination.
| S-EPMC4392313 | biostudies-literature
Project description:Recent years have witnessed much progress in computational modelling for protein subcellular localization. However, the existing sequence-based predictive models demonstrate moderate or unsatisfactory performance, and the gene ontology (GO) based models may take the risk of performance overestimation for novel proteins. Furthermore, many human proteins have multiple subcellular locations, which renders the computational modelling more complicated. Up to the present, there are far few researches specialized for predicting the subcellular localization of human proteins that may reside in multiple cellular compartments. In this paper, we propose a multi-label multi-kernel transfer learning model for human protein subcellular localization (MLMK-TLM). MLMK-TLM proposes a multi-label confusion matrix, formally formulates three multi-labelling performance measures and adapts one-against-all multi-class probabilistic outputs to multi-label learning scenario, based on which to further extends our published work GO-TLM (gene ontology based transfer learning model for protein subcellular localization) and MK-TLM (multi-kernel transfer learning based on Chou's PseAAC formulation for protein submitochondria localization) for multiplex human protein subcellular localization. With the advantages of proper homolog knowledge transfer, comprehensive survey of model performance for novel protein and multi-labelling capability, MLMK-TLM will gain more practical applicability. The experiments on human protein benchmark dataset show that MLMK-TLM significantly outperforms the baseline model and demonstrates good multi-labelling ability for novel human proteins. Some findings (predictions) are validated by the latest Swiss-Prot database. The software can be freely downloaded at http://soft.synu.edu.cn/upload/msy.rar.
| S-EPMC3374840 | biostudies-literature
Project description:Urdu is a widely used language in South Asia and worldwide. While there are similar datasets available in English, we created the first multi-label emotion dataset consisting of 6,043 tweets and six basic emotions in the Urdu Nastalíq script. A multi-label (ML) classification approach was adopted to detect emotions from Urdu. The morphological and syntactic structure of Urdu makes it a challenging problem for multi-label emotion detection. In this paper, we build a set of baseline classifiers such as machine learning algorithms (Random forest (RF), Decision tree (J48), Sequential minimal optimization (SMO), AdaBoostM1, and Bagging), deep-learning algorithms (Convolutional Neural Networks (1D-CNN), Long short-term memory (LSTM), and LSTM with CNN features) and transformer-based baseline (BERT). We used a combination of text representations: stylometric-based features, pre-trained word embedding, word-based n-grams, and character-based n-grams. The paper highlights the annotation guidelines, dataset characteristics and insights into different methodologies used for Urdu based emotion classification. We present our best results using micro-averaged F1, macro-averaged F1, accuracy, Hamming loss (HL) and exact match (EM) for all tested methods.
| S-EPMC9044368 | biostudies-literature