Search

Federated learning is a renowned technique for utilizing decentralized data while preserving privacy. However, real-world applications often face challenges like partially labeled datasets, where only a few locations have certain expert annotations, leaving large portions of unlabeled data unused. Leveraging these could enhance transformer architectures' ability in regimes with small and diversely annotated sets. We conduct the largest federated cardiac CT analysis to date (n = 8, 104) in a real-world setting across eight hospitals. Our two-step semi-supervised strategy distills knowledge from task-specific CNNs into a transformer. First, CNNs predict on unlabeled data per label type and then the transformer learns from these predictions with label-specific heads. This improves predictive accuracy and enables simultaneous learning of all partial labels across the federation, and outperforms UNet-based models in generalizability on downstream tasks. Code and model weights are made openly available for leveraging future cardiac CT analysis. [ABSTRACT FROM AUTHOR]

Background: The diagnosis of rare uterine leiomyosarcoma (uLMS) remains a challenge given the high incidence rates of benign uterine tumors such as leiomyoma (LM). In the last decade, several clinical scores and blood serum markers have been proposed. The aim of this study is to validate and update the pLMS clinical scoring system, evaluating the accuracy of the scoring system by Zhang et al. and examining the discriminatory ability of blood markers such as serum lactate dehydrogenase (LDH), neutrophil-to-lymphocyte ratio (NLR), and platelet-to-lymphocyte ratio (PLR). Methods: In a case-control study, 90 new uLMS from the DKSM consultation registry and 659 prospectively recruited LM cases from the Saarland University Hospital were used for validation. Welch's t-test and Hedges' g were used to evaluate blood markers and optimal thresholds and diagnostic odds ratios were calculated. Scoring systems were compared using receiver operating characteristics and proposed diagnostic cut-offs were reviewed. Missing values were imputed by random forest imputation to create the updated scoring system 'pLMS2' using penalized logistic regression based on the pooled data sets of 384 uLMS and 1485 LM. Results: pLMS achieved an AUC of 0.97 on the validation data, but sensitivity and specificity varied at the proposed thresholds due to a shift in the score distributions. 43 uLMS and 578 LM were included in the comparison of pLMS with Zhang's scoring system, with pLMS being superior (AUC 0.960 vs 0.845). LDH, NLR, and PLR achieved a diagnostic odds ratios of 18.03, 8.64 and 4.81, respectively. pLMS2 is based on subscores for menopausal status interacting with age, tumor diameter, intermenstrual bleeding, hypermenorrhea, dysmenorrhea, postmenstrual bleeding, rapid tumor growth, and suspicious sonography. Conclusions: Validation of the pLMS shows stable discriminatory ability as expressed by AUC, although caution should be taken with cut-off values, as sensitivity and specificity may vary. Data collection of the updated clinical score pLMS2 remains simple and convenient, with no additional cost. The proposed thresholds of 1.5 and 5.5 can be used as a guide to avoid unnecessary or inappropriate surgery and to make the use of further diagnostic measures cost-effective. LDH, NLR and PLR provide further evidence to differentiate uLMS from LM in conjunction with clinical data. [ABSTRACT FROM AUTHOR]

The data presented in this article are related to the research paper entitled "The biological age linked to oxidative stress modifies breast cancer aggressiveness" (M.M. Sáez-Freire, A. Blanco-Gómez, S. Castillo-Lluva, A. Gómez-Vecino, J.M. Galvis-Jiménez, C. Martín-Seisdedos, M. Isidoro-García, L. Hontecillas-Prieto, M.B. García-Cenador, F.J. García-Criado, M.C. Patino-Alonso, P. Galindo-Villardón, J.H. Mao, C. Prieto, A. Castellanos-Martín, L. Kaderali, J. Pérez-Losada). The data shown were obtained from a population of transgenic mice, MMTV-Erbb2/Neu, with different susceptibility to breast cancer and a mixed genetic background generated by backcrossing. It was observed that the aggressiveness of breast cancer negatively correlates with age, being lower in chronologically old mice, similar to what occurs in humans. Given that oxidative stress is associated with tumour susceptibility and the degree of aging, the association between the aggressiveness of breast cancer and multiple intermediate phenotypes directly or indirectly related to oxidative stress was studied. Using a mathematical model, we defined biological age and the degree of aging as the difference between biological and chronological ages. As a result, we observed that biologically old mice predominated among those that developed the disease early on, that is, those that were chronologically young. We then identified the specific and common genetic components of Quantitative Trait loci or QTL associated with different evolution of breast cancer, the intermediate phenotypes related to oxidative stress studied, the biological age and the degree of aging. Lastly, we showed that the expression pattern in the livers of biologically old mice were enriched in signalling pathways related to inflammation and response to infections; whereas the biologically young mice exhibited enriched pathways related to mitochondrial activity. For the explanation and discussion of these data refer to the research article cited above.

The data presented in this article are related to the research paper entitled "The biological age linked to oxidative stress modifies breast cancer aggressiveness" (M.M. Sáez-Freire, A. Blanco-Gómez, S. Castillo-Lluva, A. Gómez-Vecino, J.M. Galvis-Jiménez, C. Martín-Seisdedos, M. Isidoro-García, L. Hontecillas-Prieto, M.B. García-Cenador, F.J. García-Criado, M.C. Patino-Alonso, P. Galindo-Villardón, J.H. Mao, C. Prieto, A. Castellanos-Martín, L. Kaderali, J. Pérez-Losada). The data shown were obtained from a population of transgenic mice, MMTV-Erbb2/Neu, with different susceptibility to breast cancer and a mixed genetic background generated by backcrossing. It was observed that the aggressiveness of breast cancer negatively correlates with age, being lower in chronologically old mice, similar to what occurs in humans. Given that oxidative stress is associated with tumour susceptibility and the degree of aging, the association between the aggressiveness of breast cancer and multiple intermediate phenotypes directly or indirectly related to oxidative stress was studied. Using a mathematical model, we defined biological age and the degree of aging as the difference between biological and chronological ages. As a result, we observed that biologically old mice predominated among those that developed the disease early on, that is, those that were chronologically young. We then identified the specific and common genetic components of Quantitative Trait loci or QTL associated with different evolution of breast cancer, the intermediate phenotypes related to oxidative stress studied, the biological age and the degree of aging. Lastly, we showed that the expression pattern in the livers of biologically old mice were enriched in signalling pathways related to inflammation and response to infections; whereas the biologically young mice exhibited enriched pathways related to mitochondrial activity. For the explanation and discussion of these data refer to the research article cited above.

The incidence of breast cancer increases with age until menopause, and breast cancer is more aggressive in younger women. The existence of epidemiological links between breast cancer and aging indicates that both processes share some common mechanisms of development. Oxidative stress is associated with both cancer susceptibility and aging. Here we observed that ERBB2-positive breast cancer, which developed in genetically heterogeneous ERBB2-positive transgenic mice generated by a backcross, is more aggressive in chronologically younger than in older mice (differentiated by the median survival of the cohort that was 79 weeks), similar to what occurs in humans. In this cohort, we estimated the oxidative biological age using a mathematical model that integrated several subphenotypes directly or indirectly related to oxidative stress. The model selected the serum levels of HDL-cholesterol and magnesium and total AKT1 and glutathione concentrations in the liver. The grade of aging was calculated as the difference between the predicted biological age and the chronological age. This comparison permitted the identification of biologically younger and older mice compared with their chronological age. Interestingly, biologically older mice developed more aggressive breast cancer than the biologically younger mice. Genomic regions on chromosomes 2 and 15 linked to the grade of oxidative aging were identified. The levels of expression of Zbp1 located on chromosome 2, a gene related to necroptosis and inflammation, positively correlated with the grade of aging and tumour aggressiveness. Moreover, the pattern of gene expression of genes linked to the inflammation and the response to infection pathways was enriched in the livers of biologically old mice. This study shows part of the complex interactions between breast cancer and aging.

The Explainable Modular Neural Network (XModNN) enables the identification of biomarkers, facilitating the classification of diseases and clinical parameters in transcriptomic datasets. The modules within XModNN represent specific pathways or genes of a functional hierarchy. The incorporation of biological insights into the architectural design reduced the number of parameters. This is further reinforced by the weighted multi-loss progressive training, which enables successful classification with a reduced number of replicates. The combination of this workflow with layer-wise relevance propagation ensures a robust post hoc explanation of the individual module contribution. Two use cases were employed to predict sex and neuroblastoma cell states, demonstrating that XModNN, in contrast to standard statistical approaches, results in a reduced number of candidate biomarkers. Moreover, the architecture enables the training on a limited number of examples, attaining the same performance and robustness as support vector machine and random forests. The integrated pathway relevance analysis improves a standard gene set overrepresentation analysis, which relies solely on gene assignment. Two crucial genes and three pathways were identified for sex classification, while 26 genes and six pathways are highly important to discriminate adrenergic–mesenchymal cell states in neuroblastoma cancer. [ABSTRACT FROM AUTHOR]

Background: Limited data on SARS-CoV-2 seroprevalence in rural areas of northern Germany necessitate comprehensive cohort studies. We aimed to evaluate the seroprevalence, silent infection (SI) rates and risk factors for infections among children and adolescents in Western Pomerania from December 2020 to August 2022. Methods: In this cross-sectional study, serum or plasma samples from children and adolescents (6 months to 17 years) were collected during routine blood draw. SARS-CoV-2 specific antibodies (S1 and nucleocapsid) and their neutralizing capacity were analyzed using commercially available enzyme-linked immunosorbent and neutralization assays. Socio-demographic data and information about SARS-CoV-2 infection or vaccination were obtained. Multivariable logistic regression was used to identify independent risk factors for SARS-CoV-2 infections and SI. Results: A total of 1,131 blood samples were included into the study. Overall, SARS-CoV-2 seroprevalence was 25.1%, strongly influenced by the pandemic course, predominant virus variants, age and approval of vaccination. SI rate was 5.4% (95%-CI [3.7%–6.8%]) among unvaccinated and undiagnosed children over the entire study period with highest rates among adolescents. Main risk factor despite the time at risk for silent infections was an infected household member (Odds ratio = 9.88, 95%-CI [4.23–22.9], p < 0.001). Factors associated with overall infections (known and silent) also include the infection of a household member (Odds ratio = 17.8, 95%-CI [10.7–29.6], p < 0.001). Conclusions: We believe that the introduction of governmental measures and systematic test strategies in schools strongly impacted on the SI rate, as we suspect that asymptomatic cases have already been identified, resulting in surprisingly low SI identified in our study. [ABSTRACT FROM AUTHOR]

Malaria, caused by protozoan Plasmodium parasites, remains a prevalent infectious human disease due to the lack of an efficient and safe vaccine. This is directly related to the persisting gaps in our understanding of the parasite's interactions with the infected host, especially during the clinically silent yet essential liver stage of Plasmodium development. Previously, we and others showed that genetically attenuated parasites (GAP) that arrest in the liver induce sterile immunity, but only upon multiple administrations. Here, we comprehensively studied hepatic gene and miRNA expression in GAP-injected mice, and found both a broad activation of IFNγ-associated pathways and a significant increase of murine microRNA-155 (miR-155), that was especially pronounced in non-parenchymal cells including liver-resident macrophages (Kupffer cells). Remarkably, ectopic upregulation of this miRNA in the liver of mice using robust hepatotropic adeno-associated virus 8 (AAV8) vectors enhanced GAP's protective capacity substantially. In turn, this AAV8-mediated miR-155 expression permitted a reduction of GAP injections needed to achieve complete protection against infectious parasite challenge from previously three to only one. Our study highlights a crucial role of mammalian miRNAs in Plasmodium liver infection in vivo and concurrently implies their great potential as future immune-augmenting agents in improved vaccination regimes against malaria and other diseases.