Full Dataset Research Articles

Detection of Algorithmically Generated Malicious Domain Names with Feature Fusion of Meaningful Word Segmentation and N-Gram Sequences

Domain generation algorithms (DGAs) play an important role in network attacks and can be mainly divided into two types: dictionary-based and character-based. Dictionary-based algorithmically generated domains (AGDs) are similar in composition to normal domains and are harder to detect. Although methods based on meaningful word segmentation and n-gram sequence features exhibit good detection performance for AGDs, they are inadequate for mining meaningful word features of domain names, and the performance of hybrid detection of character-based and dictionary-based AGDs needs to be further improved. Therefore, in this paper, we first describe the composition of dictionary-based AGDs using meaningful word segmentation, introduce the standard deviation to better measure the word distribution features, and construct additional 11-dimensional statistical features for word segmentation results as a supplement. Then, by combining 3-gram and 1-gram sequence features, we improve the detection performance for both character-based and dictionary-based AGDs. Finally, we perform feature fusion of the above four kinds of features to achieve an end-to-end detection method for both kinds of AGDs. Experimental results showed that our method achieved an accuracy of 97.24% on the full dataset and better accuracy and F1 values than existing methods on both dictionary-based and character-based AGD datasets.

Ground Displacements in NY Using Persistent Scatterer Interferometric Synthetic Aperture Radar and Comparison of X- and C-Band Data

In this study, we investigated the quality of Interferometric Synthetic Aperture Radar (InSAR) data to measure surface displacements in upstate New York, an area with dense vegetation, snowy winters, and strong seasonal signals. We used data from the German Space Agency’s TerraSAR-X and TanDEM-X satellites (X-band, 3.1 cm radar wavelength) as well as the European Space Agency’s Sentinel-1 satellite (C-band, 5.6 cm radar wavelength); both datasets covered a ~3-year time period from 2018 to 2021. Using persistent scatterer interferometry (PSI), we were able to observe several deforming features in the region with sub-centimeter/year deformation rates. We also examined a version of the X-band data that we spatially averaged to the same pixel size as the Sentinel-1 imagery in order to separate out the effects of wavelength and pixel size on PSI accuracy and coverage. Overall, the largest number of stable PS points was found in the full-resolution X-band data, which was followed by the C-band data and then by the downsampled X-band data. Our analysis also included a subset of snow-free imagery so that we could assess the effect that snow-covered images had on the distribution and accuracy of PS points and the resulting time series. This analysis revealed that PS populations increased by 50–60% for the snow-free data when compared with analyses using the full datasets. The average deformation rates inferred from the time series generated using only snow-free images were nearly identical to those estimated from the full time series. We assessed the accuracy of the inferred rates through comparisons between the results of different datasets and with limited ground survey data. We found that all of the inferred deformation rates from each of the datasets agreed with in situ measurements in an area of known ground subsidence above an underground salt mine in Lansing, NY. The S1 datasets, however, had higher levels of noise.

Detection of millihertz quasi-periodic oscillations in the low-mass X-ray binary 4U 1730–22 with NICER

ABSTRACT We report the discovery of millihertz quasi-periodic oscillations (mHz QPOs) from the neutron star (NS) low-mass X-ray binary 4U 1730–22 using the Neutron Star Interior Composition Explorer (NICER). After being inactive for almost 50 years, 4U 1730–22 went into outburst twice between June and August 2021, and between February and July 2022. We analyse all the NICER observations of this source, and detect mHz QPOs with a significance > 4 σ in 35 observations. The QPO frequency of the full data set ranged between ∼ 4.5 and ∼ 8.1 mHz with an average fractional rms amplitude of the order of ∼2 per cent. The X-ray colour analysis strongly suggests that 4U 1730–22 was in a soft spectral state during the QPO detections. Our findings are consistent with those reported for other sources where the mHz QPOs have been interpreted as the result of a special mode of He burning on the NS surface called marginally stable nuclear burning (MSNB). We conclude that the mHz QPOs reported in this work are also associated with the MSNB, making 4U 1730–22 the eighth source that shows this phenomenology. We discuss our findings in the context of the heat flux from the NS crust.

Concomitant use of statins and sodium-glucose co-transporter 2 inhibitors and the risk of myotoxicity reporting: A disproportionality analysis.

Recent case reports have suggested that sodium-glucose co-transporter 2 (SGLT2) inhibitors may interact with statins to increase their risk of myotoxicity. We assessed the risk of myotoxicity reporting associated with concomitant use of SGLT2 inhibitors and statins. We queried the US Food and Drug Administration Adverse Event Reporting System (FAERS) from 2013 to 2021 for reports including SGLT2 inhibitors, statins or both. We estimated several measures of disproportionate reporting of myopathy and rhabdomyolysis associated with concomitant use of SGLT2 inhibitors and statins: reporting odds ratio (ROR) with 95% confidence interval (CI), Ω shrinkage measure (safety signal if >0) and an extension of the proportional reporting ratio (PRR) (two-criteria set, safety signal if both criteria are met), using the full FAERS dataset as the reference set. In sensitivity analyses, we focussed on specific SGLT2 inhibitor-statin pairs with higher interaction potential (canagliflozin-rosuvastatin, empagliflozin-rosuvastatin) and accounted for stimulated reporting. There were 456 myopathy and 77 rhabdomyolysis reports involving both an SGLT2 inhibitor and a statin. Concomitant use of SGLT2 inhibitors and statins was not associated with an increased risk of myopathy (ROR 0.79, 95% CI 0.70 to 0.89) or rhabdomyolysis (ROR 0.58, 95% CI 0.41 to 0.83) reporting. For both outcomes, the Ω shrinkage measure was negative and only one criterion of the PRR extension was met. SGLT2 inhibitor-statin pairs with higher interaction potential yielded potential signals for rhabdomyolysis; these signals disappeared after accounting for stimulated reporting. There was no increased risk of myotoxicity reporting associated with concomitant use of SGLT2 inhibitors and statins or for specific drug pairs.

An exponential increase in QTL detection with an increased sample size.

Power analyses are often used to determine the number of animals required for a genome-wide association study (GWAS). These analyses are typically intended to estimate the sample size needed for at least 1 locus to exceed a genome-wide significance threshold. A related question that is less commonly considered is the number of significant loci that will be discovered with a given sample size. We used simulations based on a real data set that consisted of 3,173 male and female adult N/NIH heterogeneous stock rats to explore the relationship between sample size and the number of significant loci discovered. Our simulations examined the number of loci identified in subsamples of the full data set. The subsampling analysis was conducted for 4 traits with low (0.15 ± 0.03), medium (0.31 ± 0.03 and 0.36 ± 0.03), and high (0.46 ± 0.03) SNP-based heritabilities. For each trait, we subsampled the data 100 times at different sample sizes (500, 1,000, 1,500, 2,000, and 2,500). We observed an exponential increase in the number of significant loci with larger sample sizes. Our results are consistent with similar observations in human GWAS and imply that future rodent GWAS should use sample sizes that are significantly larger than those needed to obtain a single significant result.

Digital Cognitive Behavioral Therapy for Insomnia Using a Smartphone Application in China

Digital cognitive behavioral therapy for insomnia (DCBT-I) requires adaptation to different sociocultural contexts. Moreover, studies comparing DCBT-I and sleep education in the same operating interface are lacking. To investigate the efficacy of a smartphone-based Chinese culture-adapted DCBT-I application (app) for insomnia compared with sleep education using the same app. This was a single-blinded, randomized clinical trial conducted from March 2021 to January 2022. Screening and randomization were conducted at Peking University First Hospital. Follow-up visits were performed online or in the same hospital. After assessing for eligibility, eligible participants were enrolled and allocated (1:1) to DCBT-I or sleep education groups. Data were analyzed from January to February 2022. A Chinese smartphone-based app with the same interface was used in both DCBT-I and sleep education groups over 6 weeks, with 1-, 3-, and 6-month follow-ups. The primary outcome was Insomnia Severity Index (ISI) scores with the intention-to-treat principle. Secondary and exploratory outcomes included sleep diary measures; self-reported scales assessing dysfunctional beliefs about sleep, mental health, and quality of life; and smart bracelet measures. Of 82 participants (mean [SD] age, 49.67 [14.49] years; 61 [74.4%] females), with 41 randomized to sleep education and 41 randomized to DCBT-I; 77 participants completed the 6-week intervention (39 participants in the sleep education group and 38 participants in the DCBT-I group; full analysis data set) and 73 completed the 6-month follow-up (per protocol data set). Mean (SD) ISI scores in the DCBT-I group were significantly lower than those in the sleep education group after the 6-week intervention (12.7 [4.8] points vs 14.9 [5.0] points; Cohen d = 0.458; P = .048) and at the 3-month follow-up (12.1 [5.4] points vs 14.8 [5.5] points; Cohen d = 0.489; P = .04). There were significant improvements from before to after the intervention for both the sleep education and DCBT-I groups, with large effect sizes(sleep education: d = 1.13; DCBT-I: d = 1.71). Some of the sleep diary measures and self-reported scales showed more improvements in the DCBT-I group than sleep education group, such as total sleep time (mean [SD]: 3 months, 403.9 [57.6] minutes vs 363.2 [72.3] minutes; 6 months, 420.3 [58.0] minutes vs 389.7 [59.4] minutes) and sleep efficiency (mean [SD]: 3 months, 87.4% [8.3%] vs 76.7% [12.1%]; 6 months, 87.5% [8.2%] vs 78.1% [10.9%]). In this randomized clinical trial, the smartphone-based Chinese culture-adapted DCBT-I improved insomnia severity compared with sleep education. Future multicenter clinical trials with large sample sizes are needed to validate its effectiveness in the Chinese population. ClinicalTrials.gov Identifier: NCT04779372.

Identifying Hair Biomarker Candidates for Alzheimer's Disease Using Three High Resolution Mass Spectrometry-Based Untargeted Metabolomics Strategies.

High-resolution mass spectrometry (HRMS)-based untargeted metabolomics strategies have emerged as an effective tool for discovering biomarkers of Alzheimer's disease (AD). There are various HRMS-based untargeted metabolomics strategies for biomarker discovery, including the data-dependent acquisition (DDA) method, the combination of full scan and target MS/MS, and the all ion fragmentation (AIF) method. Hair has emerged as a potential biospecimen for biomarker discovery in clinical research since it might reflect the circulating metabolic profiles over several months, while the analytical performances of the different data acquisition methods for hair biomarker discovery have been rarely investigated. Here, the analytical performances of three data acquisition methods in HRMS-based untargeted metabolomics for hair biomarker discovery were evaluated. The human hair samples from AD patients (N = 23) and cognitively normal individuals (N = 23) were used as an example. The most significant number of discriminatory features was acquired using the full scan (407), which is approximately 10-fold higher than that using the DDA strategy (41) and 11% higher than that using the AIF strategy (366). Only 66% of discriminatory chemicals discovered in the DDA strategy were discriminatory features in the full scan dataset. Moreover, compared to the deconvoluted MS/MS spectra with coeluted and background ions from the AIF method, the MS/MS spectrum obtained from the targeted MS/MS approach is cleaner and purer. Therefore, an untargeted metabolomics strategy combining the full scan with the targeted MS/MS method could obtain most discriminatory features along with a high quality MS/MS spectrum for discovering the AD biomarkers.

Active feature elicitation: An unified framework.

We consider the problem of active feature elicitation in which, given some examples with all the features (say, the full Electronic Health Record), and many examples with some of the features (say, demographics), the goal is to identify the set of examples on which more information (say, lab tests) need to be collected. The observation is that some set of features may be more expensive, personal or cumbersome to collect. We propose a classifier-independent, similarity metric-independent, general active learning approach which identifies examples that are dissimilar to the ones with the full set of data and acquire the complete set of features for these examples. Motivated by four real clinical tasks, our extensive evaluation demonstrates the effectiveness of this approach. To demonstrate the generalization capabilities of the proposed approach, we consider different divergence metrics and classifiers and present consistent results across the domains.

Using ancillary data to model the terrestrial distribution of gopher frogs

AbstractAn inherent challenge in managing rare or cryptic species is data deficiency. For this reason, ancillary data is a potentially valuable resource for generating key population estimates for priority species. We compiled ancillary commensal data collected between 1982 and 2020 during surveys of gopher tortoise (Gopherus polyphemus) burrows to estimate the terrestrial distribution of gopher frogs (Rana capito) from potential breeding wetlands. Gopher frogs were detected in gopher tortoise burrows 30‒3,879 m from identified wetlands. A global model of all records from all sites indicated that the probability of a gopher frog residing in a gopher tortoise burrow declined with increasing distance from a wetland. This pattern also held for 4 of 5 sites with a sufficient number of gopher frog detections to model independently. Based on the full data set, we estimated that 50%, 90%, and 99% of gopher frog observations occurred within 392 m, 1,019 m, and 2,752 m of the nearest wetland, respectively. Our results indicate a higher proportion of gopher frogs emigrate longer distances from wetlands than was previously reported using other methods, such as radio‐telemetry. This information can directly assist with management decisions, notably the spatial extent for application of habitat management surrounding breeding wetlands. More generally, this study illustrates the capacity of ancillary data to fill data deficiencies for a rare and cryptic species and highlights the importance of these data.

Impact of Transcatheter Edge-to-Edge Mitral Valve Repair on Guideline-Directed Medical Therapy Uptitration

BackgroundGuideline-directed medical therapy (GDMT) optimization is mandatory before transcatheter edge-to-edge mitral valve repair (M-TEER) in patients with secondary mitral regurgitation (SMR) and heart failure (HF) with reduced ejection fraction (HFrEF). However, the effect of M-TEER on GDMT is unknown. ObjectivesThe authors sought to evaluate frequency, prognostic implications and predictors of GDMT uptitration after M-TEER in patients with SMR and HFrEF. MethodsThis is a retrospective analysis of prospectively collected data from the EuroSMR Registry. The primary events were all-cause death and the composite of all-cause death or HF hospitalization. ResultsAmong the 1,641 EuroSMR patients, 810 had full datasets regarding GDMT and were included in this study. GDMT uptitration occurred in 307 patients (38%) after M-TEER. Proportion of patients receiving angiotensin-converting enzyme inhibitors/angiotensin receptor blockers/angiotensin receptor-neprilysin inhibitors, beta-blockers, and mineralocorticoid receptor antagonists was 78%, 89%, and 62% before M-TEER and 84%, 91%, and 66% 6 months after M-TEER (all P < 0.001). Patients with GDMT uptitration had a lower risk of all-cause death (adjusted HR: 0.62; 95% CI: 0.41-0.93; P = 0.020) and of all-cause death or HF hospitalization (adjusted HR: 0.54; 95% CI: 0.38-0.76; P < 0.001) compared with those without. Degree of MR reduction between baseline and 6-month follow-up was an independent predictor of GDMT uptitration after M-TEER (adjusted OR: 1.71; 95% CI: 1.08-2.71; P = 0.022). ConclusionsGDMT uptitration after M-TEER occurred in a considerable proportion of patients with SMR and HFrEF and is independently associated with lower rates for mortality and HF hospitalizations. A greater decrease in MR was associated with increased likelihood for GDMT uptitration.

Differential voltage analysis for battery manufacturing process control

Voltage-based battery metrics are ubiquitous and essential in battery manufacturing diagnostics. They enable electrochemical “fingerprinting” of batteries at the end of the manufacturing line and are naturally scalable, since voltage data is already collected as part of the formation process which is the last step in battery manufacturing. Yet, despite their prevalence, interpretations of voltage-based metrics are often ambiguous and require expert judgment. In this work, we present a method for collecting and analyzing full cell near-equilibrium voltage curves for end-of-line manufacturing process control. The method builds on existing literature on differential voltage analysis (DVA or dV/dQ) by expanding the method formalism through the lens of reproducibility, interpretability, and automation. Our model revisions introduce several new derived metrics relevant to manufacturing process control, including lithium consumed during formation and the practical negative-to-positive ratio, which complement standard metrics such as positive and negative electrode capacities. To facilitate method reproducibility, we reformulate the model to account for the “inaccessible lithium problem” which quantifies the numerical differences between modeled versus true values for electrode capacities and stoichiometries. We finally outline key data collection considerations, including C-rate and charging direction for both full cell and half cell datasets, which may impact method reproducibility. This work highlights the opportunities for leveraging voltage-based electrochemical metrics for online battery manufacturing process control.

Evaluation of surface reflectance bandpass adjustment techniques

There is an increasing number of on-orbit earth observation sensors in the solar domain (visible to middle infrared). While most the sensors are designed with apparent analogous bandpasses, they often exhibit differences in terms of Relative Spectral Response (RSR). Although the differences are well characterized, there is a need for techniques to adjust for the effect of these differences on surface reflectance after accounting for calibration, directional, and atmospheric effects. This research evaluated eight bandpass adjustment techniques using synthetic and actual remote sensing observations on seven widely used sensors. The scope of the study is global.The eight evaluated techniques were: a) spectral band adjustment factor technique (SBAF, i.e., one-coefficient regression), b) linear fit technique (i.e., two-coefficient regression), c) three Adaptive SBAF techniques (ASBAF), which are defined according to three different cost functions and apply a target-dependent SBAF, d) two artificial neural network approaches using (i) a single band (ANN-1B) and (ii) all bands (ANN-NB) as inputs, and e) a Classification and Regression Tree method (CART) using all bands as inputs. All these techniques require a calibration phase, which was carried out separately using three hyperspectral data sets: (i) USGS spectral measurements, (ii) simulated reflectance by the ProSail radiative transfer model, and (iii) real-world Hyperion observations. The full data set is publically available.The evaluation using the three hyperspectral data sets alone showed that the techniques with a single band as input (SBAF, LIN, and ANN-1B) generated higher uncertainty than the techniques with all bands as input (ASBAF, ANN-N, CART). The ASBAF techniques performed well with a few differences among the three ASBAF variants. The ANN-N was the technique that performed the best. The evaluation based on actual remote sensing data sets (Landsat-8 and Sentinel-2 data were used) suggested that all techniques reduced the deviation between the two data sets. Nonetheless, the deviation between the two data sets caused by factors other than RSR differences did not permit a clearcut comparison of the techniques. Some recommendations regarding the techniques to select for a particular application are provided.Finally, this research introduces the concept of RSR difference index (RDI) to quantify the spectral differences between two spectral bands.

Automation in architecture, engineering and construction: a scientometric analysis and implications for management

PurposeMany economic, political and socio-cultural events in the 2020s have been strong headwinds for architecture, engineering and construction (AEC). Nevertheless, technological advancements (e.g. artificial intelligence (AI), big data and robotics) provide promising avenues for the development of AEC. This study aims to map the state of the literature on automation in AEC and thereby be of value not only to those researching automation and its composition of a variety of distinct technological and system classes within AEC, but also to practitioners and policymakers in shaping the future of AEC.Design/methodology/approachThis review adopts scientometric methods, which have been effective in the research of large intra and interdisciplinary domains in the past decades. The full dataset consists of 1,871 articles on automation in AEC.FindingsThis overarching scientometric review offers three interdisciplinary streams of research: technological frontiers, project monitoring and applied research in AEC. To support the scientometric analysis, the authors offer a critical integrative review of the literature to proffer a multilevel, multistage framework of automation in AEC, which demonstrates an abundance of technological paradigm discussions and the inherent need for a holistic managerial approach to automation in AEC.Originality/valueThe authors underline employee well-being, business sustainability and social growth outcomes of automation and provide several managerial implications, such as the strategic management approach, ethical management view and human resource management perspective. In doing so, the authors seek to respond to the Sustainable Development Goals proposed by the United Nations as this becomes more prevalent for the industry and all levels of society in general.

Causes of death identified in neonates enrolled through Child Health and Mortality Prevention Surveillance (CHAMPS), December 2016 -December 2021.

Each year, 2.4 million children die within their first month of life. Child Health and Mortality Prevention Surveillance (CHAMPS) established in 7 countries aims to generate accurate data on why such deaths occur and inform prevention strategies. Neonatal deaths that occurred between December 2016 and December 2021 were investigated with MITS within 24-72 hours of death. Testing included blood, cerebrospinal fluid and lung cultures, multi-pathogen PCR on blood, CSF, nasopharyngeal swabs and lung tissue, and histopathology examination of lung, liver and brain. Data collection included clinical record review and family interview using standardized verbal autopsy. The full set of data was reviewed by local experts using a standardized process (Determination of Cause of Death) to identify all relevant conditions leading to death (causal chain), per WHO recommendations. For analysis we stratified neonatal death into 24-hours of birth, early (1-<7 days) and late (7-<28 days) neonatal deaths. We analyzed 1458 deaths, 41% occurring within 24-hours, 41% early and 18% late neonatal deaths. Leading underlying causes of death were complications of intrapartum events (31%), complications of prematurity (28%), infections (17%), respiratory disorders (11%), and congenital malformations (8%). In addition to the underlying cause, 62% of deaths had additional conditions and 14% had ≥3 other conditions in the causal chain. The most common causes considering the whole causal chain were infection (40%), prematurity (32%) and respiratory distress syndrome (28%). Common maternal conditions linked to neonatal death were maternal hypertension (10%), labour and delivery complications (8%), multiple gestation (7%), placental complications (6%) obstructed labour and chorioamnionitis (5%, each). CHAMPS' findings showing the full causal chain of events that lead to death, in addition to maternal factors, highlights the complexities involved in each death along with the multiple opportunities for prevention. Highlighting improvements to prenatal and obstetric care and infection prevention are urgently needed in high-mortality settings.

Exact Gaussian processes for massive datasets via non-stationary sparsity-discovering kernels

A Gaussian Process (GP) is a prominent mathematical framework for stochastic function approximation in science and engineering applications. Its success is largely attributed to the GP’s analytical tractability, robustness, and natural inclusion of uncertainty quantification. Unfortunately, the use of exact GPs is prohibitively expensive for large datasets due to their unfavorable numerical complexity of O(N^3) in computation and O(N^2) in storage. All existing methods addressing this issue utilize some form of approximation—usually considering subsets of the full dataset or finding representative pseudo-points that render the covariance matrix well-structured and sparse. These approximate methods can lead to inaccuracies in function approximations and often limit the user’s flexibility in designing expressive kernels. Instead of inducing sparsity via data-point geometry and structure, we propose to take advantage of naturally-occurring sparsity by allowing the kernel to discover—instead of induce—sparse structure. The premise of this paper is that the data sets and physical processes modeled by GPs often exhibit natural or implicit sparsities, but commonly-used kernels do not allow us to exploit such sparsity. The core concept of exact, and at the same time sparse GPs relies on kernel definitions that provide enough flexibility to learn and encode not only non-zero but also zero covariances. This principle of ultra-flexible, compactly-supported, and non-stationary kernels, combined with HPC and constrained optimization, lets us scale exact GPs well beyond 5 million data points.

Scaling whale monitoring using deep learning: A human-in-the-loop solution for analyzing aerial datasets

To ensure effective cetacean management and conservation policies, it is necessary to collect and rigorously analyze data about these populations. Remote sensing allows the acquisition of images over large observation areas, but due to the lack of reliable automatic analysis techniques, biologists usually analyze all images by hand. In this paper, we propose a human-in-the-loop approach to couple the power of deep learning-based automation with the expertise of biologists to develop a reliable artificial intelligence assisted annotation tool for cetacean monitoring. We tested this approach to analyze a dataset of 5334 aerial images acquired in 2017 by Fisheries and Oceans Canada to monitor belugas (Delphinapterus leucas) from the threatened Cumberland Sound population in Clearwater Fjord, Canada. First, we used a test subset of photographs to compare predictions obtained by the fine-tuned model to manual annotations made by three observers, expert marine mammal biologists. With only 100 annotated images for training, the model obtained between 90% and 91.4% mutual agreement with the three observers, exceeding the minimum inter-observer agreement of 88.6% obtained between the experts themselves. Second, this model was applied to the full dataset. The predictions were then verified by an observer and compared to annotations made completely manually and independently by another observer. The annotating observer and the human-in-the-loop pipeline detected 4051 belugas in common, out of a total of 4572 detections for the observer and 4298 for our pipeline. This experiment shows that the proposed human-in-the-loop approach is suitable for processing novel aerial datasets for beluga counting and can be used to scale cetacean monitoring. It also highlights that human observers, even experienced ones, have varied detection bias, underlining the need to discuss standardization of annotation protocols.

Radiosensitivity Index is Not Fit to be Used for Dose Adjustments: A Pan-Cancer Analysis

AimsTo explore the preclinical and latest clinical evidence of the radiation sensitivity signature termed ‘radiosensitivity index’ (RSI), to assess its suitability as an input into dose-adjustment algorithms. Materials and methodsThe original preclinical test-set data from the publication where RSI was derived were collected and reanalysed by comparing the observed versus predicted survival fraction at 2 Gy (SF2). In addition, the predictive capability of RSI was also compared to random guessing. Clinical data were collected from a recently published dataset that included RSI values, overall survival outcomes, radiotherapy dose and tumour site for six cancers (glioma, triple-negative breast, endometrial, melanoma, pancreatic and lung cancer). Cox proportional hazards models were used to assess: (i) does adjusting for RSI elucidate a dose response and (ii) does an interaction between RSI and dose exist with good precision. ResultsPreclinically, RSI showed a negative correlation (Spearman's rho = –0.61) between observed and predicted SF2, which remained negative after removing leukaemia cell lines. Furthermore, random guesses showed better correlation to SF2 than RSI, 98% of the time on the full dataset and 80% after removing leukaemia cell lines. The preclinical data show that RSI does not explain the variance in SF2 better than random guessing. Clinically, a dose response was not seen after adjusting for RSI (hazard ratio = 1.00, 95% confidence interval 0.97–1.04; P = 0.876) and no evidence of an interaction between RSI and dose was found (P = 0.844). ConclusionsThese results suggest that RSI does not explain a sufficient amount of the outcome variance to be used within dose-adjustment algorithms.

Comparing the Performance of Corporate Bankruptcy Prediction Models Based on Imbalanced Financial Data

Forecasts of corporate defaults are used in various fields across the economy. Several recent studies attempt to forecast corporate bankruptcy using various machine learning techniques. We collected financial information on 13 variables of 1020 companies listed on the KOSPI and KOSDAQ to capture the possibility of corporate bankruptcy. We propose a data processing method for small-sample domestic corporate financial data. We investigate the case of random sampling of non-bankrupt companies versus sampling non-bankrupt companies based on approximate entropy and optimized threshold based on AUC to address the imbalance between the number of bankrupt companies and the number of non-bankrupt companies. We compare the performance measures of corporate bankruptcy prediction models for the small sample data structured in two ways and the full dataset. The experimental results of this study contribute to the selection of an appropriate corporate bankruptcy prediction model.

An Injection System for the CHIME/FRB Experiment

Dedicated surveys searching for fast radio bursts (FRBs) are subject to selection effects that bias the observed population of events. Software injection systems are one method of correcting for these biases by injecting a mock population of synthetic FRBs directly into the real-time search pipeline. The injected population may then be used to map intrinsic burst properties onto an expected signal-to-noise ratio (S/N), so long as telescope characteristics such as the beam model and calibration factors are properly accounted for. This paper presents an injection system developed for the Canadian Hydrogen Intensity Mapping Experiment Fast Radio Burst Project (CHIME/FRB). The system was tested to ensure high detection efficiency, and the pulse calibration method was verified. Using an injection population of ∼85,000 synthetic FRBs, we found that the correlation between fluence and S/N for injected FRBs was consistent with that of CHIME/FRB detections in the first CHIME/FRB catalog. We noted that the sensitivity of the telescope varied strongly as a function of the broadened burst width, but not as a function of the dispersion measure. We conclude that some of the machine-learning based Radio Frequency Interference mitigation methods used by CHIME/FRB can be retrained using injection data to increase sensitivity to wide events, and that planned upgrades to the presented injection system will allow for determining a more accurate CHIME/FRB selection function in the near future. We also provide the full injection data sets along with usage tutorials.

Presence of immunogenic alternatively spliced insulin gene product in human pancreatic delta cells

Aims/hypothesisTranscriptome analyses revealed insulin-gene-derived transcripts in non-beta endocrine islet cells. We studied alternative splicing of human INS mRNA in pancreatic islets.MethodsAlternative splicing of insulin pre-mRNA was determined by PCR analysis performed on human islet RNA and single-cell RNA-seq analysis. Antisera were generated to detect insulin variants in human pancreatic tissue using immunohistochemistry, electron microscopy and single-cell western blot to confirm the expression of insulin variants. Cytotoxic T lymphocyte (CTL) activation was determined by MIP-1β release.ResultsWe identified an alternatively spliced INS product. This variant encodes the complete insulin signal peptide and B chain and an alternative C-terminus that largely overlaps with a previously identified defective ribosomal product of INS. Immunohistochemical analysis revealed that the translation product of this INS-derived splice transcript was detectable in somatostatin-producing delta cells but not in beta cells; this was confirmed by light and electron microscopy. Expression of this alternatively spliced INS product activated preproinsulin-specific CTLs in vitro. The exclusive presence of this alternatively spliced INS product in delta cells may be explained by its clearance from beta cells by insulin-degrading enzyme capturing its insulin B chain fragment and a lack of insulin-degrading enzyme expression in delta cells.Conclusions/interpretationOur data demonstrate that delta cells can express an INS product derived from alternative splicing, containing both the diabetogenic insulin signal peptide and B chain, in their secretory granules. We propose that this alternative INS product may play a role in islet autoimmunity and pathology, as well as endocrine or paracrine function or islet development and endocrine destiny, and transdifferentiation between endocrine cells. INS promoter activity is not confined to beta cells and should be used with care when assigning beta cell identity and selectivity.Data availabilityThe full EM dataset is available via www.nanotomy.org (for review: http://www.nanotomy.org/OA/Tienhoven2021SUB/6126-368/). Single-cell RNA-seq data was made available by Segerstolpe et al [13] and can be found at https://sandberglab.se/pancreas. The RNA and protein sequence of INS-splice was uploaded to GenBank (BankIt2546444 INS-splice OM489474).Graphical abstract