Confidence In Predictions Research Articles

Core performance predictions in projected SPARC first-campaign plasmas with nonlinear CGYRO

This work characterizes the core transport physics of SPARC early-campaign plasmas using the PORTALS-CGYRO framework. Empirical modeling of SPARC plasmas with L-mode confinement indicates an ample window of breakeven (Q > 1) without the need of H-mode operation. Extensive modeling of multi-channel (electron energy, ion energy, and electron particle) flux-matched conditions with the nonlinear CGYRO code for turbulent transport coupled to the macroscopic plasma evolution using PORTALS reveals that the maximum fusion performance to be attained will be highly dependent on the near-edge pressure. Stiff core transport conditions are found, particularly when fusion gain approaches unity, and predicted density peaking is found to be in line with empirical databases of particle source-free H-modes. Impurity optimization is identified as a potential avenue to increase fusion performance while enabling core-edge integration. Extensive validation of the quasilinear TGLF model builds confidence in reduced-model predictions. The implications of projecting L-mode performance to high-performance and burning-plasma devices is discussed, together with the importance of predicting edge conditions.

Multi-cancer early detection (MCED) in circulating tumor DNA (ctDNA) using an ultrasensitive mutation-based amplicon next-generation sequencing (NGS) assay.

e15042 Background: Cancer remains a leading cause of mortality globally. For most cancers, routine screening options are still unavailable. Blood-based MCED has emerged as a promising strategy to bridge this gap. Mutation-based NGS in ctDNA detects highly specific cancer signals, an attribute critical for MCED assays. We developed an ultrasensitive mutation-based NGS assay optimized to detect cancer signals in 50 cancers and predict tissue-of-origin (TOO) for up to 21 cancer classes. Methods: An amplicon-based NGS assay incorporating mirror barcodes for error correction was designed to detect cancer-associated alterations (CAAs), including single nucleotide variants (SNVs), insertion-deletion mutations (indels), and viral DNA (EBV and 20 genotypes of HPV), in 50 cancers. Analytical validation was performed using reference standards and clinical validation was performed in a retrospective case-control cohort of 601 subjects comprising cell-free DNA (cfDNA) from 313 self-declared healthy donors and 288 cancer patients (30.9% early-stage, 50.3% late-stage and 9.4% hematolymphoid cancers). Tissue-of-origin (TOO) was predicted for 21 cancer classes using a weighted sum model. Where relevant, matched buffy coat genomic DNA (gDNA) was analyzed to rule out clonal hematopoiesis. Results: Analytical validation demonstrated a limit of detection (LoD) of 0.05% variant allele frequency (VAF) for indels and 0.07% VAF for SNVs in cfDNA. LoD was matched in gDNA (0.04% and 0.05% VAF for indels and SNVs respectively). In clinical validation, the assay demonstrated a specificity of 99.0% (310/313) and an overall sensitivity of 80.9% (233/288), with sensitivities of 64.0%, 91.7%, and 85.2% in early-stage, late-stage, and hematolymphoid cancers respectively. The mean number of CAAs detected in early-stage, late-stage, and hematolymphoid CAA-positive cancers was 2.1, 3.1, and 4.6. Inclusion of viral targets aided the detection of nasopharyngeal (NPC), cervical, and p16+ oropharyngeal (oPSCC) cancers. EBV DNA was detected in 91.7% of NPC cases. In cervical cancer and p16+ oPSCC, HPV DNA (genotypes 16, 18, 33 and 39) was detected in 71.4% and 90.9% of cases respectively. Conversely, none of the p16- oPSCC cancers had detectable HPV DNA. Across the 21 TOO classes, a high confidence prediction could be made in 62.4% (mean 73.6%, range 26.7%-100%) of cases with an overall accuracy of 87.7% (mean 84.9%, range 33.3%-100%). Overall TOO prediction performance was modeled in a simulated cohort based on cancer incidence in Singapore and Hong Kong, giving an accuracy of 87.3% and 88.7% respectively. Conclusions: The mutation-based NGS assay sensitively detects cancer signals and accurately predicts the tumor origin for early-stage, late-stage, and hematolymphoid cancers. Inclusion of viral targets additionally enables early detection of virus-associated cancers.

Accurately predicting rare and poorly detectable species habitat for spatial protection

Abstract With the loss of biodiversity worldwide, understanding species distribution is essential for species management, but modelling the distribution of rare and poorly detectable species can be challenging because of data gaps and observer biases. Over‐ or under‐predictions are frequent, leading to uncertainty in spatial management measures, particularly for highly mobile and data‐poor species. Here we developed a ‘Combined Model for Accurate Prediction’ to accurately predict the distribution of data‐poor and rare species. This modelling framework aims to improve the accuracy of both predicted ‘core’ and ‘unsuitable’ habitats, to help support managers with spatial protection measures. We tested the combined modelling approach on 11 data‐poor and rare diadromous fish during their at‐sea life history phase and used the combined model to analyse the adequacy of existing marine protected areas (MPAs) for these fish. The combined modelling approach modelled both ‘core’ and ‘unsuitable’ habitats with high accuracy. Of the seven diadromous species modelled, most MPAs designated to protect diadromous fish are outside their core habitats. Furthermore, when their core habitat was within an MPA, only 50% of this area was designated to protect them. These results highlight inadequate protection of the existing networks of MPAs for protected and threatened species. Synthesis and applications. Being able to accurately model species distribution is critical to reliable and transparent biodiversity and conservation assessments. By modelling accurate ‘core’ and ‘unsuitable’ habitats with models that minimise omission and commission rates respectively, conservation measures could be targeted in specific spatial areas that maximise the protection of rare and poorly detected species. This method therefore helps minimise impacts on stakeholders, while providing managers with increased confidence in the model predictions.

Exploiting Rating Prediction Certainty for Recommendation Formulation in Collaborative Filtering

Collaborative filtering is a popular recommender system (RecSys) method that produces rating prediction values for products by combining the ratings that close users have already given to the same products. Afterwards, the products that achieve the highest prediction values are recommended to the user. However, as expected, prediction estimation may contain errors, which, in the case of RecSys, will lead to either not recommending a product that the user would actually like (i.e., purchase, watch, or listen) or to recommending a product that the user would not like, with both cases leading to degraded recommendation quality. Especially in the latter case, the RecSys would be deemed unreliable. In this work, we design and develop a recommendation algorithm that considers both the rating prediction values and the prediction confidence, derived from features associated with rating prediction accuracy in collaborative filtering. The presented algorithm is based on the rationale that it is preferable to recommend an item with a slightly lower prediction value, if that prediction seems to be certain and safe, over another that has a higher value but of lower certainty. The proposed algorithm prevents low-confidence rating predictions from being included in recommendations, ensuring the recommendation quality and reliability of the RecSys.

Parabolic Modeling Forecasts of Space and Time European Hydropower Production

Renewable sources of energy production are some of the main targets today to protect the environment through reduced fossil fuel consumption and CO2 emissions. Alongside wind, solar, marine, biomass and nuclear sources, hydropower is among the oldest but still not fully explored renewable energy sources. Compared with other sources like wind and solar, hydropower is more stable and consistent, offering increased predictability. Even so, it should be analyzed considering water flow, dams capacity, climate change, irrigation, navigation, and so on. The aim of this study is to propose a forecast model of hydropower production capacity and identify long-term trends. The curve fit forecast parabolic model was applied to 33 European countries for time series data from 1990 to 2021. Space-time cube ArcGIS representation in 2D and 3D offers visualization of the prediction and model confidence rate. The quadratic trajectory fit the raw data for 14 countries, validated by visual check, and in 20 countries, validated by FMRSE 10% threshold from the maximal value. The quadratic model choice is good for forecasting future values of hydropower electric capacity in 22 countries, with accuracy confirmed by the VMRSE 10% threshold from the maximal value. Seven local outliers were identified, with only one validated as a global outlier based on the Generalized Extreme Studentized Deviate (GESD) test at a 5% maximal number of outliers and a 90% confidence level. This result achieves our objective of estimating a level with a high degree of occurrence and offering a reliable forecast of hydropower production capacity. All European countries show a growing trend in the short term, but the trends show a stagnation or decrease if policies do not consider intensive growth through new technology integration and digital adoption. Unfortunately, Europe does not have extensive growth potential compared with Asia–Pacific. Public policies must boost hybrid hydro–wind or hydro–solar systems and intensive technical solutions.

Efficacy and safety of anti-angiogenic drug monotherapy and combination therapy for ovarian cancer: a meta-analysis and trial sequential analysis of randomized controlled trials.

As the development of novel anti-angiogenic drugs and the continuous evolution of guideline recommendations, the efficacy and safety of anti-angiogenic agents in ovarian cancer (OC) remains unclear. Consequently, a meta-analysis was carried out to assess the efficacy and safety of anti-angiogenic drug monotherapy and combination therapy for OC. An exhaustive literature review was performed across multiple databases, including PubMed, Embase, Web of Science, and Cochrane, encompassing all relevant randomized controlled trials (RCTs) up until 6 April 2024. The evaluation of efficacy outcomes incorporated progression-free survival (PFS), overall survival (OS), and objective response rate (ORR). Safety was assessed through the occurrence of any grade adverse events (AEs) and grade ≥3 AEs. Synthesis of the data involved the calculation of hazard ratios (HRs), relative risks (RRs), and their corresponding 95% confidence intervals (CIs) and prediction intervals (PIs). Trial sequential analysis was executed employing TSA v0.9.5.10 Beta software, STATA 12.0, and R software 4.3.1. In this meta-analysis, 35 RCTs were included, encompassing 16,199 subjects in total. The overall analysis indicated that anti-angiogenic drug combination therapy significantly improved PFS (HR [95% CI] = 0.678 [0.606-0.759], 95% PI: 0.415-1.108), OS (HR [95% CI] = 0.917 [0.870-0.966], 95% PI: 0.851-0.984), and ORR (RR [95% CI] = 1.441 [1.287-1.614], 95% PI: 1.032-2.014), but also increased the incidence of grade ≥3 AEs (RR [95% CI] = 1.137 [1.099-1.177], 95% PI: 1.011-1.252). The analysis did not corroborate any benefit of anti-angiogenic monotherapy over placebo concerning PFS (HR [95% CI] = 0.956 [0.709-1.288], 95% PI: 0.345-2.645) and OS (HR [95% CI] = 1.039 [0.921-1.173], 95% PI: 0.824-1.331). However, it was observed that monotherapy with anti-angiogenic drugs did increase the incidence of any grade AEs (RR [95% CI] = 1.072 [1.036-1.109], 95% PI: 0.709-1.592). Our study confirmed the PFS, OS, and ORR benefits of anti-angiogenic drug combination therapy for OC patients. The efficacy results of anti-angiogenic monotherapy necessitates further evaluation as more RCTs become available. Clinicians should be vigilant of AEs when administering anti-angiogenic agents in a clinical setting.

Active Learning of Ligands That Enhance Perovskite Nanocrystal Luminescence.

Ligands play a critical role in the optical properties and chemical stability of colloidal nanocrystals (NCs), but identifying ligands that can enhance NC properties is daunting, given the high dimensionality of chemical space. Here, we use machine learning (ML) and robotic screening to accelerate the discovery of ligands that enhance the photoluminescence quantum yield (PLQY) of CsPbBr3 perovskite NCs. We developed a ML model designed to predict the relative PL enhancement of perovskite NCs when coordinated with a ligand selected from a pool of 29,904 candidate molecules. Ligand candidates were selected using an active learning (AL) approach that accounted for uncertainty quantified by twin regressors. After eight experimental iterations of batch AL (corresponding to 21 initial and 72 model-recommended ligands), the uncertainty of the model decreased, demonstrating an increased confidence in the model predictions. Feature importance and counterfactual analyses of model predictions illustrate the potential use of ligand field strength in designing PL-enhancing ligands. Our versatile AL framework can be readily adapted to screen the effect of ligands on a wide range of colloidal nanomaterials.

Confidence-aware multi-modality learning for eye disease screening

Multi-modal ophthalmic image classification plays a key role in diagnosing eye diseases, as it integrates information from different sources to complement their respective performances. However, recent improvements have mainly focused on accuracy, often neglecting the importance of confidence and robustness in predictions for diverse modalities. In this study, we propose a novel multi-modality evidential fusion pipeline for eye disease screening. It provides a measure of confidence for each modality and elegantly integrates the multi-modality information using a multi-distribution fusion perspective. Specifically, our method first utilizes normal inverse gamma prior distributions over pre-trained models to learn both aleatoric and epistemic uncertainty for uni-modality. Then, the normal inverse gamma distribution is analyzed as the Student’s t distribution. Furthermore, within a confidence-aware fusion framework, we propose a mixture of Student’s t distributions to effectively integrate different modalities, imparting the model with heavy-tailed properties and enhancing its robustness and reliability. More importantly, the confidence-aware multi-modality ranking regularization term induces the model to more reasonably rank the noisy single-modal and fused-modal confidence, leading to improved reliability and accuracy. Experimental results on both public and internal datasets demonstrate that our model excels in robustness, particularly in challenging scenarios involving Gaussian noise and modality missing conditions. Moreover, our model exhibits strong generalization capabilities to out-of-distribution data, underscoring its potential as a promising solution for multimodal eye disease screening.

Medical calculators derived synthetic cohorts: a novel method for generating synthetic patient data

This study shows that we can use synthetic cohorts created from medical risk calculators to gain insights into how risk estimations, clinical reasoning, data-driven subgrouping, and the confidence in risk calculator scores are connected. When prediction variables aren't evenly distributed in these synthetic cohorts, they can be used to group similar cases together, revealing new insights about how cohorts behave. We also found that the confidence in predictions made by these calculators can vary depending on patient characteristics. This suggests that it might be beneficial to include a "normalized confidence" score in future versions of these calculators for healthcare professionals. We plan to explore this idea further in our upcoming research.

Modelling CO2 storage in oil and gas reservoirs in the Gippsland Basin

The Gippsland Basin has world class geology for carbon capture and storage (CCS) and a long history of oil and gas production. Depleted oil and gas fields within the Gippsland Basin that are candidates for carbon dioxide (CO2) storage are in close proximity to existing infrastructure that could be repurposed as part of a CCS project. Modelling of CO2 storage in the depleted Bream oil and gas reservoir is being progressed. Bream reservoir properties are very well understood due to extensive geological and geophysical data sets available from wells and seismic data. Additionally, the field has been through three key phases of development during its production history; oil production and gas re-injection in 1988, gas cap blowdown started in 2002, and seasonal gas storage and withdrawal started in 2012 through to field shut-in in 2020. This provides a wealth of dynamic data that is used to calibrate the reservoir models to improve our confidence in the CO2 plume prediction. However, there are also challenges in modelling CO2 storage in depleted fields. Unlike saline aquifers, CO2 can be injected into a three-phase depleted reservoir that contains residual oil and gas saturation. The key aspects of our workflow to evaluate the plume behaviour are presented in this paper.

Climate change, malaria and neglected tropical diseases: a scoping review.

To explore the effects of climate change on malaria and 20 neglected tropical diseases (NTDs), and potential effect amelioration through mitigation and adaptation, we searched for papers published from January 2010 to October 2023. We descriptively synthesised extracted data. We analysed numbers of papers meeting our inclusion criteria by country and national disease burden, healthcare access and quality index (HAQI), as well as by climate vulnerability score. From 42 693 retrieved records, 1543 full-text papers were assessed. Of 511 papers meeting the inclusion criteria, 185 studied malaria, 181 dengue and chikungunya and 53 leishmaniasis; other NTDs were relatively understudied. Mitigation was considered in 174 papers (34%) and adaption strategies in 24 (5%). Amplitude and direction of effects of climate change on malaria and NTDs are likely to vary by disease and location, be non-linear and evolve over time. Available analyses do not allow confident prediction of the overall global impact of climate change on these diseases. For dengue and chikungunya and the group of non-vector-borne NTDs, the literature privileged consideration of current low-burden countries with a high HAQI. No leishmaniasis papers considered outcomes in East Africa. Comprehensive, collaborative and standardised modelling efforts are needed to better understand how climate change will directly and indirectly affect malaria and NTDs.

Toward confident prostate cancer detection using ultrasound: a multi-center study.

Deep learning-based analysis of micro-ultrasound images to detect cancerous lesions is a promising tool for improving prostate cancer (PCa) diagnosis. An ideal model should confidently identify cancer while responding with appropriate uncertainty when presented with out-of-distribution inputs that arise during deployment due to imaging artifacts and the biological heterogeneity of patients and prostatic tissue. Using micro-ultrasound data from 693 patients across 5 clinical centers who underwent micro-ultrasound guided prostate biopsy, we train and evaluate convolutional neural network models for PCa detection. To improve robustness to out-of-distribution inputs, we employ and comprehensively benchmark several state-of-the-art uncertainty estimation methods. PCa detection models achieve performance scores up to average AUROC with a 10-fold cross validation setup. Models with uncertainty estimation obtain expected calibration error scores as low as , indicating that confident predictions are very likely to be correct. Visualizations of the model output demonstrate that the model correctly identifies healthy versus malignant tissue. Deep learning models have been developed to confidently detect PCa lesions from micro-ultrasound. The performance of these models, determined from a large and diverse dataset, is competitive with visual analysis of magnetic resonance imaging, the clinical benchmark to identify PCa lesions for targeted biopsy. Deep learning with micro-ultrasound should be further studied as an avenue for targeted prostate biopsy.

Hindcasting and forecasting marsh ecomorphodynamics by integration of model with stratigraphic record

Marsh morphodynamics models are usually calibrated using sparse short-term measurements. Hence, their ability to predict long-term trajectories (centuries to millennia) is highly uncertain. Here I improved an existing model for marsh ecomorphodynamics and used it to hindcast the whole morphodynamic trajectory (2500 years) of an actual salt marsh in New England (USA). The model includes enough processes – among which tidal currents and waves, sand and mud transport, and vegetation population dynamics – to simulate a realistic marsh analogue that closely matches the modern topobathymetry and stratigraphy. Marsh trajectory is highly sensitive to initial bathymetry, mud supply from the nearshore, wave erosion processes, and vegetation dynamics. Each one of these factors is important yet difficult to calibrate on its own. Matching the long-term evolution with the stratigraphic record provides a global constraint for the processes and parameters in the model, increasing the confidence in future long-term predictions. This type of comparison complements those made at short times scales, and it should be used more routinely.

WordBlitz: An Efficient Hard-Label Textual Adversarial Attack Method Jointly Leveraging Adversarial Transferability and Word Importance

Existing textual attacks mostly perturb keywords in sentences to generate adversarial examples by relying on the prediction confidence of victim models. In practice, attackers can only access the prediction label, meaning that the victim model can easily defend against such hard-label attacks by denying access based on the attack’s frequency. In this paper, we propose an efficient hard-label attack approach, called WordBlitz. First, based on the adversarial transferability, we train a substitute model to initialize the attack parameter set, including a candidate pool and two weight tables of keywords and candidate words. Then, adversarial examples are generated and optimized under the guidance of the two weight tables. During optimization, we design a hybrid local search algorithm with word importance to find the globally optimal solution while updating the two weight tables according to the attack results. Finally, the non-adversarial text generated during perturbation optimization is added to the training of the substitute model as data augmentation to improve the adversarial transferability. Experimental results show that WordBlitz surpasses the baseline in terms of better effectiveness, higher efficiency, and lower cost. Its efficiency is especially pronounced in scenarios with broader search spaces, and its attack success rate on a Chinese dataset is higher than on baselines.

Further adoption of conservation tillage can increase maize yields in the western US Corn Belt

Conservation tillage can reduce soil erosion, increase soil health, and decrease labor and fuel input costs. Despite these benefits, potential yield impacts remain an important concern for farmers considering adoption. Previous research suggests that conservation tillage is likely to have the largest yield benefits in more arid conditions, but a lack of field-level analyses across climatic, management and soil conditions limits confidence in such predictions. Satellite imagery provides the opportunity to monitor agricultural lands at sub-field resolution across large spatial scales and wide environmental gradients. Here we investigate the maize yield impacts of conservation tillage in the semi-arid western US Corn Belt, using sub-field resolution datasets on tillage practices and crop yields derived from satellite data spanning four states (Nebraska, Kansas, South Dakota, and North Dakota) between 2008 and 2020. On these datasets, we estimate heterogenous yield outcomes for several thousand maize fields across gradients in climate, soil quality and irrigation status by using a causal forests analysis, an adaptation of the random forests machine-learning algorithm for causal inference on observational data. We find that long-term adoption of conservation tillage increased rainfed maize yields by an average of 9.9% in the region. Impacts on irrigated yields were small and not statistically significant. These results, along with an analysis of variables related to greater than average yield benefits, indicate that improved water infiltration and retention are the primary reasons for conservation tillage benefits. Despite yield benefits, many fields estimated to see increased yields under long term low till have not adopted the practice. Therefore, we identify specific counties likely to benefit most from increased levels of adoption. Our results strengthen the understanding of the impacts of conservation agriculture on crop yields and help define environments and counties most likely to benefit from conservation tillage.

NetFense: Adversarial Defenses against Privacy Attacks on Neural Networks for Graph Data

Recent advances in protecting node privacy on graph data and attacking graph neural networks (GNNs) gain much attention. The eye does not bring these two essential tasks together yet. Imagine an adversary can utilize the powerful GNNs to inferusers’ private labels in a social network. How can we adversarial defend against such privacy attacks while maintaining the utility of perturbed graphs? In this work, we propose a novel research task, adversarial defenses against GNN-based privacy attacks, and present a graph perturbation-based approach, NetFense, to achieve the goal. NetFense can simultaneously keep graph data unnoticed ability (i.e., having limited changes on the graph structure), maintain the prediction confidence of targeted label classification (i.e., preserving data utility), and reduce the prediction confidence of private label classification (i.e., protecting the privacy of nodes). Experiments conducted on ingle- and multiple-target perturbations using three real graph data exhibit that the perturbed graphs by NetFense can effectively maintain data utility (i.e., model unnoticed ability) on targeted label classification and significantly decrease the prediction confidence of private label classification (i.e., privacy protection). Extensive studies also bring several insights, such as the flexibility of NetFense, preserving local neighbourhoods in data unnoticed ability, and better privacy protection for high-degree nodes.

Implementation of dynamic start-up test experimental data as a main part of the nuclear code validation procedure: Developed RELAP5 model for VVER-1000

The main purposes of start-up tests in nuclear power plants (NPPs) are to ensure safe and reliable operation, verify system functionality, comply with regulatory requirements, optimize performance, and establish a foundation for ongoing plant operation and maintenance. However, the start-up tests of NPPs also could be used as a main part of the nuclear code validation procedure for several reasons including: realistic simulation, comprehensive evaluation, detection of code limitations, validation of safety margins and confidence in code predictions. The main purpose of the current study is to define and assess the validation procedure based on actual start-up test data. In this regard, the developed RELAP5 model has been validated against the actual data of VVER-1000 plant during a dynamic start-up test. The results of this full-scale validation show a good agreement between the developed RELAP5 model results and actual plant data. Finally, by defining a step by step validation procedure, it has been recommended to use the start-up phase test data as a more robust validation process which allow for full-scale validation of the nuclear code by comparing its predictions with actual plant measurements and also other advantages which have been demonstrated in the current study.

Physiologically Based Pharmacokinetic Modeling to Predict the Impact of Liver Cirrhosis on Glucuronidation via UGT1A4 and UGT2B7/2B4-A Case Study with Midazolam.

Hepatic impairment, due to liver cirrhosis, decreases the activity of cytochrome P450 enzymes (CYPs). The use of physiologically based pharmacokinetic (PBPK) modeling to predict this effect for CYP substrates has been well-established, but the effect of cirrhosis on uridine-glucuronosyltransferase (UGT) activities is less studied and few PBPK models have been reported. UGT enzymes are involved in primary N-glucuronidation of midazolam and glucuronidation of 1'-OH-midazolam following CYP3A hydroxylation. In this study, Simcyp was used to establish PBPK models for midazolam, its primary metabolites midazolam-N-glucuronide (UGT1A4) and 1'-OH midazolam (CYP3A4/3A5), and the secondary metabolite 1'-OH-midazolam-O-glucuronide (UGT2B7/2B4), allowing to simulate the impact of liver cirrhosis on the primary and secondary glucuronidation of midazolam. The model was verified in noncirrhotic subjects before extrapolation to cirrhotic patients of Child-Pugh (CP) classes A, B, and C. Our model successfully predicted the exposures of midazolam and its metabolites in noncirrhotic and cirrhotic patients, with 86% of observed plasma concentrations within 5th-95th percentiles of predictions and observed geometrical mean of area under the plasma concentration curve between 0 hours to infinity and maximal plasma concentration within 0.7- to 1.43-fold of predictions. The simulated metabolic ratio defined as the ratio of the glucuronide metabolite AUC over the parent compound AUC (AUCglucuronide/AUCparent, metabolic ratio [MR]), was calculated for midazolam-N-glucuronide to midazolam (indicative of UGT1A4 activity) and decreased by 40% (CP A), 48% (CP B), and 75% (CP C). For 1'-OH-midazolam-O-glucuronide to 1'-OH-midazolam, the MR (indicative of UGT2B7/2B4 activity) dropped by 35% (CP A), 51% (CP B), and 64% (CP C). These predicted MRs were corroborated by the observed data. This work thus increases confidence in Simcyp predictions of the effect of liver cirrhosis on the pharmacokinetics of UGT1A4 and UGT2B7/UGT2B4 substrates. SIGNIFICANCE STATEMENT: This article presents a physiologically based pharmacokinetic model for midazolam and its metabolites and verifies the accurate simulation of pharmacokinetic profiles when using the Simcyp hepatic impairment population models. Exposure changes of midazolam-N-glucuronide and 1'-OH-midazolam-O-glucuronide reflect the impact of decreases in UGT1A4 and UGT2B7/2B4 glucuronidation activity in cirrhotic patients. The approach used in this study may be extended to verify the modeling of other uridine glucuronosyltransferase enzymes affected by liver cirrhosis.

Improving multiple sclerosis lesion segmentation across clinical sites: A federated learning approach with noise-resilient training

Accurately measuring the evolution of Multiple Sclerosis (MS) with magnetic resonance imaging (MRI) critically informs understanding of disease progression and helps to direct therapeutic strategy. Deep learning models have shown promise for automatically segmenting MS lesions, but the scarcity of accurately annotated data hinders progress in this area. Obtaining sufficient data from a single clinical site is challenging and does not address the heterogeneous need for model robustness. Conversely, the collection of data from multiple sites introduces data privacy concerns and potential label noise due to varying annotation standards. To address this dilemma, we explore the use of the federated learning framework while considering label noise. Our approach enables collaboration among multiple clinical sites without compromising data privacy under a federated learning paradigm that incorporates a noise-robust training strategy based on label correction. Specifically, we introduce a Decoupled Hard Label Correction (DHLC) strategy that considers the imbalanced distribution and fuzzy boundaries of MS lesions, enabling the correction of false annotations based on prediction confidence. We also introduce a Centrally Enhanced Label Correction (CELC) strategy, which leverages the aggregated central model as a correction teacher for all sites, enhancing the reliability of the correction process. Extensive experiments conducted on two multi-site datasets demonstrate the effectiveness and robustness of our proposed methods, indicating their potential for clinical applications in multi-site collaborations to train better deep learning models with lower cost in data collection and annotation.

Public Reactions to Communication of Uncertainty: How Long-Term Benefits Can Outweigh Short-Term Costs

Abstract Uncertainty is a fact of political life but not a fact of political communication. Elites are prone to make confident predictions and downplay uncertainty about future outcomes, presumably fearing that the acknowledgement of uncertainty would undermine public confidence in their predictions and the evidence they are based on. But this calculation might both exaggerate the costs and downplay the potential benefits of reporting uncertainty. On costs, the evidence from previous studies is mixed; on benefits, previous research has neglected the possibility that, by acknowledging that outcomes may be worse than expected, those communicating uncertainty will dampen public reactions to the bad news. Here, based on a two-stage online survey experiment (N = 2,165) from December 2020 about COVID-19 vaccines, we find results suggesting that governments are well advised to communicate uncertainty. The costs at Stage 1 were low: reporting a confidence interval around the safety and effectiveness of a hypothetical COVID-19 vaccine did not undermine belief in the statistics or intentions to take the vaccine. And there were indeed benefits at Stage 2: when outcomes turned out to be worse than expected but within that confidence interval, confidence in the vaccine was partly insulated from negative effects.