Combined Dataset Research Articles

Exploring the efficacy of targeted fingerprinting marker techniques in genetic studies of the Indian poppy (Papaver somniferum L.): A comparative analysis

Targeted fingerprinting marker techniques have emerged as promising tools in plant sciences, particularly in plant breeding. To initiate an effective breeding program for a licit but industrially potential plant, the opium poppy, a comparative assessment of these markers was conducted. In this study, we compared the Start Codon Targeted polymorphism (SCoT), CAAT box-derived polymorphism (CBDP), and Expressed sequence tag-simple sequence repeats (EST-SSRs) markers for their competence and informativeness in revealing genetic identity, relatedness, and diversity among ten released varieties of the Indian opium poppy. Among the markers studied, EST-SSR markers demonstrated higher discriminating power, attributed to their ability to unveil comparatively higher polymorphic information content, marker index, and effective marker index. While all three marker systems effectively discriminated the poppy varieties, EST-SSR markers showed greater genetic diversity and a more informative ancestral history-based dendrogram topology. Significant correlation coefficients were observed between EST-SSR and CBDP markers, particularly in the combined marker dataset. Furthermore, the co-phenetic correlation was significant for three marker systems, with EST-SSR markers showing maximum correlation. Notably, the dendrograms constructed using EST-SSR markers and combined marker data displayed a significant and positive correlation with phenotypic traits assessed for the ten poppy varieties. Overall, the findings highlight the superiority of EST-SSR markers in deciphering genetic relationships among Indian opium poppy varieties and suggest their potential as valuable tools in breeding and genetic diversity studies.

Three New Species of Microdochium (Microdochiaceae, Xylariales) on Bambusaceae sp. and Saprophytic Leaves from Hainan and Yunnan, China.

Species of the genus Microdochium (Microdochiaceae, Xylariales) have been reported from the whole world and separated from multiple plant hosts. The primary aim of the present study is to describe and illustrate three new species isolated from the leaf spot of Bambusaceae sp. and saprophytic leaves in Hainan and Yunnan provinces, China. The proposed three species, viz., Microdochium bambusae, M. nannuoshanense and M. phyllosaprophyticum, are based on multi-locus phylogenies from a combined dataset of ITS rDNA, LSU, RPB2 and TUB2 in conjunction with morphological characteristics. Descriptions and illustrations of three new species in the genus are provided.

StableDNAm: towards a stable and efficient model for predicting DNA methylation based on adaptive feature correction learning

BackgroundDNA methylation, instrumental in numerous life processes, underscores the paramount importance of its accurate prediction. Recent studies suggest that deep learning, due to its capacity to extract profound insights, provides a more precise DNA methylation prediction. However, issues related to the stability and generalization performance of these models persist.ResultsIn this study, we introduce an efficient and stable DNA methylation prediction model. This model incorporates a feature fusion approach, adaptive feature correction technology, and a contrastive learning strategy. The proposed model presents several advantages. First, DNA sequences are encoded at four levels to comprehensively capture intricate information across multi-scale and low-span features. Second, we design a sequence-specific feature correction module that adaptively adjusts the weights of sequence features. This improvement enhances the model’s stability and scalability, or its generality. Third, our contrastive learning strategy mitigates the instability issues resulting from sparse data. To validate our model, we conducted multiple sets of experiments on commonly used datasets, demonstrating the model’s robustness and stability. Simultaneously, we amalgamate various datasets into a single, unified dataset. The experimental outcomes from this combined dataset substantiate the model’s robust adaptability.ConclusionsOur research findings affirm that the StableDNAm model is a general, stable, and effective instrument for DNA methylation prediction. It holds substantial promise for providing invaluable assistance in future methylation-related research and analyses.

Gut microbiota signatures in inflammatory bowel disease.

Inflammatory bowel diseases (IBD), including Crohn's disease (CD) and ulcerative colitis (UC), affect millions of people worldwide with increasing incidence. Several studies have shown a link between gut microbiota composition and IBD, but results are often limited by small sample sizes. We aimed to re-analyze publicly available fecal microbiota data from IBD patients. We extracted original fecal 16S rRNA amplicon sequencing data from 45 cohorts of IBD patients and healthy individuals using the BioProject database at the National Center for Biotechnology Information. Unlike previous meta-analyses, we merged all study cohorts into a single dataset, including sex, age, geography, and disease information, based on which microbiota signatures were analyzed, while accounting for varying technical platforms. Among 2518 individuals in the combined dataset, we discovered a hitherto unseen number of genera associated with IBD. A total of 77 genera associated with CD, of which 38 were novel associations, and a total of 64 genera associated with UC, of which 28 represented novel associations. Signatures were robust across different technical platforms and geographic locations. Reduced alpha diversity in IBD compared to healthy individuals, in CD compared to UC, and altered microbiota composition (beta diversity) in UC and especially in CD as compared to healthy individuals were found. Combining original microbiota data from 45 cohorts, we identified a hitherto unseen large number of genera associated with IBD. Identification of microbiota features robustly associated with CD and UC may pave the way for the identification of new treatment targets.

Clinico-biological-radiomics (CBR) based machine learning for improving the diagnostic accuracy of FDG-PET false-positive lymph nodes in lung cancer

BackgroundThe main problem of positron emission tomography/computed tomography (PET/CT) for lymph node (LN) staging is the high false positive rate (FPR). Thus, we aimed to explore a clinico-biological-radiomics (CBR) model via machine learning (ML) to reduce FPR and improve the accuracy for predicting the hypermetabolic mediastinal–hilar LNs status in lung cancer than conventional PET/CT.MethodsA total of 260 lung cancer patients with hypermetabolic mediastinal–hilar LNs (SUVmax ≥ 2.5) were retrospectively reviewed. Patients were treated with surgery with systematic LN resection and pathologically divided into the LN negative (LN-) and positive (LN +) groups, and randomly assigned into the training (n = 182) and test (n = 78) sets. Preoperative CBR dataset containing 1738 multi-scale features was constructed for all patients. Prediction models for hypermetabolic LNs status were developed using the features selected by the supervised ML algorithms, and evaluated using the classical diagnostic indicators. Then, a nomogram was developed based on the model with the highest area under the curve (AUC) and the lowest FPR, and validated by the calibration plots.ResultsIn total, 109 LN− and 151 LN + patients were enrolled in this study. 6 independent prediction models were developed to differentiate LN− from LN + patients using the selected features from clinico-biological-image dataset, radiomics dataset, and their combined CBR dataset, respectively. The DeLong test showed that the CBR Model containing all-scale features held the highest predictive efficiency and the lowest FPR among all of established models (p < 0.05) in both the training and test sets (AUCs of 0.90 and 0.89, FPRs of 12.82% and 6.45%, respectively) (p < 0.05). The quantitative nomogram based on CBR Model was validated to have a good consistency with actual observations.ConclusionThis study presents an integrated CBR nomogram that can further reduce the FPR and improve the accuracy of hypermetabolic mediastinal–hilar LNs evaluation than conventional PET/CT in lung cancer, thereby greatly reducing the risk of overestimation and assisting for precision treatment.

Evaluating the Performance of Automated Machine Learning (AutoML) Tools for Heart Disease Diagnosis and Prediction

Globally, over 17 million people annually die from cardiovascular diseases, with heart disease being the leading cause of mortality in the United States. The ever-increasing volume of data related to heart disease opens up possibilities for employing machine learning (ML) techniques in diagnosing and predicting heart conditions. While applying ML demands a certain level of computer science expertise—often a barrier for healthcare professionals—automated machine learning (AutoML) tools significantly lower this barrier. They enable users to construct the most effective ML models without in-depth technical knowledge. Despite their potential, there has been a lack of research comparing the performance of different AutoML tools on heart disease data. Addressing this gap, our study evaluates three AutoML tools—PyCaret, AutoGluon, and AutoKeras—against three datasets (Cleveland, Hungarian, and a combined dataset). To evaluate the efficacy of AutoML against conventional machine learning methodologies, we crafted ten machine learning models using the standard practices of exploratory data analysis (EDA), data cleansing, feature engineering, and others, utilizing the sklearn library. Our toolkit included an array of models—logistic regression, support vector machines, decision trees, random forest, and various ensemble models. Employing 5-fold cross-validation, these traditionally developed models demonstrated accuracy rates spanning from 55% to 60%. This performance is markedly inferior to that of AutoML tools, indicating the latter’s superior capability in generating predictive models. Among AutoML tools, AutoGluon emerged as the superior tool, consistently achieving accuracy rates between 78% and 86% across the datasets. PyCaret’s performance varied, with accuracy rates from 65% to 83%, indicating a dependency on the nature of the dataset. AutoKeras showed the most fluctuation in performance, with accuracies ranging from 54% to 83%. Our findings suggest that AutoML tools can simplify the generation of robust ML models that potentially surpass those crafted through traditional ML methodologies. However, we must also consider the limitations of AutoML tools and explore strategies to overcome them. The successful deployment of high-performance ML models designed via AutoML could revolutionize the treatment and prevention of heart disease globally, significantly impacting patient care.

Uncovering Novel Modifiers of Tau Aggregation and Pathology using a Proximity Proteomics Approach (BioID)

AbstractBackgroundAlzheimer’s disease (AD) is characterized by the accumulation of neurofibrillary tangles (NFTs), comprised primarily of microtubule‐associated protein tau. NFTs are defining hallmarks of AD and its accumulation into pathological oligomers and fibrils has been associated with cognitive decline in AD and other tauopathies. Despite its pivotal role in the disease process, we still have a poor understanding of how their formation, toxicity, and spread across brain regions is regulated via tau‐associated proteins that regulate the oligomerization, pathological accumulation, and seeding of tau in affected cell types and vulnerable brain regions.MethodWe have established two proximity‐dependent biotin‐identification (BioID) models as novel approaches to identify the composition and proximal molecular environment of tau aggregates. Using an in vitro and ex vivo model approach coupled with mass spectrometry, we have identified proteins that were primarily associated with mutant tau aggregates vs. soluble wildtype tau and stratified them based on significance, molecular pathway and druggability. Immunoblotting and immunofluorescence were performed to assess the interaction of endogenous and overexpressed top candidates with tau, along with visualization with NFTs in the context of human AD pathology to assess their impact on tau‐aggregate formation and toxicity.ResultTo determine the physiological and aggregate specific interacting partners of tau, we are comparing wild type tau (TauWT) with mutant tau (Tau3xMUT) carrying three tauopathy‐associated mutations (A152T/P301L/S320F). The proteomic analysis from our combined proteomic datasets revealed that significantly changed proteins included candidates involved in RNA processing, as well as protein ubiquitination and proteasome degradation. We have begun validating proteins that are enriched in mutant tau in human tissue of AD, as well as primary tauopathies, that may be druggable targets for tauopathies.ConclusionThis study provides novel insights into cellular pathways and molecular mechanisms of neurodegeneration, by identifying in the context of living neurons and brain tissue different functional classes of tau‐associated proteins with relevance for AD pathophysiology. These are expected to include proteins that may mediate the toxic effect of NFTs, facilitate the formation or degradation of pathological tau aggregates, and catalyze posttranslational modifications of tau oligomers and associated proteins.

Lactifluus subg. Lactariopsis (Russulales/Basidiomycota) from Brazilian tropical forests

Lactifluus subg. Lactariopsis is a well-supported group of milkcaps occurring predominantly in tropical and subtropical ecosystems. This subgenus contains agaricoid and pleurotoid representatives and is the only group with members containing a secondary velum. The results of taxonomic and phylogenetic research into Lactifluus subg. Lactariopsis collected in Brazilian tropical forests (Amazon, Atlantic Rainforests, and Caatinga biomes) of the North and Northeast is presented here. Two new species are described: Lf. aurantioceraceus in sect. Neotropici, and Lf. reniformis in sect. Multicipites, newly delimited here . In addition, Lf. neotropicus has had its distribution expanded to the Caatinga biome, Lf. dunensis was confirmed in Lf. sect. Neotropici with new records from French Guiana, and Lf. annulifer was re-collected from the type locality. A combined dataset of nuclear rDNA internal transcribed spacer region and nuclear large subunit amplified sequences was analyzed and, based on its phylogenetic hypothesis, the two novel taxa and the new section were confirmed .

Stock Price Prediction Based on Daily News Headlines: Logistic Regression Model and LSTM Model

The influence of daily news on the economy, especially stock market, cannot be underestimated with big data gaining its momentum. In this article, the author uses two years of daily news headlines to predict the stock market movements of the Dow Jones Industrial Average. The first treatments on the dataset are collecting news from Kaggle.com and stock data from Yahoo Finance. The two datasets are then combined into one CSV file and split into training and test sets. Two machine learning models, Logistic Regression and Long Short-Term Memory, are built to fit the combined dataset, and the test index is the accuracy of prediction. The test accuracy is 0.58 with the three-word phrase by Logistic Regression and 0.65 after ten times training with the LSTM model. The final result demonstrates that the two models are feasible and effective for seeking the relationship between daily news and stock market movements and, thus, valuable for stock prediction. The attempts to set parameters give reference to further study, especially the word count of phrases and the number of training circulation.

Dementias Platform UK: Bringing genetics into life

AbstractBackgroundDPUK is a data repository bringing together a wide range of population and clinical cohorts from the UK, Europe and South Korea. It enables data discovery, variable selection, and data access for multi‐modal, multi‐cohort analysis in a shared and secure environment. Neurodegenerative dementias are group of diseases with highly heterogeneous pathology and with an overlapping genetic component that is poorly understood. Combing genetic data from different studies is important due to the increase in power this provides, but this presents an additional challenge due to differences in genotyping arrays and lack of overlapping variants.MethodWe have curated and harmonized genome‐wide data from 6 studies (Brains for Dementia Research, EPIC Norfolk, Generation Scotland, Airwave‐chip 1, Airwave‐chip 2, MRC National Survey of Health Development (NSHD)) within the DPUK platform. We created a pipeline that performs rigorous quality‐control (QC) analysis and imputes genetic variance with the 1000 Genome reference panel using the minimac algorithm. Further, standard pruning and thresholding Polygenic Risk Score (PRS) have been generated with 5 summary statistics related to neurodegenerative diseases (clinical AD, clinical/proxy AD, FTD, ALS, PD) for all 6 studies separately and for the combined dataset.ResultResearch‐ready imputed genetic data that have undergone QC and the PRS, based upon the latest neurodegenerative GWAS, are available for the 6 cohorts separately, and for the combined dataset of 60,670 individuals.ConclusionPreparing multiple datasets to a common standard for research‐readiness increases scientific opportunity and allows the wider research community to access and process data at scale and pace.

Equibiaxial flexural strength determination of UO2 using a ball-on-ring test

To increase nuclear fuel performance and reliability, their mechanical properties require an accurate and statistically relevant assessment. Biaxial flexural strength tests provide an alternative to bend bar techniques for assessing mechanical behavior; namely, the transverse rupture strength (TRS) of ceramic samples. Biaxial test samples require simple geometries and minimal surface preparation, reducing fabrication costs and handling hazards. This study investigated the TRS of polycrystalline UO2 fuel forms at room temperature using a ball-on-ring test fixture. Pellets were fabricated from UO2 powder using conventional powder processing and sintering techniques. The TRS and Weibull parameters were obtained through Weibull statistics on over 60 UO2 samples tested under equibiaxial flexure. The larger sample size in this study enabled a more robust Weibull statistical analysis than alternative test methods, which may not capture the stochastic failure of typical ceramics. Furthermore, two different loading ball diameters were employed to assess the impact of contact damage on fracture strength. While Hertzian contact damage was observed with the smaller loading ball, the fracture strength remained unaffected. A fracture analysis of the tested UO2 samples indicated a mixture of intergranular and transgranular fracture that transitions to transgranular fracture with increasing distance from the fracture origin. The characteristic strength of the combined data sets was determined to be 148 MPa, and the Weibull modulus was determined to be 9.1. The TRS values and Weibull parameters were close to values found in the literature for alternative testing techniques using samples with similar microstructure and density. The findings in this study validate the ball-on-ring method used to obtain the TRS of UO2 with a sample geometry more representative of nuclear fuels. Additionally, experimental TRS results from this study can be implemented in modeling codes to predict fuel performance, which is critical to fuel burnup extension and advanced nuclear fuel technologies.

Machine Learning for Identification of High-Risk Clonal Haematopoiesis Using Blood Count Data

Introduction: Clonal haematopoiesis (CH) is a common, age-related phenomenon associated with an increased risk of myeloid malignancies (MM), raising the prospect of screening to identify individuals at high risk of progression from CH to MM. The nature of the driver mutation(s) is the major determinant of the risk of progression to MM. Presently, the identification of these mutations relies on next generation sequencing (NGS) of blood DNA, making it impractical to identify those at risk at scale. We previously reported that the presence of CH is associated with changes in complete blood count (CBC) parameters such as the red cell distribution width (RDW), mean corpuscular volume (MCV) and others. However, individual parameters have limited discriminative power to distinguish those with CH versus those without. Here, we present a machine learning (ML) pipeline applied to a combined CBC dataset, which can be harnessed to improve our ability to identify individuals with high likelihood of carrying CH, who can subsequently be prioritised for genetic testing. Methods: Whole exome sequencing (WES) data from 454,052 UK Biobank (UKB) participants were analysed using Mutect2 to detect CH-driver mutations across the exons of 43 CH-associated genes. Germline variants were filtered using a panel of ormal and a one-sided exact binomial test. A final set of driver mutations was derived by filtering variants based on a pre-defined list of recurrent CH mutations or recurrence of &amp;gt;= 7 times in haematological cancers in the catalogue of somatic mutations in cancer. Fields pertaining to blood count variables were extracted and filtered to remove individuals with missing data, extreme outliers, highly correlated parameters and variables with near zero-variance resulting in 434,057 participants with complete datasets. Individuals were annotated as “CH” or “no-CH”; a Random Forest (RF) binary classifier was optimised on age, sex and 13 blood count variables as features. To quantify the robustness and stability of the model, the RF optimisation was iterated ten times, on a customised cross validation approach by using a different random sample from the majority (control) class each time. All ML models were constructed using the Caret package (R version 3.6.3). Results: We first found that RF classifiers to identify the presence of CH overall performed poorly (median AUC: 0.640/0.671 for any/large clone VAF &amp;gt;=10% CH respectively). However, gene-specific models showed that the presence of CH associated with mutations in certain high risk genes ( SF3B1, SRSF2, JAK2, CALR) could be predicted from blood counts more confidently (median AUC 0.776, 0.834, 0.933, 0.849 respectively). By contrast, mutations in the more common CH driver genes ( DNMT3A and TET2) could not be accurately predicted from blood count data (median AUC 0.618 and 0.666 respectively) (Figure 1). In light of these findings, we proceeded to build a single RF model to predict the presence of high-risk CH associated with mutations in any of SF3B1, SRSF2, JAK2 or CALR. This performed well, with median AUC of 0.816/0.888 for any/large clone high-risk CH respectively (Figure 2). Our RF models were also able to give insights into novel blood count association with specific types of CH; in particular we found that the basophil count was an important predictor of the presence of CH driven by mutation in codon K57 of the GNB1 gene. To investigate this further, we analysed basophil counts in the UKB and confirmed that basophilia (basophils &amp;gt;0.2 x 10 9/L) was more common amongst carriers of GNB1-K57 mutations (14/234 = 5.9%) than controls (2597/433,863 = 0.5%). Conclusions: Data-driven ML models applied to blood count data can identify individuals with high-risk CH driven by genes associated with progression to myelodysplastic syndrome and myeloproliferative neoplasms, who should be prioritised for genetic testing for somatic mutations. The future integration of raw blood analyser and longitudinal blood count data into predictive models would be an important next step towards improving our ability to identify individuals that may harbour CH and help facilitate the introduction of large-scale screening for the phenomenon.

Immunological Footprints of Pediatric Acute Myeloid Leukemia in Local Bone Marrow T Cells

Introduction: Despite notable advancements, childhood acute myeloid leukemia (AML) remains associated with one of the poorest prognoses among pediatric malignancies. Conventional chemotherapies carry significant toxicity, underscoring the urgent need for novel e.g. immunotherapy-based approaches. Despite AML being previously considered non-immunogenic due to its low mutational burden, the complex interplay between leukemic blasts and T cells within the bone marrow in pediatric patients remains largely unexplored. This study investigates the AML-induced footprint on local bone marrow T cells (bmT cells) representing tumor-infiltrating lymphocytes, with the aim to reveal functional points of action for future immunotherapy. Methods: Cryopreserved bone marrow samples from both pediatric AML patients (n=29) and age-matched healthy bone marrow donors (HD, n=9) were analyzed. Multicolor flow cytometry quantified surface expression of inhibitory signaling and activation receptors, as well as T cell differentiation stages. RNA-Seq and Assay for Transposase-Accessible Chromatin using sequencing (ATAC-Seq) were performed of sorted CD8 + bmT cells. Bulk RNA and ATAC-Seq data was visualized by principal component analysis (PCA). Another PCA was performed on the combined dataset of RNA- and ATAC-Seq. Differential analysis using DESeq2 (fitType “parametric”, testType “Wald”) identified 135 differentially expressed genes in CD8 + bmT cells between HD and AML patients based on fold change (≥1/≤-1) and statistical criteria (s-value&amp;lt;0.05). Results: Based on CD45RO, CD62L and CD95 surface expression, late differentiation stages of bmT cells were enriched in AML. Namely, the frequency of naïve bmT cells was reduced (33.9% vs. 57.5%, mean values, p=0.0018), while the frequency of effector memory bmT cells was increased in AML patients (21.5% vs. 9.4%, p=0.0010). This could be validated by RNA-Seq of sorted CD8 + bmT cells which documented higher expression of cytotoxic effector T cell genes such as Perforin, Granulysin and Granzyme B. In addition, our analysis also revealed higher expression of functional relevant genes associated with T cell cytotoxity (ADGRG1), effector function (BATF/TPX2) or exhaustion (SLAMF7) in AML patients. Next, surface expression of receptors for inhibitory signaling and activation was analyzed. Several of these receptors have also been described as immune checkpoints (ICP). Analyzing differential expression between HD and AML patients showed pronounced upregulation of inhibitory markers (TIM-3: 7.1% vs. 1.7%, p&amp;lt;0.0001, CD39: 7.7% vs. 2.0%, p=0.0103, LAG-3: 8.0% vs. 2.9%; CTLA-4: 3.9% vs. 1.6%, p=0.0024; PD-1: 34.8% vs. 21.3%, p=0.0016). Since CD8 + bmT cells showed the most promising effects in the flow cytometry analysis, this subpopulation of bmT cells was further characterized by RNA- and ATAC-Seq. Here, clustering of HD and AML patients' CD8 + bmT cells was observed in PCA. Interestingly, PCA also revealed differential clustering of CD8 + bmT cells from AML primary and relapse samples. Confirming the observed differences in phenotype of bmT cells in AML patients, we identified distinct bmT cell populations specifically induced in AML patients compared to HD, based on phenotype and differential ICP expression patterns. Through tSNE analysis, we discovered a CD8 + effector memory T cell population with a TIM-3 +PD-1 +2B4 + phenotype, comprising 2.3% of all bmT cells in AML samples compared to only 0.03% in HD. Conclusion: This study reveals significant alterations in the activation and inhibition potential of bmT cells in pediatric AML compared to HD, as confirmed by comprehensive analysis of the epigenome, transcriptome and surfaceome. We confirm the observed differences through the identification of a T cell population with a typically exhausted phenotype specific for AML patients when compared to HD, implying persistent antigen exposure and potential direct or indirect interactions with leukemic blasts. Although prospective clinical trials will have to confirm functional relevance in patient cohorts, this study provides crucial insights into the immune landscape of pediatric AML, underscoring the potential for targeted immunotherapeutic interventions.

Cryoprecipitate-Poor Plasma Instead of Fresh Frozen Plasma As Replacement Therapy in Thrombotic Thrombocytopenic Purpura: A Systematic Review and Meta-Analysis

Background: Thrombotic thrombocytopenic purpura (TTP) is a life-threatening condition characterized by the presence of small-vessel thrombi enriched with platelets, leading to thrombocytopenia, thrombotic microangiopathy (TMA), and on occasion, resulting in end-organ damage. The standard therapeutic plasma exchange (TPE) approach, initiated after suspicion of TTP, is done with fresh frozen plasma (FFP). Yet, the use of cryoprecipitate-poor plasma (CPP) has been suggested as alternative in recent decades given its absence of von Willebrand factor, which would benefit the pathophysiology of TTP. However, its efficacy over FFP remains unclear. Therefore, we aimed to conduct a meta-analysis to compare the safety and efficacy of CPP versus FFP in patients with TTP. Methods: We systematically searched PubMed, Cochrane Central and Embase for clinical studies published up to June 2023, comparing CPP to FFP as replacement fluids in TPP or TMA patients. Outcomes assessed were overall mortality, relapse rate, response to treatment and number of TPE sessions. We pooled outcomes as odds ratios (OR) and mean differences (MD) with 95% confidence intervals (CI). We also performed a subgroup analysis with randomized controlled trials (RCTs). Statistical analysis was performed with Review Manager version 5.4.1. Heterogeneity was assessed with I 2 statistics. The protocol was registered in PROSPERO (ID: CRD42023440665). Results: A total of eight articles met inclusion criteria, of which three were RCTs. The combined dataset involved 290 patients, of which 144 (49.7%) were in CPP group and 146 (50.3%) in the FFP group. CPP significantly reduced overall mortality compared to FFP (OR 0.27; 95% CI 0.13-0.59; p=0.001; I 2=0%; Figure 1A). However, our subgroup analysis of RCTs showed no significant difference in overall mortality between groups (OR 0.49; 95% CI 0.13-1.84; p=0.29; I 2=0%; Figure 1B). Additionally, there were no significant differences between groups regarding relapses (OR 0.81; 95% CI 0.37-1.77; p=0.60; I 2=0%; Figure 1C), response to treatment (OR 2.20; 95% CI 0.69-7.07; p=0.18; I 2=44%; Figure 1D) or number of TPE sessions (MD 1.63; 95% CI -2.83 - 6.08; p=0.47; I 2=55%; Figure 1E). Conclusion: Our meta-analysis shows that the safety and efficacy of CPP in TTP patients is at least similar to FFP in reducing mortality, relapse rate, and improving response to treatment. The reduced mortality with CPP found in the whole pool of studies, along with the trend towards reduced mortality found in the RCT-only analysis indicates the potential benefit of CPP over FFP. However, larger population-based randomized trials should contribute to narrow confidence intervals and decrease bias of observational studies, providing clearer conclusions for treatment efficacy in the future.

Dynamic model-assisted transferable network for liquid rocket engine fault diagnosis using limited fault samples

The accurate detection and diagnosis of faults in Liquid Rocket Engines (LREs) are critical for ensuring space mission safety. However, the limited availability of actual fault samples and the diversification of potential faults present significant challenges in achieving precise diagnosis. To overcome these obstacles, we propose a dynamic model-assisted transfer learning approach. In this study, we first modularize the LRE and establish a dynamic model based on the mathematical principles of each module. Subsequently, we artificially established a fault module model, injected faults into the normal model, and simulated various fault modes to expand the fault sample library. Leveraging this augmented dataset in combination with a limited number of actual fault samples, we employ transfer learning to fine-tune a Convolutional Neural Network (CNN). Compared with other classic methods, the migrated CNN effectively adapts to the distribution of real data and significantly improves the accuracy of LRE fault diagnosis.

Right or left-handed: are locomotion type, body size, and microhabitat related to forelimb laterality in anurans?

Laterality, the tendency of animals to use the structures on one side of the body more frequently or efficiently than the other side, can reflect the evolutionary history of species, their ecological characteristics, and cognitive abilities. Using anurans as a model system, it has been hypothesized that the strength of laterality (measured as the percentage of right-handed individuals) is higher in terrestrial and arboreal species, which exhibit asynchronous locomotion type (e.g., walking, climbing), than in aquatic species where synchronous locomotion type (e.g., swimming) predominates. We tested these predictions, including body size as an additional factor, in 19 Neotropical anuran species. For this, we obtained data on forelimb use preference for individuals of such species. We then combined these data with published data for another 14 anuran species obtained through a systematic review of the literature. The combined dataset was used to perform a phylogenetic comparative analysis to verify the strength of laterality in the species. We found that four out of the 19 tested species show a percentage of right-handed individuals that was higher or lower than expected by chance. Individuals of these four species exhibit asynchronous movements of forelimbs during displacements, foraging behaviors, sexual displays, or agonistic interactions. Results of a phyloANOVA and a phyloANCOVA indicated that the strength of laterality was higher in species with asynchronous movements, but this was unrelated to microhabitat and body size. Asynchronous (i.e., non-simultaneous) movements of limbs imply alternating muscle contractions and a higher neural asymmetry than synchronous (i.e., simultaneous) movements of limbs. Therefore, species with an asynchronous locomotion type could be reflecting such complexity and neural asymmetry in higher levels of forelimb laterality. Our results mostly support the predictions based on models that cover the expected relationship between the strength of laterality and the evolutionary ecology of vertebrates.

Improving mixed-integer temporal modeling by generating synthetic data using conditional generative adversarial networks: A case study of fluid overload prediction in the intensive care unit

ObjectiveThe challenge of mixed-integer temporal data, which is particularly prominent for medication use in the critically ill, limits the performance of predictive models. The purpose of this evaluation was to pilot test integrating synthetic data within an existing dataset of complex medication data to improve machine learning model prediction of fluid overload. Materials and methodsThis retrospective cohort study evaluated patients admitted to an ICU ≥ 72 h. Four machine learning algorithms to predict fluid overload after 48–72 h of ICU admission were developed using the original dataset. Then, two distinct synthetic data generation methodologies (synthetic minority over-sampling technique (SMOTE) and conditional tabular generative adversarial network (CTGAN)) were used to create synthetic data. Finally, a stacking ensemble technique designed to train a meta-learner was established. Models underwent training in three scenarios of varying qualities and quantities of datasets. ResultsTraining machine learning algorithms on the combined synthetic and original dataset overall increased the performance of the predictive models compared to training on the original dataset. The highest performing model was the meta-model trained on the combined dataset with 0.83 AUROC while it managed to significantly enhance the sensitivity across different training scenarios. DiscussionThe integration of synthetically generated data is the first time such methods have been applied to ICU medication data and offers a promising solution to enhance the performance of machine learning models for fluid overload, which may be translated to other ICU outcomes. A meta-learner was able to make a trade-off between different performance metrics and improve the ability to identify the minority class.

Machine Learning for Film Thickness Prediction in Elastohydrodynamic Lubricated Elliptical Contacts

This study extends the use of Machine Learning (ML) approaches for lubricant film thickness predictions to the general case of elliptical elastohydrodynamic (EHD) contacts, by considering wide and narrow contacts over a wide range of ellipticity and operating conditions. Finite element (FEM) simulations are used to generate substantial training and testing datasets that are used within the proposed ML framework. The complete dataset entails 915 samples; split into an 823-sample training dataset and a 92-sample testing dataset, corresponding to 90% and 10% of the combined dataset samples, respectively. The proposed ML model consists of a pre-processing stage in which conventional EHD dimensionless groups are used to minimize the number of inputs into the model, reducing them to only three. The core of the model is based on Gaussian Process Regression (GPR), a powerful ML regression tool, well-suited for small-sized datasets, producing output central and minimum film thicknesses, also in dimensionless form. The last stage is a post-processing one, in which the output film thicknesses are retrieved in dimensional from. The results reveal the capabilities and potential of the proposed ML framework, producing quasi-instantaneous predictions that are far more accurate than conventional film thickness analytical formulae. In fact, the produced central and minimum film thickness predictions are on average within 0.3% and 1.0% of the FEM results, respectively.

Machine learning insights of anthropogenic and natural influences on riverbed deformation in a large lowland river

As escalating environmental pressures threaten the world's river systems, understanding the driving factors influencing their geomorphological changes is of critical global importance. This study illuminates the complex interaction between natural and anthropogenic factors that drove geomorphological changes along the upper Apalachicola River (UAR) from 1960 to 2010. This study utilized LiDAR point cloud data and hydrographic surveys conducted between 1960 and 2010 to identify the primary factors driving riverbed changes. A triangular irregular network (TIN) was used to convert the combined datasets from each survey, following coordinate transformation, into a 1.5 m-resolution digital elevation model (DEM) raster. The DEM of difference (DEMoD) was computed by subtracting the 1960 DEM from that of 2010, and the total sediment loss and gain were then determined by multiplying the DEMoD by the DEM resolution. A random forest (RF) regression model was used to determine the most influential factors contributing to riverbed deformation. The results revealed substantial differences in sediment loss and gain across the river. The UAR, divided into three segments, experienced sediment loss of about −6.56 × 106 m3 between 1960 and 2010, with the farthest upstream section (RM 106 to 98) losing −3.14 × 106 m3, which is ~48 % of the sediment losses in the study area. Some sections of the channel have deepened significantly, with a pronounced depth difference of −17.20 m between RM 92 and 94, potentially caused by the presence of limestone and the enlargement of a sinkhole or spring on the riverbed. The RF model identified the dam (>28 %) as the most crucial anthropogenic factor in riverbed deformation, followed by the presence of dikes (~17—18 %), then historical dredging (~7—8 %), highlighting the substantial impact of artificial structures on riverbed deformation. In contrast, rock removal, while potentially impactful in certain locations, had negligible effects (>0.5 %) at the system scale. This study highlights the interaction of natural and anthropogenic factors in river geomorphology and the potential for machine learning to identify these dynamics and inform data-driven river management and conservation strategies.

Towards a generalized hybrid deep learning model with optimized hyperparameters for malicious traffic detection in the Industrial Internet of Things

Detecting malicious attacks in Industrial Internet of Things (IIoT) is crucial to minimize downtime and financial losses. However, existing deep learning (DL) research faces limitations in reliability and generalization due to reliance on a single dataset. This single dataset approach hinders the effectiveness of the models when applied to new and unseen datasets. This paper addresses the issue using multiple-domain learning, creating a generalized DL framework for IIoT traffic classification. It combines three datasets: Edge-IIoTSet, WUSTL-IIoT-2021, and X-IIoTID. The proposed framework has two stages. The first stage involves the fusion of different datasets into a common feature space. To achieve this, an autoencoder architecture is proposed to match the dimensionality of the datasets into a common feature space. Subsequently, a modified locally linear embedding is used as a manifold alignment to ensure statistical matching of multiple datasets such that the maximum mean discrepancy between multiple datasets is minimized to compare and combine datasets effectively. The second stage presents a hybrid DL model with a convolutional neural network (CNN) and gated recurrent unit (GRU) for IIoT traffic classification. Bayesian optimization fine-tunes hyperparameters. Experiments on combined, single, and cross-datasets show the superiority of the framework. The CNN-GRU model achieved 97.68% accuracy, 97.70% recall, 97.67% precision, and 97.68% F1-score for binary classification on the combined dataset. Transfer learning improved accuracy from 50.95% to 97.80% and F1-score from 48.52% to 97.79% when trained on X-IIoTID and tested on Edge-IIoTSet. Multi-classification accuracy and weighted averages of recall, precision, and F1-score were approximately 98.85%.