Reliable Performance Prediction Research Articles

Building reliable AI for quantifying uncertainty in particulate matter predictions with deep learning

Predicting the concentrations of particulate matter (PM) has recently become crucial because of its significant impact on air pollution. As the adoption of artificial intelligence (AI) in this domain increases, ensuring the reliability of AI-driven predictions has become paramount. Although many previous studies have harnessed the power of AI for PM prediction, the inherent uncertainty in the results, which is influenced by spatial data imbalances and varying meteorological factors, poses a challenge. Addressing this uncertainty is central to building reliable AI systems. Therefore, we employed the Monte Carlo dropout (MCDO) approach, which is a technique that integrates dropout layers in neural networks during both the training and inference phases, to estimate the prediction uncertainty. Our objective was to address the challenge of prediction uncertainty in PM concentrations, with the ultimate aim of enhancing the reliability and trustworthiness of AI-driven predictions in air quality forecasting. By applying the MCDO approach to grids formed from multidimensional arrays of air quality and weather data, we obtained a 95% confidence interval for the prediction results, thereby demonstrating the trustworthiness of our model. Our evaluation revealed an R-square greater than 0.97 for PM10 and PM2.5, showcasing the robust and reliable predictive performance of the model. This study highlights not only the accuracy of this approach but also the critical role of quantifying uncertainty in building reliable AI systems. Our findings mark a significant step towards a holistic approach to predictive modeling in which reliability and uncertainty quantification are at the forefront.

Effective design of sustainable energy productivity based on the experimental investigation of the humidification-dehumidification-desalination system using hybrid optimization

Reliable and cost-effective industrial-based sustainable desalination technologies are crucial for efficient and effective desalination of saline water. Seawater desalination emerges as a vital treatment, with humidification-dehumidification identified as a promising technology due to its simple, low-maintenance design and compatibility with renewable energy sources. The present work was developed using a large-scale experimental test rig based on the performance of a humidification-dehumidification system operated by the heat pump. As such, the study presents an advanced approach for optimizing humidification-dehumidification desalination systems using hybrid machine learning optimization techniques. The effectiveness of a neuro-fuzzy model, a decision algorithm based on a boosted tree, and a simple averaging ensemble in enhancing the performance of humidification-dehumidification systems were investigated. Experimental data includes cold water flowrate (L/min), hot water flowrate (L/min), inlet air temperature (°C), inlet cold water temperature (°C), inlet hot water temperature (°C), and inlet air relative humidity (%) as input variables combination and freshwater productivity (L/h), recovery ratio (%) as target variables. Several statistical indicators were used to evaluate the models’ prediction skills; similarly, a 2-dimensional Taylor diagram and error-radial plot were also utilized. In the calibration phase, boosted tree models slightly outperform neuro-fuzzy models, particularly for recovery ratio, demonstrating high accuracy and low bias. However, all models exhibit robust predictive accuracy in the verification phase, especially for recovery ratio, emphasizing their potential in generalizing to unseen data. The result indicated that the boosted tree outperformed others with Nash–Sutcliffe Efficiency = 94 % and 88 % for recovery ratio and freshwater productivity, respectively. The simple averaging ensemble shows marginal improvement with 95 % accuracy for the recovery ratio. Integrating hybrid artificial intelligence optimization with humidification-dehumidification desalination systems marks a transformative step in addressing freshwater shortages. By utilizing cutting-edge machine learning techniques, we achieve a more efficient, accurate, and reliable prediction of desalination performance.

Predictive Machine Learning Model to Assess the Adsorption Efficiency of Biochar-Heavy Metals for Effective Remediation of Soil-Plant Environment.

Biochar is crucial for agricultural output and plays a significant role in effectively eliminating heavy metals (HMs) from the soil, which is essential for maintaining a soil-plant environment. This work aimed to assess machine learning models to analyze the impact of soil parameters on the transformation of HMs in biochar-soil-plant environments, considering the intricate non-linear relationships involved. A total of 211 datasets from pot or field experiments were evaluated. Fourteen factors were taken into account to assess the efficiency and bioavailability of HM-biochar amendment immobilization. Four predictive models, namely linear regression (LR), partial least squares (PLS), support vector regression (SVR), and random forest (RF), were compared to predict the immobilization efficiency of biochar-HM. The findings revealed that the RF model was created using 5-fold cross-validation, which exhibited a more reliable prediction performance. The results indicated that soil features accounted for 79.7% of the absorption of HM by crops, followed by biochar properties at 17.1% and crop properties at 3.2%. The main elements that influenced the result have been determined as the characteristics of the soil (including the presence of different HM species and the amount of clay) and the quantity and attributes of the biochar (such as the temperature at which it was produced by pyrolysis). Furthermore, the RF model was further developed to predict bioaccumulation factors (BAF) and variations in crop uptake (CCU). The R2 values were found to be 0.7338 and 0.6997, respectively. Thus, machine learning (ML) models could be useful in understanding the behavior of HMs in soil-plant ecosystems by employing biochar additions.

Towards Reliable Prediction of Performance for Polymer Electrolyte Membrane Fuel Cells via Machine Learning-Integrated Hybrid Numerical Simulations

Towards Reliable Prediction of Performance for Polymer Electrolyte Membrane Fuel Cells via Machine Learning-Integrated Hybrid Numerical Simulations

An experimental validation of diffusion-based devolatilization models in extruders using post-industrial and post-consumer plastic waste

Abstract Extrusion is a key process in mechanical recycling. In a degassing step, volatile components, including all impurities and moisture, are removed from a polymer melt to ensure consistently high quality of the recyclates. Predicting devolatilization performance is therefore of interest in the design of degassing screws; in the plastics industry, it also plays an important role in the transition from a linear to a circular economy. Using two different devolatilization models, we first modelled the degassing process of a lab-scale twin-screw extruder and an industrial-scale recycling single-screw extruder. We then predicted the devolatilization performance of both machines, validated the results with experimental data obtained from emissions tests carried out with post-industrial and post-consumer polypropylene materials and performed linear regression analysis to compare our two models in terms of predictive quality. Our results showed that both models are equally suitable for reliable prediction of the devolatilization performance.

Dynamic prediction of high-temperature points in longwall gobs under a multi-field coupling framework

Spontaneous coal combustion (SCC) in longwall gobs is a serious disaster and requires prediction tools to accurately and reliably predict the temperature in gobs, but currently, there has been a lack of simple and efficient predictors. To address this issue, the temperature distributions in a longwall gob under different mining state parameters were simulated using a multi-fields coupled model. The temperature distribution and the mining state parameters were extracted to serve as input variables for prediction. Two machine learning models, integrating parameter optimization and ensemble learning strategies, respectively, were proposed to establish the functional relationships between the dynamic change boundary temperature and the maximum temperature in the gob. The results show that these two models have good prediction accuracy on the high-temperature points in longwall gobs, indicating that the ensemble learning has improved prediction accuracy. There is a positive correlation between the number of measuring points and the prediction accuracy, and three measurement points at the key locations can deliver reliable predictive performance.

Artificial Neural Network-based model to predict the International Roughness Index of national highways in Nepal

Reliable predictions of pavement performance are crucial for road maintenance, rehabilitation, and reconstruction planning. To facilitate predictions of the International Roughness Index (IRI) changes over time on national highways in Nepal, this study develops a comprehensive overall model, along with regional models that consider climatic and traffic variations among the highways. The study models IRI over time using the Artificial Neural Network (ANN) approach and compares the results with those obtained from a multiple linear regression-based model. The models are developed using pavement IRI, traffic, and climatic (rainfall and temperature) data specific to national highways of Nepal, encompassing 1745 sections and 3710 total observations. The ANN-based overall model has a coefficient of determination (R2) value of 0.82 and outperforms the regression-based model, which has an R2 value of 0.76. The regional models developed for the Terai, Hill, high volume Terai and low volume Terai highways have R2 values of 0.87, 0.91, 0.85 and 0.88, respectively, indicating a good fit. Analysis of the IRI trend over time, as observed from the performance curves generated from the ANN-based model, revealed an S-shaped pattern and lower Root Mean Square Error (RMSE) compared to the regression-based model. Sensitivity analysis highlighted the initial pavement IRI as the most significant parameter in all cases. High temperature days emerged as the second most influential parameter in most models, except for the high volume Terai model, where the number of commercial vehicles serves as the second most sensitive parameter after the initial IRI.

Pore fluid salinity effects on post-stabilization performance of cement-treated Marl

Soil improvement using chemical admixtures has been extensively investigated in the recent past and these admixtures are abundantly being used on various construction projects. However, there is a need to elucidate the long-term performance of stabilized geomaterials under aggressive pore fluid environments such as high salinity of groundwater in coastal areas. This study presents the effects of the concentration of pore fluid salinity on the post-stabilization behavior of cement-treated Marl specimens to brine solution from Sabkha deposits and rejected water from a water treatment plant. Unconfined compressive strength (UCS), mineralogy, chemical, and microstructural evaluations have been carried out in this regard. minerals like albite and muscovite completely leached out of the samples after 28 days of soaking in a saline environment. A new mineral Attapulgite was formed at various concentration levels of the pore fluid. Significant changes in the chemical composition were also observed in various phases of the experiments. Moreover, at all salinity levels, the UCS initially increased to a maximum of 75 % which later reduced to a value less than the original strength whereas the elastic modulus (E50) increased as the soaking period progressed. Lastly, it is emphasized that the engineering properties of treated soils exposed to aggressive pore fluids should be carefully assessed for a reliable prediction of the long-term performance of geotechnical structures.

Axial Compressor Transonic Rotor Shock Interaction With Upstream Stator Row

Abstract Modern multistage axial compressors are designed for high aerodynamic loading and wide range with front stages rotor inlet relative Mach number above one over a large portion of span. Rotor shock waves propagate and reflect in the upstream stator where they provoke extra aerodynamic losses and aeromechanic risks. This paper discusses stator–rotor shock interactions using both steady and unsteady numerical simulations. The computations focus on three different stage designs with the same capacity and total pressure ratio, but different axial gaps and load distribution in stator and rotor. Speedlines from steady and unsteady computations compare the designs, and the detailed flow fields are analyzed at design and close to stall conditions. The steady and unsteady calculations predict similar stage performance, but different stator–rotor loss split. The analyses reveal how the rotor shocks orientation with respect to the upstream stator camber line controls the intensity of shocks interaction with the stator row and its propagation upstream. Finally, the paper indicates the ways to reduce the shock-driven rotor–stator interaction and suggests simple criteria to determine when to switch from steady to unsteady calculations for a reliable transonic stage performance prediction.

Predicting Photodetector Responsivity through Machine Learning

AbstractThis study introduces a novel methodology for predicting photodetector responsivity, specifically targeting challenging materials like borophene. The synthesis of these materials faces substantial experimental complexities, necessitating reliable performance predictions before fabrication. To address this, a comprehensive approach leveraging advanced machine learning techniques, specifically artificial neural networks (ANN), is developed. Integration of X‐ray diffraction (XRD) and Raman spectra data into AI models enables efficient prediction of photodetector efficiency prior to device fabrication. The innovation lies in strategically incorporating Generative Adversarial Networks (GANs) for dataset augmentation, significantly expanding the dataset size and enhancing the robustness of the ANN model. Sensitivity analyses highlighted influential factors such as bias voltage and spectral coefficients, validating the approach and aligning with recent experimental findings. This methodology not only advances optoelectronics, but also holds promise for materials science and device engineering. Predictions for Wavelength‐Responsivity plots, considering borophene allotropes as active layers and n‐Si as substrates, show peaks around 300–400 nm, ranging from 0.04 to 0.36 AW−1at bias voltages between 1 and 5 volts. These estimations assume a borophene layer thickness of approximately 1.6 nm and a radiation power intensity of 5000 µ Wcm−2.

Optimization of a high through-flow design turbine using response surface method

Enhancing the through-flow capability of the turbine facilitates the potential to decrease the external dimension of the engine, leading to a reduction in weight and an increase in the thrust-to-weight ratio. The optimization of blade profile in terms of aerodynamic design is an essential strategy to improve the overall performance of the high through-flow design turbine. This study aims to develop an automated optimization technique, utilizing the response surface method (RSM), to enable reliable prediction and rapid optimization of aerodynamic performance. A parametric modeling method is devised to generate blade profiles with continuous curvature throughout the automated optimization process. The Box–Behnken experimental design, in combination with Reynolds-averaged Navier–Stokes numerical calculation, is employed to construct a second-order polynomial RSM approximation model. The optimization process comprises two levels: improving the through-flow capability and reducing blade profile loss. The optimized profile, Opt1, achieves a substantial 13.19% improvement in through-flow capability at the cost of a 12.27% increase in blade profile loss. Thus, further optimization is performed to minimize blade profile loss based on the Opt1 scheme. Geometric constraints are applied to the most influential parameters affecting through-flow capability to mitigate their impacts. Compared to the Opt1, the final optimized profile, Opt2, achieves a substantial 31.83% decrease in blade profile loss with a negligible sacrifice of 0.23% in through-flow capability.

Performance evaluation of cable-stayed bridge expansion joints based on Lasso dimensionality reduction and temperature-displacement-correlation model

Expansion joints play a crucial role in accommodating the longitudinal movement of the main beam, which is mainly caused by temperature variation. This paper establishes an accurate model that relates the temperature field of the main beam to the displacement of the expansion joint, enabling reliable performance prediction and early warning of the expansion joint. Firstly, three commonly used methods for characterizing the temperature field of the main beam are introduced, along with their advantages and disadvantages. Secondly, a novel method is proposed using the Lasso algorithm to calculate critical temperatures. The objective is to select temperature channels data that have significant impact on the longitudinal displacement of the main beam. The selected channels data is then linearly weighted based on feature importance to obtain critical temperature. Based on this, a precise relationship model between the main beam temperature and the expansion joint displacement is derived through regression. For the residual term in the model fitting, an expansion joint performance early warning procedure is developed based on the X-bar control chart. Finally, using one-year long-term monitoring data from a newly constructed cable-stayed bridge as an example, the proposed method demonstrates superior capability in predicting the predefined damage of the expansion joint compared to the other two commonly used methods.

High somatic mutations in circulating tumor DNA predict response of metastatic pancreatic ductal adenocarcinoma to first-line nab-paclitaxel plus S-1: prospective study

AimsWe previously showed that the nab-paclitaxel plus S-1 (NPS) regimen had promising effects against metastatic pancreatic ducal adenocarcinoma (mPDAC), whose efficacy however could not be precisely predicted by routine biomarkers. This prospective study aimed to investigate the values of mutations in circulating tumor DNA (ctDNA) and their dynamic changes in predicting response of mPDAC to NPS chemotherapy.MethodsPaired tumor tissue and blood samples were prospectively collected from patients with mPDAC receiving first-line NPS chemotherapy, and underwent next-generation sequencing with genomic profiling of 425 genes for ctDNA. High mutation allelic frequency (MAF) was defined as ≥ 30% and ≥ 5% in tumor tissue and blood, respectively. Kappa statistics were used to assess agreement between mutant genes in tumor and ctDNA. Associations of mutations in ctDNA and their dynamic changes with tumor response, overall survival (OS), and progression-free survival (PFS) were assessed using the Kaplan–Meier method, multivariable-adjusted Cox proportional hazards regression, and longitudinal data analysis.Results147 blood samples and 43 paired tumor specimens from 43 patients with mPDAC were sequenced. The most common driver genes with high MAF were KRAS (tumor, 35%; ctDNA, 37%) and TP53 (tumor, 37%; ctDNA, 33%). Mutation rates of KRAS and TP53 in ctDNA were significantly higher in patients with liver metastasis, with baseline CA19-9 ≥ 2000 U/mL, and/or without an early CA19-9 response. κ values for the 5 most commonly mutated genes between tumor and ctDNA ranged from 0.48 to 0.76. MAFs of the genes mostly decreased sequentially during subsequent measurements, which significantly correlated with objective response, with an increase indicating cancer progression. High mutations of KRAS and ARID1A in both tumor and ctDNA, and of TP53, CDKN2A, and SMAD4 in ctDNA but not in tumor were significantly associated with shorter survival. When predicting 6-month OS, AUCs for the 5 most commonly mutated genes in ctDNA ranged from 0.59 to 0.84, larger than for genes in tumor (0.56 to 0.71) and for clinicopathologic characteristics (0.51 to 0.68). Repeated measurements of mutations in ctDNA significantly differentiated survival and tumor response. Among the 31 patients with ≥ 2 ctDNA tests, longitudinal analysis of changes in gene MAF showed that ctDNA progression was 60 and 58 days ahead of radiologic and CA19-9 progression for 48% and 42% of the patients, respectively.ConclusionsHigh mutations of multiple driving genes in ctDNA and their dynamic changes could effectively predict response of mPDAC to NPS chemotherapy, with promising reliable predictive performance superior to routine clinicopathologic parameters. Inspiringly, longitudinal ctDNA tracking could predict disease progression about 2 months ahead of radiologic or CA19-9 evaluations, with the potential to precisely devise individualized therapeutic strategies for mPDAC.

High-Throughput Prediction of the Thermal and Electronic Transport Properties of Large Physical and Chemical Spaces Accelerated by Machine Learning: Charting the ZT of Binary Skutterudites.

Thermal and electronic transport properties are the keys to many technological applications of materials. Thermoelectric, TE, materials can be considered a singular case in which not only one but three different transport properties are combined to describe their performance through their TE figure of merit, ZT. Despite the availability of high-throughput experimental techniques, synthesizing, characterizing, and measuring the properties of samples with numerous variables affecting ZT are not a cost- or time-efficient approach to lead this strategy. The significance of computational materials science in discovering new TE materials has been running in parallel to the development of new frameworks and methodologies to compute the electron and thermal transport properties linked to ZT. Nevertheless, the trade-off between computational cost and accuracy has hindered the reliable prediction of TE performance for large chemical spaces. In this work, we present for the first time the combination of new ab initio methodologies to predict transport properties with machine learning and a high-throughput framework to establish a solid foundation for the accurate prediction of thermal and electron transport properties. This strategy is applied to a whole family of materials, binary skutterudites, which are well-known as good TE candidates. Following this methodology, it is possible not only to connect ZT with the experimental synthetic (carrier concentration and grain size) and operando (temperature) variables but also to understand the physical and chemical phenomena that act as driving forces in the maximization of ZT for p-type and n-type binary skutterudites.

A Method for Calculating the Leaf Area of Pak Choi Based on an Improved Mask R-CNN

The leaf area of pak choi is a critical indicator of growth rate, nutrient absorption, and photosynthetic efficiency, and it is required to be precisely measured for an optimal agricultural output. Traditional methods often fail to deliver the necessary accuracy and efficiency. We propose a method for calculating the leaf area of pak choi based on an improved Mask R-CNN. We have enhanced Mask R-CNN by integrating an advanced attention mechanism and a two-layer fully convolutional network (FCN) into its segmentation branch. This integration significantly improves the model’s ability to detect and segment leaf edges with increased precision. By extracting the contours of reference objects, the conversion coefficient between the pixel area and the actual area is calculated. Using the mask segmentation output from the model, the area of each leaf is calculated. Experimental results demonstrate that the improved model achieves mean average precision (mAP) scores of 0.9136 and 0.9132 in detection and segmentation tasks, respectively, representing improvements of 1.01% and 1.02% over the original Mask R-CNN. The model demonstrates excellent recognition and segmentation capabilities for pak choi leaves. The error between the calculation result of the segmented leaf area and the actual measured area is less than 4.47%. These results indicate that the proposed method provides a reliable segmentation and prediction performance. It eliminates the need for detached leaf measurements, making it suitable for real-life leaf area measurement scenarios and providing valuable support for automated production technologies in plant factories.

MVNMDA: A Multi-View Network Combing Semantic and Global Features for Predicting miRNA-Disease Association.

A growing body of experimental evidence suggests that microRNAs (miRNAs) are closely associated with specific human diseases and play critical roles in their development and progression. Therefore, identifying miRNA related to specific diseases is of great significance for disease screening and treatment. In the early stages, the identification of associations between miRNAs and diseases demanded laborious and time-consuming biological experiments that often carried a substantial risk of failure. With the exponential growth in the number of potential miRNA-disease association combinations, traditional biological experimental methods face difficulties in processing massive amounts of data. Hence, developing more efficient computational methods to predict possible miRNA-disease associations and prioritize them is particularly necessary. In recent years, numerous deep learning-based computational methods have been developed and have demonstrated excellent performance. However, most of these methods rely on external databases or tools to compute various auxiliary information. Unfortunately, these external databases or tools often cover only a limited portion of miRNAs and diseases, resulting in many miRNAs and diseases being unable to match with these computational methods. Therefore, there are certain limitations associated with the practical application of these methods. To overcome the above limitations, this study proposes a multi-view computational model called MVNMDA, which predicts potential miRNA-disease associations by integrating features of miRNA and diseases from local views, global views, and semantic views. Specifically, MVNMDA utilizes known association information to construct node initial features. Then, multiple networks are constructed based on known association to extract low-dimensional feature embedding of all nodes. Finally, a cascaded attention classifier is proposed to fuse features from coarse to fine, suppressing noise within the features and making precise predictions. To validate the effectiveness of the proposed method, extensive experiments were conducted on the HMDD v2.0 and HMDD v3.2 datasets. The experimental results demonstrate that MVNMDA achieves better performance compared to other computational methods. Additionally, the case study results further demonstrate the reliable predictive performance of MVNMDA.

Pavement Analysis with the Consideration of Unbound Granular Material Nonlinearity

Accurate pavement design and evaluation requires the execution of response analysis. Pavement materials’ behavior does not necessarily conform to the assumptions of the multi-linear elastic theory usually adopted during pavement analysis. In particular, the unbound granular materials located in the base and sub-base layers behave in a nonlinear elastic manner, which can be captured through advanced constitutive modeling of their resilient modulus. The finite element method enables us to code constitutive models and quantify potential variations in pavement responses because of different mechanistic assumptions. In this study, variations in response are investigated for a typical structure of a flexible pavement considering the nonlinear anisotropic behavior of the unbound materials together with their initial stress–strain state. To demonstrate the impact of their behavior on the outcome of pavement analysis, variable asphalt concrete layer thicknesses and moduli are assumed, such that they cover a large spectrum of roadways. It was found that pavement responses can be calculated up to 3.5 times higher than those retrieved from the conventional linear analysis. This comparison means that the alterative mechanistic modeling of the unbound granular materials can be proved to be more conservative (i.e., leading to higher strains) in terms of pavement design and analysis. From a practical perspective, this study alerts pavement scientists and engineers engaged in pavement design to a more reliable performance prediction, which is needed to bridge the gap between advanced modeling and routine analysis.

Development of an Intratumoral and Peritumoral Radiomics Nomogram Using Digital Breast Tomosynthesis for Preoperative Assessment of Lymphovascular Invasion in Invasive Breast Cancer

This study aimed to create a nomogram model that combines clinical factors with radiomics analysis of both intra- and peritumoral regions extracted from preoperative digital breast tomosynthesis (DBT) images, in order to develop a reliable method for predicting the lymphovascular invasion (LVI) status in invasive breast cancer (IBC) patients. A total of 178 patients were randomly split into a training dataset (N=124) and a validation dataset (N=54). Comprehensive clinical data, encompassing DBT features, were gathered for all cases. Radiomics features were extracted and selected from intra- and peritumoral region to establish radiomics signature (Radscore). To construct the clinical model and nomogram model, univariate and multivariate logistic regression analyses were utilized to identify independent risk factors. To assess and validate these models, various analytical methods were employed, including receiver operating characteristic (ROC) curve analysis, calibration curve analysis, decision curve analysis (DCA), net reclassification improvement (NRI), and integrated discriminatory improvement (IDI). The clinical model is constructed based on two independent risk factors: tumor margin and the DBT-reported lymph node metastasis (DBT_reported_LNM). Incorporating Radscore_Combine(utilizing both intra- and peritumoral radiomics features), tumor margin, and DBT_reported_LNM into the nomogram achieved a reliable predictive performance, with area under the curve (AUC) values of 0.906 and 0.905 in both datasets, respectively. The significant improvement demonstrated by the NRI and IDI indicates that the Radscore_Combine could be a valuable biomarker for effectively predicting the status of LVI. The nomogram demonstrated a reliable ability to predict LVI in IBC patients.

Uncertainty-aware fatigue-life prediction of additively manufactured Hastelloy X superalloy using a physics-informed probabilistic neural network

Microstructural inhomogeneity in additively manufactured (AM) components leads to uncertainty in their fatigue performance. While purely data-driven methods can only provide deterministic outcomes and lack physical interpretability. Furthermore, considering the dispersion of fatigue life, a probabilistic neural network framework integrating physical information, namely a physics-informed probabilistic neural network (PIPNN), is proposed for predicting the fatigue life of AM parts. The framework describes the dispersion of fatigue life in the parametric form of probability statistics. It incorporates physical laws and models to constrain neurons and loss function, enabling the network to learn deeper physical laws that align with the fatigue process, thus enhancing the interpretability and prediction reliability of the model. Fatigue experiments were performed on Hastelloy X superalloy specimens fabricated using laser powder bed fusion, serving as the basis for validating and comparing the PIPNN model with a probabilistic neural network. The results indicate that PIPNN adeptly captures the heteroskedasticity of fatigue life and exhibits superior prediction accuracy and more reliable prediction performance in fatigue-life prediction. PIPNN offers a physically consistent method for fatigue-life prediction considering probabilistic statistics.

Identification and validation of a dysregulated TME-related gene signature for predicting prognosis, and immunological properties in bladder cancer.

During tumor growth, tumor cells interact with their tumor microenvironment (TME) resulting in the development of heterogeneous tumors that promote tumor occurrence and progression. Recently, there has been extensive attention on TME as a possible therapeutic target for cancers. However, an accurate TME-related prediction model is urgently needed to aid in the assessment of patients' prognoses and therapeutic value, and to assist in clinical decision-making. As such, this study aimed to develop and validate a new prognostic model based on TME-associated genes for BC patients. Transcriptome data and clinical information for BC patients were extracted from The Cancer Genome Atlas (TCGA) database. Gene Expression Omnibus (GEO) and IMvigor210 databases, along with the MSigDB, were utilized to identify genes associated with TMEs (TMRGs). A consensus clustering approach was used to identify molecular clusters associated with TMEs. LASSO Cox regression analysis was conducted to establish a prognostic TMRG-related signature, with verifications being successfully conducted internally and externally. Gene ontology (GO), KEGG, and single-sample gene set enrichment analyses (ssGSEA) were performed to investigate the underlying mechanisms. The potential response to ICB therapy was estimated using the Tumor Immune Dysfunction and Exclusion (TIDE) algorithm and Immunophenoscore (IPS). Additionally, it was found that the expression level of certain genes in the model was significantly correlated with objective responses to anti-PD-1 or anti-PD-L1 treatment in the IMvigor210, GSE111636, GSE176307, or Truce01 (registration number NCT04730219) cohorts. Finally, real-time PCR validation was performed on 10 paired tissue samples, and in vitro cytological experiments were also conducted on BC cell lines. In BC patients, 133 genes differentially expressed that were associated with prognosis in TME. Consensus clustering analysis revealed three distinct clinicopathological characteristics and survival outcomes. A novel prognostic model based on nine TMRGs (including C3orf62, DPYSL2, GZMA, SERPINB3, RHCG, PTPRR, STMN3, TMPRSS4, COMP) was identified, and a TMEscore for OS prediction was constructed, with its reliable predictive performance in BC patients being validated. MultiCox analysis showed that the risk score was an independent prognostic factor. A nomogram was developed to facilitate the clinical viability of TMEscore. Based on GO and KEGG enrichment analyses, biological processes related to ECM and collagen binding were significantly enriched among high-risk individuals. In addition, the low-risk group, characterized by a higher number of infiltrating CD8+ T cells and a lower burden of tumor mutations, demonstrated a longer survival time. Our study also found that TMEscore correlated with drug susceptibility, immune cell infiltration, and the prediction of immunotherapy efficacy. Lastly, we identified SERPINB3 as significantly promoting BC cells migration and invasion through differential expression validation and in vitro phenotypic experiments. Our study developed a prognostic model based on nine TMRGs that accurately and stably predicted survival, guiding individual treatment for patients with BC, and providing new therapeutic strategies for the disease.