Manifold Approximation Research Articles

Surface-Enhanced Raman Scattering Combined with Machine Learning for Rapid and Sensitive Detection of Anti-SARS-CoV-2 IgG

This work reports an efficient method to detect SARS-CoV-2 antibodies in blood samples based on SERS combined with a machine learning tool. For this purpose, gold nanoparticles directly conjugated with spike protein were used in human blood samples to identify anti-SARS-CoV-2 antibodies. The comprehensive database utilized Raman spectra from all 594 blood serum samples. Machine learning investigations were carried out using the Scikit-Learn library and were implemented in Python, and the characteristics of Raman spectra of positive and negative SARS-CoV-2 samples were extracted using the Uniform Manifold Approximation and Projection (UMAP) technique. The machine learning models used were k-Nearest Neighbors (kNN), Support Vector Machine (SVM), Decision Trees (DTs), logistic regression (LR), and Light Gradient Boosting Machine (LightGBM). The kNN model led to a sensitivity of 0.943, specificity of 0.9275, and accuracy of 0.9377. This study showed that combining Raman spectroscopy and a machine algorithm can be an effective diagnostic method. Furthermore, we highlighted the advantages and disadvantages of each algorithm, providing valuable information for future research.

Exploring tumor endothelial cells heterogeneity in hepatocellular carcinoma: insights from single-cell sequencing and pseudotime analysis

Objective This study aimed to explore the heterogeneity of tumor endothelial cells (TECs) in hepatocellular carcinoma (HCC) and their role in tumor progression, with the goal of identifying new therapeutic targets and strategies to improve patient prognosis. Methods Single-cell RNA sequencing data from nine primary liver cancer samples were analyzed, obtained from the Gene Expression Omnibus (GEO) database. Data preprocessing, normalization, dimensionality reduction, and batch effect correction were performed based on the Seurat package. HCC cell types were identified using uniform manifold approximation and projection (UMAP) and cluster analysis, and the different cell types were annotated using the CellMarker database. Pseudotime trajectory analysis was conducted with Monocle to explore the differentiation trajectory of TECs. MAPK signaling pathway activity and copy number variations (CNV) in TECs were analyzed in conjunction with data from The Cancer Genome Atlas (TCGA), the trans-well and wound healing assay was used for cell invasion and migration activity assessment. Results Two subgroups of TECs (TECs 1 and TECs 2) were identified, exhibiting distinct functional activities and signaling pathways. Specifically, TECs 1 may be involved in tumor cell proliferation and inflammatory responses, whereas TECs 2 is not only involved in cell proliferation pathways, but also enriched in pathways such as metabolic synthesis. Pseudotime analysis revealed dynamic changes in TECs subgroups during HCC progression, correlating specific gene expressions (such as PDGFRB, PGF, JUN, and NR4A1). Subsequently, the JUN gene was predicted by performing binding sites and was shown to act as a transcription factor that may regulate the expression of the PGF gene. CNV analysis highlighted key genes and pathways in TECs that might influence HCC progression, and the PGF as key regulatory factor mediated cell proliferation and migration. Conclusion The study revealed the heterogeneity of TECs in HCC and their potential roles in tumor progression, offering new perspectives and potential therapeutic targets for HCC molecular mechanisms. The findings emphasize the importance of further exploring TECs heterogeneity for understanding HCC pathogenesis and developing personalized treatment strategies.

Clustering of artificial intelligence derived plaque features for risk stratification for future atrial fibrillation development

Abstract Introduction We previously demonstrated that artificial intelligence (AI)-derived coronary artery plaque features are associated with future development of atrial fibrillation (AF) among patients undergoing CCTA(1). Our results suggest that phenotyping a large dataset of plaque features using data science techniques may reveal associations not visible from univariate analysis. Purpose Use data science analysis of AI-CT plaque characteristics to identify subpopulations at higher risk of AF. Methods We applied dimensionality reduction and visual cluster identification to phenotype our population of patients based on their AI-CT plaque features. We used a single center CCTA registry of patients with prior clinically-indicated CCTA (N=2700) from 2017-2020. Cases of new AF were defined by onset of AF on ECG or telemetry after CCTA via EMR search. Patients with AF prior to CCTA, history of mitral valve surgery, or postoperative AF were excluded. Controls were selected 1:1 by nearest neighbor propensity matching (R. v. 4.0.3.) based on age, sex, tobacco use, HTN, HLD, and diabetes. Atherosclerotic plaques were quantitatively analyzed using a commercially available FDA-approved AI coronary analysis software yielding stenosis area and diameter, remodeling index, volume of calcified/non-calcified/low-density plaques, and number of 2 or more high risk plaques. Uniform Manifold Approximation and Projection (UMAP) dimensionality reduction with cosine distance metric was used to reduce the plaque feature space from 80 dimensions (8 AI measurements per each of 10 anatomical vessel segments in all patients) to 2 dimensions on a scatter plot. Two non-contiguous clusters (i.e. plaque phenotypes) were identified based on visual analysis of UMAP feature reduction (Figure 1). Results 94 cases comprised the study cohort (47 cases and 47 controls). Mean age was 67±11 years. 38% were female. Mean CADRADS score was 1.8±1.52 and mean CAC was 325±622. Mean time from CT to AF was 281±300 days. The majority of AF patients (67%) fell into Cluster 1 (N=36), while Cluster 2 (N=58) contained most of the control subjects (75%). Treating Cluster 1 as a predictor of AF, the PPV and specificity of the UMAP reduction was 67% and 75% respectively, with NPV of 60% and sensitivity of 51% (Table 1). Left atrial volume index (LAVI) was significantly different between the 2 clusters, with mean 56.15±12.42 ml/m2 in Cluster 1 and 46.96±16.03 in Cluster 2 (p=0.0046). LAVI was also significantly different between control subjects in Cluster 1 vs 2, mean 54.84±11.6 vs 41.62±7.92 (p<0.001). Conclusions We demonstrated that UMAP-derived clustering of AI-CT plaque characteristics identifies novel phenotypes for future risk of AF among patients undergoing CCTA. Further, control patients in Cluster 1 were found to have a significantly higher LAVI compared to control subjects in Cluster 2, which supports that membership in Cluster 1 may increase risk of future development of AF.

Distributed Photovoltaic Communication Anomaly Detection Based on Spatiotemporal Feature Collaborative Modeling

As distributed photovoltaic (PV) technology rapidly develops and is widely applied, the methods of cyberattacks are continuously evolving, posing increasingly severe threats to the communication networks of distributed PV systems. Recent studies have shown that the Transformer model, which effectively integrates global information and handles long-distance dependencies, has garnered significant attention. Based on this, our research proposes a model named STformer, which is applied to the task of attack detection in distributed PV communication. Specifically, we propose a temporal attention mechanism and a variable attention mechanism. The temporal attention mechanism focuses on capturing subtle changes and trends in data sequences over time, ensuring a highly sensitive recognition of patterns inherent in time-series data. In contrast, the variable attention mechanism analyzes the intrinsic relationships and interactions between different variables, uncovering critical correlations that may indicate abnormal behavior or potential attacks. Additionally, we incorporate the Uniform Manifold Approximation and Projection (UMAP) dimensionality reduction technique. This technique not only helps reduce computational complexity but, in certain cases, can enhance anomaly detection performance. Finally, compared to classical and advanced methods, STformer demonstrates satisfactory performance in simulation experiments.

Elucidation of endogenous and exogenous chemicals in maternal serum using high-resolution mass spectrometry

The increasing exposure to environmental chemicals calls for comprehensive non-targeted analysis to detect unrecognized substances in human samples. We examined human serum samples to classify compounds as endogenous or exogenous using public databases and to explore the relationships between exposure markers and metabolic patterns. Serum samples from 84 pregnant women at 32 weeks gestation were analyzed using LC-QToFMS. Using the PubChemLite for Exposomics database, we annotated and classified 106 compounds (51 endogenous, 55 exogenous). The compound patterns were analyzed using three dimensional reduction methods: Principal Component Analysis (PCA), regularized Generalized Canonical Correlation Analysis (rGCCA), and Uniform Manifold Approximation and Projection (UMAP). OPTICS clustering applied to these methods revealed two distinct clusters, with 89 % of significant compounds overlapping between clusters. The detected exogenous compounds included dietary substances, phthalates, nitrogenous compounds, and parabens. Pathway enrichment analysis showed that chemical exposure was linked to changes in amino acid metabolism, protein and mineral transport, and energy metabolism. While we found associations between exposure and metabolite changes, we could not establish causality. Our approach of analyzing both exogenous and endogenous chemicals from the same dataset using PubChemLite database presents a new method for exposome research, despite limitations in sample size and peak annotation accuracy. These findings contribute to understanding multiple chemical exposures and their metabolic effects in human biomonitoring.

Optimizing Methane Uptake on N/O Functionalized Graphene via DFT, Machine Learning, and Uniform Manifold Approximation and Projection (UMAP) Techniques

Optimizing Methane Uptake on N/O Functionalized Graphene via DFT, Machine Learning, and Uniform Manifold Approximation and Projection (UMAP) Techniques

CIC/ATXN1-rearranged tumors in the central nervous system are mainly represented by sarcomas: A comprehensive clinicopathological and epigenetic series.

CIC fusions have been described in two different central nervous system (CNS) tumor entities. On one hand, fusions of CIC or ATXN1 genes belonging to the same complex of transcriptional repressors, were reported in the CIC-rearranged, sarcoma (SARC-CIC). The diagnosis of this tumor type, which was recently added to the World Health Organization (WHO) Classification of CNS tumors, is difficult mainly because the data concerning its histopathology (as compared to its soft tissue counterpart), immunoprofile, and clinical as well as radiological characteristics are scarce in the literature. On the other hand, a recent study, based on DNA-methylation profiling, has identified a novel high-grade neuroepithelial tumor characterized by recurrent CIC fusions (HGNET-CIC). The aim of this multicentric study was to characterize a cohort of 15 primary CNS tumors harboring a CIC or ATXN1 fusion in terms of clinical, radiological, histopathological, immunophenotypical, and epigenetic characteristics. According to the integrated diagnoses, 14/15 tumors corresponded to SARC-CIC, and only one to HGNET-CIC. The tumors showed similar clinical (mainly pediatric), radiological (mostly supratentorial, cystic, and contrast enhancing), immunophenotypical (common expression of glioneuronal markers), and genetic (similar spectrum of fusions) profiles but their histopathological appearance was clearly distinct. Moreover, we found a novel fusion transcript (CIC::EWSR1) in a SARC-CIC. Most DNA methylation profiles using the Heidelberg Brain Tumor Classifier (v12.8) annotated the samples to the methylation class "SARC-CIC" (9/14 tumors with available data). By using uniform manifold approximation and projection analysis, four other samples were classified as SARC-CIC and another clustered within the methylation class of HGNET-CIC. Our findings confirm that CNS CIC-fused tumors do not represent a single molecular tumor entity. Further analyses are needed to characterize HGNET-CIC in more detail. These results may help to refine the essential diagnostic criteria for SARC-CIC and their terminology (with a suggested consensual name of sarcoma, CIC/ATXN1-complex rearranged).

A Real-Time System Status Evaluation Method for Passive UHF RFID Robots in Dynamic Scenarios

In dynamic scenarios, the status of a Radio Frequency Identification (RFID) system fluctuates with environmental changes. The key to improving system efficiency lies in the real-time monitoring and evaluation of the system status, along with adaptive adjustments to the system parameters and read algorithms. This paper focuses on the status changes in RFID systems in dynamic scenarios, aiming to enhance system robustness and reading performance, ensuring high link quality, reasonable resource scheduling, and real-time status evaluation under varying conditions. This paper comprehensively considers the system parameter settings in dynamic scenarios, integrating the interaction model between readers and tags. The system’s real-time status is evaluated from both the physical layer and the Medium Access Control (MAC) layer perspectives. For the physical layer, a link quality evaluation model based on Uniform Manifold Approximation and Projection (UMAP) and K-Means clustering is proposed from the link quality. For the MAC layer, a multi-criteria decision-making evaluation model based on combined weighting and the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) is proposed, which comprehensively considers both subjective and objective factors, utilizing the TOPSIS algorithm for an accurate evaluation of the MAC layer system status. For the RFID system, this paper proposes a real-time status evaluation model based on the Classification and Regression Tree (CART), which synthesizes the evaluation results of the physical layer and MAC layer. Finally, engineering tests and verification were conducted on the RFID robot system in mobile scenarios. The results showed that the clustering average silhouette coefficient of the physical layer link quality evaluation model based on K-Means was 0.70184, indicating a relatively good clustering effect. The system status evaluation model of the MAC layer, based on the combined weighting-TOPSIS method, demonstrated good flexibility and generalization. The real-time status evaluation model of the RFID system, based on CART, achieved a classification accuracy of 98.3%, with an algorithm runtime of 0.003 s. Compared with other algorithms, it had a higher classification accuracy and shorter runtime, making it well suited for the real-time evaluation of the RFID robot system’s status in dynamic scenarios.

Dimensionality Reduction and Nearest Neighbors for Improving Out-of-Distribution Detection in Medical Image Segmentation

Clinically deployed deep learning-based segmentation models are known to fail on data outside of their training distributions. While clinicians review the segmentations, these models tend to perform well in most instances, which could exacerbate automation bias. Therefore, detecting out-of-distribution images at inference is critical to warn the clinicians that the model likely failed. This work applied the Mahalanobis distance (MD) post hoc to the bottleneck features of four Swin UNETR and nnU-net models that segmented the liver on T1-weighted magnetic resonance imaging and computed tomography. By reducing the dimensions of the bottleneck features with either principal component analysis or uniform manifold approximation and projection, images the models failed on were detected with high performance and minimal computational load. In addition, this work explored a non-parametric alternative to the MD, a k-th nearest neighbors distance (KNN). KNN drastically improved scalability and performance over MD when both were applied to raw and average-pooled bottleneck features. Our code is available at <a href='https://github.com/mckellwoodland/dimen_reduce_mahal'>https://github.com/mckellwoodland/dimen_reduce_mahal</a>

Feature extraction and classification of microstructure and magnetic tomography images of an advanced Nd-Fe-B sintered magnet

Abstract Although the microstructure and magnetic tomography images of an advanced Nd-Fe-B sintered magnet were previously reported [Takeuchi, SO., NPG Asia Mater. 14, 70 (2022)], the relationship between these three-dimensional images has not been well analyzed. In this work, a feature extraction method of histogram of oriented gradients and a classification method of uniform manifold approximation and projection are employed for this issue. The microstructural features with superimposing the information of magnetic domain structure are classified into two groups depending on the external magnetic field, resulting in the different microstructural features for the different magnetization states are successfully classified. These differences of the microstructural features are difficult to be found by human recognition. Further detailed analysis of these microstructural features may clarify the key microstructures for the nucleation of reversed domains and their propagations.

Wide Area VISTA Extra-galactic Survey (WAVES): Unsupervised star-galaxy separation on the WAVES-Wide photometric input catalogue using UMAP and hdbscan

Abstract Star-galaxy separation is a crucial step in creating target catalogues for extragalactic spectroscopic surveys. A classifier biased towards inclusivity risks including high numbers of stars, wasting fibre hours, while a more conservative classifier might overlook galaxies, compromising completeness and hence survey objectives. To avoid bias introduced by a training set in supervised methods, we employ an unsupervised machine learning approach. Using photometry from the Wide Area VISTA Extragalactic Survey (WAVES)-Wide catalogue comprising 9-band u - Ks data, we create a feature space with colours, fluxes, and apparent size information extracted by ProFound. We apply the non-linear dimensionality reduction method UMAP (Uniform Manifold Approximation and Projection) combined with the classifier hdbscan to classify stars and galaxies. Our method is verified against a baseline colour and morphological method using a truth catalogue from Gaia, SDSS, GAMA, and DESI. We correctly identify 99.75% of galaxies within the AB magnitude limit of Z = 21.2, with an F1 score of 0.9971 ± 0.0018 across the entire ground truth sample, compared to 0.9879 ± 0.0088 from the baseline method. Our method’s higher purity (0.9967 ± 0.0021) compared to the baseline (0.9795 ± 0.0172) increases efficiency, identifying 11% fewer galaxy or ambiguous sources, saving approximately 70,000 fibre hours on the 4MOST instrument. We achieve reliable classification statistics for challenging sources including quasars, compact galaxies, and low surface brightness galaxies, retrieving 92.7%, 84.6%, and 99.5% of them respectively. Angular clustering analysis validates our classifications, showing consistency with expected galaxy clustering, regardless of the baseline classification.

Subclonal Populations Detected by Flow Cytometry in Chronic Lymphocytic Leukemia

Abstract Introduction Chronic lymphocytic leukemia/small lymphocytic lymphoma (CLL/SLL) is a mature, small B-cell lymphoma that typically co-expresses CD5 and CD23. Molecular-targeted therapies have produced remarkable therapeutic benefits in CLL, yet many patients are resistant to treatment. This study aims to identify subpopulations in CLL by conventional flow cytometry and clustering analysis. Methods One-hundred-and-forty-one CLL cases were taken from a public database and analyzed by uniform manifold approximation and projection (UMAP). Differences in size, granularity, and intensities of CD19, CD20, CD79a, CD5, CD23, CD10, and FMC7 were plotted with BD FACSDiva Software. Eighteen cases with a new diagnosis of CLL at our institution between January 1, 2021 to December 31, 2022 were analyzed. Only peripheral blood samples were included. Markers including forward scatter, side scatter, and fluorescent intensities of CD19, CD20, CD5, CD23, CD10, surface kappa, and surface lambda were used to identify discrete CLL clusters in UMAP plots. Results: Public Dataset Clustering analysis of conventional flow cytometry data from 141 cases from a public database reveals discrete populations of cells exhibiting varying sizes, granularity and fluorescent intensities. Among the 141 cases, 33 (23.4%) had 1 cluster, 64 (45.4%) with 2 clusters, 42 (29.8%) with 3 clusters and 2 (1.4%) with 4 clusters. Among the cases with 2 clusters, a smaller population of cells deviated from the main neoplastic population. This smaller population had a higher forward and side scatter, and brighter intensities in CD45 and CD19, consistent with prolymphocytes. Among the cases with 3 clusters, a third population recurrently had a smaller size by forward scatter. A few cases additionally had varying intensities by CD23 and CD19. Among the 2 cases with 4 clusters, only slight deviations by UMAP were seen and may be artifactual. Institutional Dataset Clustering analysis of 18 cases at our institution revealed complex patterns of cell clusters exhibiting varying sizes, granularity and surface receptor expression intensity. Among the 18 cases, 1 (5.6%) had 1 cluster, 1 (5.6%) with 2 clusters, 10 (55.6%) with 3 clusters, 3 (16.7%) with 4 clusters, and 3 (16.7%) with 5 clusters. Similar to the public dataset, the third population, in addition to the predominant population and prolymphocytes, consistently varied by lower forward scatter. Additional clusters, such as a population with varying CD23 signal intensities, were identified in some cases. Conclusions In our study, multiple clonal subsets in CLL/SLL are detected by conventional flow cytometry analysis. These subclonal populations may result in therapy resistance for these patients, and identifying them by flow cytometry analysis may provide additional prognostic information in the clinical setting. Furthermore, the ability to detect these subclonal populations may allow for isolation of specific clones for future investigation in research settings.

Advanced classification of hot subdwarf binaries using artificial intelligence techniques and Gaia DR3 data

Hot subdwarf stars are compact blue evolved objects, burning helium in their cores surrounded by a tiny hydrogen envelope. In the Hertzsprung-Russell Diagram they are located by the blue end of the Horizontal Branch. Most models agree on a quite probable common envelope binary evolution scenario in the Red Giant phase. However, the current binarity rate for these objects is yet unsolved, but key, question in this field. This study aims to develop a novel classification method for identifying hot subdwarf binaries within large datasets using Artificial Intelligence techniques and data from the third Gaia data release (GDR3). The results will be compared with those obtained previously using Virtual Observatory techniques on coincident samples. The methods used for hot subdwarf binary classification include supervised and unsupervised machine learning techniques. Specifically, we have used Support Vector Machines (SVM) to classify 3084 hot subdwarf stars based on their colour-magnitude properties. Among these, 2815 objects have Gaia DR3 BP/RP spectra, which were classified using Self-Organizing Maps (SOM) and Convolutional Neural Networks (CNN). In order to ensure spectral quality, previously to SOM and CNN classification, our 2815 BP/RP set were pre-analysed with two different approaches: the cosine similarity technique and the Uniform Manifold Approximation and Projection (UMAP) technique. Additional analysis onto a golden sample of 88 well-defined objects, is also presented. The findings demonstrate a high agreement level (sim 70-90<!PCT!>) with the classifications from the Virtual Observatory Sed Analyzer (VOSA) tool. This shows that the SVM, SOM, and CNN methods effectively classify sources with an accuracy comparable to human inspection or non-AI techniques. Notably, SVM in a radial basis function achieves 70.97<!PCT!> reproducibility for binary targets using photometry, and CNN reaches 84.94<!PCT!> for binary detection using spectroscopy. We also found that the single--binary differences are especially observable on the infrared flux in our Gaia DR3 BP/BR spectra, at wavelengths larger than $ sim $700 nm. We find that all the methods used are in fairly good agreement and are particularly effective to discern between single and binary systems. The agreement is also consistent with the results previously obtained with VOSA. In global terms, considering all quality metrics, CNN is the method that provides the best accuracy. The methods also appear effective for detecting peculiarities in the spectra. While promising, challenges in dealing with uncertain compositions highlight the need for caution, suggesting further research is needed to refine techniques and enhance automated classification reliability, particularly for large-scale surveys.

Single-cell analysis reveals alternations between the aged and young mice prostates

BackgroundAging of the male prostate is an inevitable process in which the prostate undergoes hyperplasia, and this growth may lead to compression of the urethra, resulting in voiding dysfunction and associated symptoms, and an increased risk of prostate cancer. Despite the significance of prostate aging, the molecular mechanisms involved are still not fully understood.MethodsProstate split by lobes from young (2 months) and aged (24 months) mice were collected for single-cell RNA sequencing (scRNA-seq) analysis. Tissues from both anterior prostate (AP) and ventral/dorsal/lateral prostate (VDLP) were included in the study. Data analysis included unsupervised clustering using the uniform manifold approximation and projection (UMAP) algorithm to identify distinct cell types based on marker gene expression. Differential gene expression analysis was performed to identify age-related changes in gene expression across different cell types. Functional enrichment analysis was conducted to elucidate biological pathways associated with differentially expressed genes. Additionally, cellular interactions and developmental trajectories were analyzed to characterize cellular dynamics during prostate aging.ResultsThe single-cell transcriptome analysis of the mouse prostate during aging revealed heterogeneity across various cell types and their changes during the aging process. We found a significant increase in the proportion of mesenchymal and immune cells in aged mice. Our study unveiled alterations in genes and pathways associated with cellular senescence, oxidative stress, and regeneration in epithelial cells. Furthermore, we observed that basal cells may undergo epithelial-mesenchymal transition (EMT) to become mesenchymal cells, particularly prominent in aged mice. Additionally, immune cells, notably macrophages and T cells, exhibited a heightened inflammatory response in aged mice.ConclusionIn summary, our study provides a comparative analysis of the single-cell transcriptome of the aged and young mice prostates, elucidating cellular and molecular changes between the aged and young mice prostates.

Heart Disease Prediction Using CNN with Various Feature Selection Approaches

Objectives: To evaluate the performance of CNN models with feature selection methods like Relief, Uniform Manifold Approximation and Projection (UMAP), and Linear discriminant Analysis (LDA) for forecasting heart diseases. Methods: The present research for heart disease prediction compares the performance of feature selection algorithms like ReliefF, Uniform Manifold Approximation and Projection (UMAP), and Linear discriminant Analysis (LDA) with Convolution Neural networks (CNN) for prediction. The study is conducted in a dataset with 303 records collected from patients with 14 attributes. It is also validated with the publicly available Cleveland dataset (Kaggle). The software environment used for implementation is Jupyter Notebook, which uses Python. The dataset consists of 303 records collected from the patients with 14 attributes. It is validated with the publicly available Cleveland dataset. Findings: The study examines how these feature selection techniques affect CNN's accuracy. According to experimental findings, the CNN-UMAP hybrid model outperforms with an accuracy of 91.88%, precision of 0.89, and recall of 0.85 compared to ReliefF and LDA, UMAP shows up as the most successful feature selection method among the studied techniques when utilized alongside CNNs. Novelty: ReliefF, UMAP, and LDA allow CNN to learn more significant and discriminative features by minimizing the dimensions of the input data while preserving its underlying pattern. Previous studies also attempted to identify the key contributing characteristics to heart disease prediction, but less emphasis was placed on these feature selection methods in determining the effectiveness of the features for heart disease prediction. Keywords: Heart Disease, Feature Selection Methods, Convolutional Neural Network, ReliefF, UMAP and LDA

High-dimensional Immune Profiles and Machine Learning May Predict Acute Myeloid Leukemia Relapse Early following Transplant.

Identification of early immune signatures associated with acute myeloid leukemia (AML) relapse following hematopoietic stem cell transplant (HSCT) is critical for patient outcomes. We analyzed PBMCs from 58 patients with AML undergoing HSCT, focusing on T cell subsets and functional profiles. High-dimensional flow cytometry coupled with Uniform Manifold Approximation and Projection dimensionality reduction and PhenoGraph clustering revealed distinct changes in CD4+ and CD8+ T cell populations in 16 patients who relapsed within 1 y of HSCT. We observed increased IL-2, IL-10, and IL-17-producing CD4+ T cells, alongside decreased CD8+ T cell function early in relapsing patients. Notably, relapsing patients exhibited increased TCF-1intermediate cells, which lacked granzyme B or IFN-γ production in the CD4+ T cell compartment. We then developed a supervised machine learning algorithm that predicted AML relapse with 90% accuracy within 30 d after HSCT using high-throughput assays. The algorithm leverages condensed immune phenotypic data, alongside the ADASYN algorithm, for data balancing and 100 rounds of XGBoost supervised learning. This approach holds potential for detecting relapse-associated immune signatures months before clinical manifestation. Our findings demonstrate a distinct immunological signature potentially capable of predicting AML relapse as early as 30 d after HSCT.

The Wong-Zakai approximations of invariant manifolds for retarded partial differential equations with multiplicative white noise

We focus on the dynamics and Wong-Zakai approximation for a class of retarded partial differential equations subjected to multiplicative white noise. We show that when restricted to a local region and under certain conditions, there exists a unique global solution for the truncated system driven by either the white noise or the approximation noise. Such solution generates a random dynamical system, and the solutions of Wong-Zakai approximations are convergent to solutions of the stochastic retarded differential equation. We also show that there exist invariant manifolds for the truncated system driven by either the white noise or the approximation noise, which are then the local manifolds for the untruncated systems, and prove that such invariant manifolds of the Wong-Zakai approximations converge to those of the stochastic retarded differential equation.

Odontogenic Myxomas Harbor Recurrent Copy Number Alterations and a Distinct Methylation Signature.

Odontogenic myxoma is a rare, benign, and locally aggressive tumor that develops in the tooth-bearing areas of the jaw. The molecular mechanisms underlying odontogenic myxomas are unknown and no diagnostic markers are available to date. The aim of this study was to analyze DNA methylation and copy number variations in odontogenic myxomas to identify new molecular signatures for diagnostic decision-making. We collected a cohort of 16 odontogenic myxomas from 2006 to 2021 located in the mandible (n = 10) and maxilla (n = 6) with available formalin-fixed paraffin-embedded or fresh frozen tumor tissue from a biopsy or resection material. Genome-wide DNA methylation and copy number variation data were generated from 12 odontogenic myxomas using the Illumina Infinium Methylation EPIC array, interrogating >850,000 CpG sites. Unsupervised clustering and dimensionality reduction (Uniform Manifold Approximation and Projection) revealed that odontogenic myxomas formed a distinct DNA methylation class. Copy number profiling showed recurrent whole-chromosome gains (trisomies) of chromosomes 5, 8, and 20 in all cases, and of chromosomes 10, 12, and 17 in all except one case. In conclusion, odontogenic myxomas harbor recurrent copy number patterns and a distinct DNA methylation profile, which can be used as an additional diagnostic tool in the appropriate clinical and radiologic context. Further research is needed to explain the genetic mechanisms caused by these alterations that drive these locally aggressive neoplasms.

Sifting the debris: Patterns in the SNR population with unsupervised ML methods

Context. Supernova remnants (SNRs) carry vast amounts of mechanical and radiative energy that heavily influence the structural, dynamical, and chemical evolution of galaxies. To this day, more than 300 SNRs have been discovered in the Milky Way, exhibiting a wide variety of observational features. However, existing classification schemes are mainly based on their radio morphology. Aims. In this work, we introduce a novel unsupervised deep learning pipeline to analyse a representative subsample of the Galactic SNR population (~50% of the total) with the aim of finding a connection between their multi-wavelength features and their physical properties. Methods. The pipeline involves two stages: (1) a representation learning stage, consisting of a convolutional autoencoder that feeds on imagery from infrared and radio continuum surveys (WISE 22 μm, Hi-GAL 70 μm and SMGPS 30 cm) and produces a compact representation in a lower-dimensionality latent space; and (2) a clustering stage that seeks meaningful clusters in the latent space that can be linked to the physical properties of the SNRs and their surroundings. Results. Our results suggest that this approach, when combined with an intermediate uniform manifold approximation and projection (UMAP) reprojection of the autoencoded embeddings into a more clusterable manifold, enables us to find reliable clusters. Despite a large number of sources being classified as outliers, most clusters relate to the presence of distinctive features, such as the distribution of infrared emission, the presence of radio shells and pulsar wind nebulae, and the existence of dust filaments.

Deep learning disconnectomes to accelerate and improve long-term predictions for post-stroke symptoms.

This study investigates the efficacy of deep-learning models in expediting the generation of disconnectomes for individualized prediction of neuropsychological outcomes one year after stroke. Utilising a 3D U-Net network, we trained a model on a dataset of N = 1333 synthetic lesions and corresponding disconnectomes, subsequently applying it to N = 1333 real stroke lesions. The model-generated disconnection patterns were then projected into a two-dimensional 'morphospace' via uniform manifold approximation and projection for dimension reduction dimensionality reduction. We correlated the positioning within this morphospace with one-year neuropsychological scores across 86 metrics in a novel cohort of 119 stroke patients, employing multiple regression models and validating the findings in an out-of-sample group of 20 patients. Our results demonstrate that the 3D U-Net model captures the critical information of conventional disconnectomes with a notable increase in efficiency, generating deep-disconnectomes 720 times faster than current state-of-the-art software. The predictive accuracy of neuropsychological outcomes by deep-disconnectomes averaged 85.2% (R 2 = 0.208), which significantly surpassed the conventional disconnectome approach (P = 0.009). These findings mark a substantial advancement in the production of disconnectome maps via deep learning, suggesting that this method could greatly enhance the prognostic assessment and clinical management of stroke survivors by incorporating disconnection patterns as a predictive tool.