Unsupervised K-means Clustering Research Articles

The Immune Landscape and Molecular Subtypes of Pediatric Crohn's Disease: Results from In Silico Analysis.

Pediatric Crohn's disease (CD) presents a distinct phenotype from adult-onset disease. A dysregulated immune response is critical in CD pathogenesis; thus, it is clinically important to describe immune cell alterations and to identify a new molecular classification for pediatric CD. To this end, in this study, a RNA-seq derived dataset GSE101794-which contains the expression profiles of 254 treatment-naïve pediatric CD samples, including CIBERSORTx and weighted gene-co-expression network analysis (WGCNA)-were performed to estimate the ratio of immune cells and to identify modules and genes related to specific immune cell infiltration, respectively. Hub genes derived from WGCNA were further employed to create a molecular classification using unsupervised K-means clustering. In the pediatric CD samples, it was found that M2 macrophages, CD4+ memory resting T cells, CD8+ T cells, and resting mast cells were the most prominent immune cells in intestinal tissues. Then, 985 up-regulated genes and 860 down-regulated genes were identified in samples with high immune cell infiltration. Of these differential genes, 10 hub genes (APOA1, CYB5A, XPNPEP2, SLC1A7, SLC4A6, LIPE, G6PC, AGXT2, SLC13A1, and SOAT2) were associated with CD8+T cell infiltration. Clinically, the higher expression of these 10 hub genes was strongly associated with an earlier age of CD onset and colonic-type CD. Furthermore, based on these key genes, pediatric CD could be classified into three molecular subtypes, displaying a different immune landscape. Altogether, this in silico analysis provides a novel insight into the immune signature of pediatric CD, and a new classification of pediatric CD is presented, which may help us develop more personalized disease management and treatments for pediatric CD.

Tool wear identification and prediction method based on stack sparse self-coding network

In the process of metal cutting, the effective monitoring of tool wear is of great significance to ensure the machining quality of parts. Aiming at the problem of tool wear monitoring, a tool wear recognition and prediction method based on stack sparse self-coding network is proposed. This method can simplify the establishment process of monitoring model, monitor the tool wear according to different task requirements, and guide the tool replacement in the actual cutting process. Firstly, unsupervised K-means clustering is used to divide the tool wear stage, and the feature set is marked. Secondly, the parameters of stack sparse self-coding network layer are determined by trial, and the sensitive features that can reflect the tool wear process are obtained. Finally, the tool wear identification model of stack sparse self-encoder and the tool wear prediction model of BP neural network are established respectively, and the smoothing correction method is used to further improve the prediction accuracy. The experimental results show that the established tool wear identification and prediction model can accurately monitor the tool wear state and wear amount, and has a certain reference value for efficient tool change in the actual metal cutting process.

Abstract P3-05-57: Lymphatico-venous communications and rhythmic, respiratory pressure changes influence the incidence of metastatic disease at presentation in solid tumours

Abstract Background- Dissemination of cancer cells and their colonization at distant sites (metastases) is the primary cause of death in patients with solid tumors. Despite of popular theories of cancer metastases, there are several lacunae in our understanding of extent of Metastatic Disease at Presentation (MDP). The aim of this study was to evaluate if organ sites could be grouped together based on the available knowledge extent of lympho-vascular communication (LVC) and probable respiratory pressure changes they undergo and whether this correlated with incidence of MDP. Methods- We analyzed a prospectively maintained database of patients with solid tumors presenting between January and December 2017 at our tertiary care cancer center. The tumors were hypothetically divided into four regions (R) depending on extent of LVC and respiratory pressure changes. All patients had clinico-radiological staging after pathological confirmation of cancer. The hypothetical region grouping R1-R4 was independently evaluated by unsupervised k-means clustering. The incidence of MDP was compared in subgroups defined by primary tumor stage, anatomical region, histology, gender, and age using logistic regression. Results- Of 17,467 patients included with median age 54 years, 6714 (38.4%) had MDP. The incidence of M1 disease in R1 vs R2 vs R3 vs R4 regions, older (>50 years) vs younger Conclusion- Primary anatomical regions defined by LVC and respiratory pressure changes; primary tumor histology (type and grade) and male gender are significant determinants of MDP in solid tumors. Region grouping of solid organs based on extent of lymphaticovenous communications and rhythmic respiratory pressure changes and % MDP Citation Format: Rajendra Badwe, Shalaka Joshi, Rohini Hawaldar, Rohan Chaubal, Atanu Bhattacharjee, Vaibhav Vanmali, Nita S. Nair, Sudeep Gupta. Lymphatico-venous communications and rhythmic, respiratory pressure changes influence the incidence of metastatic disease at presentation in solid tumours [abstract]. In: Proceedings of the 2022 San Antonio Breast Cancer Symposium; 2022 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2023;83(5 Suppl):Abstract nr P3-05-57.

Lymph node metastasis‐dependent molecular classification in papillary thyroid carcinoma defines aggressive metastatic outgrowth

Lymph node metastasis‐dependent molecular classification in papillary thyroid carcinoma defines aggressive metastatic outgrowth

Prognosis-related molecular subtyping in head and neck squamous cell carcinoma patients based on glycolytic/cholesterogenic gene data

BackgroundHead and neck squamous cell carcinoma (HNSCC) remains an unmet medical challenge. Metabolic reprogramming is a hallmark of diverse cancers, including HNSCC.MethodsWe investigated the metabolic profile in HNSCC by using The Cancer Genome Atlas (TCGA) (n = 481) and Gene Expression Omnibus (GEO) (n = 97) databases. The metabolic stratification of HNSCC samples was identified by using unsupervised k-means clustering. We analyzed the correlations of the metabolic subtypes in HNSCC with featured genomic alterations and known HNSCC subtypes. We further validated the metabolism-related subtypes based on features of ENO1, PFKFB3, NSDHL and SQLE expression in HNSCC by Immunohistochemistry. In addition, genomic characteristics of tumor metabolism that varied among different cancer types were confirmed.ResultsBased on the median expression of coexpressed cholesterogenic and glycolytic genes, HNSCC subtypes were identified, including glycolytic, cholesterogenic, quiescent and mixed subtypes. The quiescent subtype was associated with the longest survival and was distributed in stage I and G1 HNSCC. Mutation analysis of HNSCC genes indicated that TP53 has the highest mutation frequency. The CDKN2A mutation frequency has the most significant differences amongst these four subtypes. There is good overlap between our metabolic subtypes and the HNSCC subtype.ConclusionThe four metabolic subtypes were successfully determined in HNSCC. Compared to the quiescent subtype, glycolytic, cholesterogenic and mixed subtypes had significantly worse outcome, which might offer guidelines for developing a novel treatment strategy for HNSCC.

Classifying aneuploidy in genotype intensity data using deep learning.

Aneuploidy is the loss or gain of one or more chromosomes. Although it is a rare phenomenon in liveborn individuals, it is observed in livestock breeding populations. These breeding populations are often routinely genotyped and the genotype intensity data from single nucleotide polymorphism (SNP) arrays can be exploited to identify aneuploidy cases. This identification is a time-consuming and costly task, because it is often performed by visual inspection of the data per chromosome, usually done in plots of the intensity data by an expert. Therefore, we wanted to explore the feasibility of automated image classification to replace (part of) the visual detection procedure for any diploid species. The aim of this study was to develop a deep learning Convolutional Neural Network (CNN) classification model based on chromosome level plots of SNP array intensity data that can classify the images into disomic, monosomic and trisomic cases. A multispecies dataset enriched for aneuploidy cases was collected containing genotype intensity data of 3321 disomic, 1759 monosomic and 164 trisomic chromosomes. The final CNN model had an accuracy of 99.9%, overall precision was 1, recall was 0.98 and the F1 score was 0.99 for classifying images from intensity data. The high precision assures that cases detected are most likely true cases, however, some trisomy cases may be missed (the recall of the class trisomic was 0.94). This supervised CNN model performed much better than an unsupervised k-means clustering, which reached an accuracy of 0.73 and had especially difficult to classify trisomic cases correctly. The developed CNN classification model provides high accuracy to classify aneuploidy cases based on images of plotted X and Y genotype intensity values. The classification model can be used as a tool for routine screening in large diploid populations that are genotyped to get a better understanding of the incidence and inheritance, and in addition, avoid anomalies in breeding candidates.

Mortality prediction on unsupervised and semi-supervised clusters of medical intensive care unit patients based on MIMIC-II database

IntroductionThe study aimed to propose a framework for identifying patient clusters in medical intensive care units (MICUs) based on the Medical Information Mart for Intensive Care II (MIMIC-II) database. The suggested framework makes use of the survival outcomes and physiological information available in the dataset and is hence called a semi-supervised approach. Five neural networks were trained on the clusters identified using the proposed approach to determine whether the proposed framework could improve the predictive accuracy of the deep learning models. MethodsThis study utilized data from the MIMIC-II database, which is a publicly available database that contains information on patients admitted to intensive care units. The clusters underlying the MICU patient population were identified using unsupervised and semi-supervised K-means clustering. Mortality in the resulting clusters was predicted using five deep learning-based survival models and the performance of these models was compared using two metrics. ResultsThree clusters (cluster 1, n = 1304; cluster 2, n = 474; cluster 3, n = 1079) were identified using unsupervised K-means, and another three clusters (cluster 1, n = 479; cluster 2, n = 1492; cluster 3, n = 886) were identified using semi-supervised K-means clustering. Experimental results demonstrate that, in general, the performance of deep learning models was better on semi-supervised clusters obtained by combining Cox proportional hazards (Cox-PH) model-based feature selection and K-means compared to unsupervised clusters. ConclusionsIn the present study, it was observed that deep learning-based survival models tend to perform better on clusters that are identified in a semi-supervised fashion. This approach helps to extract more meaningful patterns and associations between different clinical features and patient outcomes.

QUANTITATIVE ANALYSIS OF CARBON BLACK AGGLOMERATES MORPHOLOGY IN ELASTOMER COMPOSITES BASED ON X-RAY TOMOGRAPHY BY MEANS OF NUMERICAL CLUSTERING

ABSTRACT A novel methodology for characterizing the morphology distribution of filler agglomerates in elastomer composites is presented based on laboratory-sourced X-ray tomography. Various feature extraction methods (via, e.g., image-processing filters, segmentation) and selection tools (Spearman's rank correlation coefficient) combined with K-means unsupervised clustering algorithm were developed for identifying the distinct morphological classes in model materials (carbon-filled ethylene propylene diene monomer rubber). The interest of this methodology was demonstrated by precisely differentiating the materials compounded with different processing parameters. For instance, in this example, thanks to this analysis, it was found that introducing the filler before the elastomer in internal mixer tends to favor more structured agglomerates.

Diagnostic milk biomarkers for predicting the metabolic health status of dairy cattle during early lactation.

Data on metabolic profiles of blood sampled at d 3, 6, 9, and 21 in lactation from 117 lactations (99 cows) were used for unsupervised k-means clustering. Blood metabolic parameters included β-hydroxybutyrate (BHB), nonesterified fatty acids, glucose, insulin-like growth factor-1 (IGF-1) and insulin. Clustering relied on the average and range of the 5 blood parameters of all 4 sampling days. The clusters were labeled as imbalanced (n = 42) and balanced (n = 72) metabolic status based on the values of the blood parameters. Various random forest models were built to predict the metabolic cluster of cows during early lactation from the milk composition. All the models were evaluated using a leave-group-out cross-validation, meaning data from a single cow were always present in either train or test data to avoid any data leakage. Features were either milk fatty acids (MFA) determined by gas chromatography (MFA [GC]) or features that could be determined during a routine dairy herd improvement (DHI) analysis, such as concentration of fat, protein, lactose, fat/protein ratio, urea, and somatic cell count (determined and reported routinely in DHI registrations), either or not in combination with MFA and BHB determined by mid-infrared (MIR), denoted as MFA [MIR] and BHB [MIR], respectively, which are routinely analyzed but not routinely reported in DHI registrations yet. Models solely based on fat, protein, lactose, fat/protein ratio, urea and somatic cell count (i.e., DHI model) were characterized by the lowest predictive performance [area under the receiver operating characteristic curve (AUCROC) = 0.69]. The combination of the features of the DHI model with BHB [MIR] and MFA [MIR] powerfully increased the predictive performance (AUCROC = 0.81). The model based on the detailed MFA profile determined by GC analysis did not outperform (AUCROC = 0.81) the model using the DHI-features in combination with BHB [MIR] and MFA [MIR]. Predictions solely based on samples at d 3 were characterized by lower performance (AUCROC DHI + BHB [MIR] + MFA [MIR] model at d 3: 0.75; AUCROC MFA [GC] model at d 3: 0.73). High predictive performance was found using samples from d 9 and 21. To conclude, overall, the DHI + BHB [MIR] + MFA [MIR] model allowed to predict metabolic status during early lactation. Accordingly, these parameters show potential for routine prediction of metabolic status.

Ore-Waste Discrimination Using Supervised and Unsupervised Classification of Hyperspectral Images

Ore and waste discrimination is essential for optimizing exploitation and minimizing ore dilution in a mining operation. The conventional ore/waste discrimination approach relies on the interpretation of ore control by geologists, which is subjective, time-consuming, and can cause safety hazards. Hyperspectral remote sensing can be used as an alternative approach for ore/waste discrimination. The focus of this study is to investigate the application of hyperspectral remote sensing and deep learning (DL) for real-time ore and waste classification. Hyperspectral images of several meters of drill core samples from a silver ore deposit labeled by a site geologist as ore and waste material were used to train and test the models. A DL model was trained on the labels generated by a spectral angle mapper (SAM) machine learning technique. The performance on ore/waste discrimination of three classifiers (supervised DL and SAM, and unsupervised k-means clustering) was evaluated using Rand Error and Pixel Error as disagreement analysis and accuracy assessment indices. The results showed that the DL method outperformed the other two techniques. The performance of the DL model reached 0.89, 0.95, 0.89, and 0.91, respectively, on overall accuracy, precision, recall, and F1 score, which indicate the strong capability of the DL model in ore and waste discrimination. An integrated hyperspectral imaging and DL technique has strong potential to be used for practical and efficient discrimination of ore and waste in a near real-time manner.

Machine Learning-Assisted Gas-Specific Fingerprint Detection/Classification Strategy Based on Mutually Interactive Features of Semiconductor Gas Sensor Arrays

A high-performance machine learning-assisted gas sensor strategy based on the integration of supervised and unsupervised learning with a gas-sensitive semiconductor metal oxide (SMO) gas sensor array is introduced. A 4-SMO sensor array was chosen as a test sensor system for detecting carbon monoxide (CO) and ethyl alcohol (C2H5OH) mixtures using 15 different combinations. Gas sensing detection/classification was performed with different numbers of gas sensor and machine learning algorithms. K-Means clustering was successfully employed to rationally identify the similarity features of targeted gases among 4 different groups, i.e., matrix gas, two single-component gases, and one two-gas mixture, based on only unlabeled voltage-based gas sensing information. Detailed classification was performed through a multitude of supervised algorithms, i.e., 2-layer artificial neural networks (ANNs), 4-layer deep neural networks (DNNs), 1-dimensional convolutional neural networks (1D CNNs), and 2-dimensional CNNs (2D CNNs). The numerical-based DNNs and image-based CNNs are shown to be excellent approaches for gas detection and classification, as indicated by the highest accuracy and lowest loss indicators. Through the analysis of the influence of the number of sensors on the arrayed gas sensor system, the application of machine learning methodology to an arrayed gas sensor system demonstrates four unique features, i.e., a data augmentation methodology, machine learning approach of combining K-means clustering and neural networks, and a systematic approach to optimized sensor combinations, potentially leading to the practical sensor networks based on chemical sensors. Even two SMO sensor combinations are shown to be highly effective in gas discrimination against diverse gas environments assisted through numeric-based DNNs and image-based 1D CNNs, overcoming the simple clustering proposed through the unsupervised K-means clustering.

CNSC-15. DECIPHERING RAMAN-SPECTRAL TISSUE HETEROGENEITY IN GLIOBLASTOMA

Abstract Raman Spectroscopy is able to provide a fast identification method for healthy and pathological tissues without the need for sample preparation. This makes it highly interesting for intraoperative use in identifying tumor types and borders. However, the capability of a machine-learning classifier to recognize known tissues based on Raman spectra relies on a solid ground truth that is usually provided by measuring small tissue samples first and analyzing the sample later by histopathology. This approach is limited in very closely interspersed tissues that cannot be identified macroscopically, such as glioblastoma where vital tumor tissue, necrosis and peritumoral tissue with characteristic changes to healthy brain tissue are tightly arranged. These areas have been shown to be distinct in their spectral properties, which makes tumor border identification aided by Raman Spectroscopy challenging. As infiltrative tumors leaving no tumor-cell-free brain tissue, borders need to be defined on a clinical scale, where extensive resection must be weighed against retaining functional brain tissue. For this, spectral analysis beyond tumor identification is necessary. In a computational approach we use methods of unsupervised learning and K-means Clustering to investigate heterogeneity within glioblastomas in the native state in order to reduce the complexity of this tumor type. Thereby, it was possible to extract spectra of necrotic tumor areas and to calculate a prediction probability for necrotic tissue within glioblastomas. Several clusters with spectral similarity to each other could be identified, which might represent the different areas of a glioblastoma. Two of these clusters were identified as probable grey and white matter, respectively, by comparison with autopsy tissue. In addition, this assignment independently confirmed the prediction of the classifier as non-necrotic tissue. Deciphering the spectral heterogeneity of glioblastoma in this way may be a great advantage for the surgeon to identify tumor borders and assist in resection control.

Identification and immune characteristics of molecular subtypes related to protein glycosylation in Alzheimer's disease.

Protein glycosylation has been confirmed to be involved in the pathological mechanisms of Alzheimer's disease (AD); however, there is still a lack of systematic analysis of the immune processes mediated by protein glycosylation-related genes (PGRGs) in AD. Transcriptomic data of AD patients were obtained from the Gene Expression Omnibus database and divided into training and verification datasets. The core PGRGs of the training set were identified by weighted gene co-expression network analysis, and protein glycosylation-related subtypes in AD were identified based on k-means unsupervised clustering. Protein glycosylation scores and neuroinflammatory levels of different subtypes were compared, and functional enrichment analysis and drug prediction were performed based on the differentially expressed genes (DEGs) between the subtypes. A random forest model was used to select important DEGs as diagnostic markers between subtypes, and a line chart model was constructed and verified in other datasets. We evaluated the differences in immune cell infiltration between the subtypes through the single-sample gene set enrichment analysis, analyzed the correlation between core diagnostic markers and immune cells, and explored the expression regulation network of the core diagnostic markers. Eight core PGRGs were differentially expressed between the training set and control samples. AD was divided into two subtypes with significantly different biological processes, such as vesicle-mediated transport in synapses and neuroactive ligand-receptor interactions. The high protein glycosylation subtype had a higher level of neuroinflammation. Riluzole and sulfasalazine were found to have potential clinical value in this subtype. A reliable construction line chart model was constructed based on nine diagnostic markers, and SERPINA3 was identified as the core diagnostic marker. There were significant differences in immune cell infiltration between the two subtypes. SERPINA3 was found to be closely related to immune cells, and the expression of SERPINA3 in AD was found to be regulated by a competing endogenous RNA network that involves eight long non-coding RNAs and seven microRNAs. Protein glycosylation and its corresponding immune process play an important role in the occurrence and development of AD. Understanding the role of PGRGs in AD may provide a new potential therapeutic target for AD.

Mapping Lunar Swirls with Machine Learning: The Application of Unsupervised and Supervised Image Classification Algorithms in Reiner Gamma and Mare Ingenii

Lunar swirls are recognized as broad, bright albedo features in various regions of the Moon. These features are often separated by dark off-swirl lanes or terminate against the dark background, such as lunar maria. Prior mapping of swirls has been done primarily by albedo contrast, which is prone to subjectivity. Closer examination of on-swirl areas shows that they are not uniform, making the boundary between on- and off-swirl difficult to map with certainty. We have applied machine learning techniques to address these issues by identifying the number of swirl units and then mapping them based on actual reflectance, or I/F data. Using LROC NAC paired stereo images that are converted to I/F reflectance at a range of incidence angles, we applied both unsupervised K-means clustering and supervised Maximum Likelihood Classification algorithms to classify and map portions of lunar swirls in Reiner Gamma and Mare Ingenii. Results show that the classification maps are a reasonable match to the representative albedos for the two study regions. A third transitionary swirl unit, termed diffuse-swirl, is present in both the maps and the cumulative distribution plots of the reflectance values. Overall, we find that the use of both algorithms provides independent confirmation of both the number and location of these units and their interrelation. More importantly, the algorithms remove mapping subjectivity by using quantitative information. The data and the statistics generated from the maps also have value in future studies by placing limits for categorizing swirl units in different regions on the Moon.

Identification and immune characteristics of molecular subtypes related to fatty acid metabolism in idiopathic pulmonary fibrosis.

BackgroundAlthough fatty acid metabolism has been confirmed to be involved in the pathological process of idiopathic pulmonary fibrosis (IPF), systematic analyses on the immune process mediated by fatty acid metabolism-related genes (FAMRGs) in IPF remain lacking.MethodsThe gene expression data of 315 patients with IPF were obtained from Gene Expression Omnibus database and were divided into the training and verification sets. The core FAMRGs of the training set were identified through weighted gene co-expression network analysis. Then, the fatty acid metabolism-related subtypes in IPF were identified on the basis of k-means unsupervised clustering. The scores of fatty acid metabolism and the expression of the fibrosis biomarkers in different subtypes were compared, and functional enrichment analysis was carried out on the differentially expressed genes between subtypes. A random forest model was used to select important FAMRGs as diagnostic markers for distinguishing between subtypes, and a line chart model was constructed and verified by using other datasets and rat models with different degrees of pulmonary fibrosis. The difference in immune cell infiltration among subtypes was evaluated with CIBERSORT, and the correlation between core diagnostic markers and immune cells were analyzed.ResultsTwenty-four core FAMRGs were differentially expressed between the training set and normal samples, and IPF was divided into two subtypes. Significant differences were observed between the two subtypes in biological processes, such as linoleic acid metabolism, cilium movement, and natural killer (NK) cell activation. The subtype with high fatty acid metabolism had more severe pulmonary fibrosis than the other subtype. A reliable construction line chart model based on six diagnostic markers was constructed, and ABCA3 and CYP24A1 were identified as core diagnostic markers. Significant differences in immune cell infiltration were found between the two subtypes, and ABCA3 and CYP24A1 were closely related to NK cells.ConclusionFatty acid metabolism and the immune process that it mediates play an important role in the occurrence and development of IPF. The analysis of the role of FAMRGs in IPF may provide a new potential therapeutic target for IPF.

Abstract B020: Multimodal single-cell analyses reveal identification of unique transcriptional subgroups in ewing sarcoma

Abstract Ewing sarcoma (EwS) tumors are driven by pathognomonic fusions between FET proteins and ETS family transcription factors (most frequently EWS::FLI1). EWS::FLI1 drives tumorigenesis through massive transcriptomic and epigenomic rewiring. Despite mutational homogeneity, EwS are highly transcriptionally heterogeneous suggesting that cell context is a key determinant of EWS::FLI1-driven gene signatures. Fusion protein activity has been revealed as a core source of EwS heterogeneity. However, deep investigation into the transcriptomic and epigenomic heterogeneity of EwS cells has yet to be described. Molecular subtyping of EwS has thus far proven to be difficult to achieve with bulk RNA-sequencing approaches. We hypothesized that single-cell characterization of EwS cells would provide the necessary resolution to allow identification and characterization of transcriptionally distinct tumor subgroups. We subjected eight established EwS cell lines, one PDX-derived EwS line, and five non-EwS cell lines to single-cell multiomic (ATAC + RNA) sequencing. A range of 1300-3800 live cells were captured per sample, with an average of 25,000 transcript and accessible reads per cell. Initial analysis of transcriptomes confirmed the EwS-specific, EWS::FLI1-dependent signature gene sets in EwS samples. Integrated analysis of the multiome data from the 9 EwS samples identified inter-tumor heterogeneity and unsupervised k-means clustering revealed three molecular subgroups that were not evident from expression data alone. Subgroup 1 comprised only the A673 cell line, while the remaining 8 EwS samples were split between subgroups 2 and 3. Gene ontology analysis defined enrichment of distinct gene sets across the 3 subgroups. Interestingly, expression of subgroup 2-defining genes was reduced following knockdown of EWSR1::FLI1, while expression of subgroup 3-defining genes increased. Thus, these data show that transcriptionally distinct subgroups of EwS exist and these are defined, in part, by whether EWS::FLI1-dependent gene activation or gene repression dominates transcriptional rewiring. Our ongoing studies suggest that these molecular subgroups are defined by cooperation between the fusion and cell context-dependent transcription factors. Future analyses will determine if these distinct multiomic programs exist in primary patients tumors and test the pathobiologic and clinical significance of these novel molecular subgroups. Citation Format: April A. Apfelbaum, Olivia Waltner, Shruti S. Bhise, Sami Kanaan, Jay F. Sarthy, Scott N. Furlan, Elizabeth R. Lawlor. Multimodal single-cell analyses reveal identification of unique transcriptional subgroups in ewing sarcoma [abstract]. In: Proceedings of the AACR Special Conference: Sarcomas; 2022 May 9-12; Montreal, QC, Canada. Philadelphia (PA): AACR; Clin Cancer Res 2022;28(18_Suppl):Abstract nr B020.

Prediction of dyslexia severity levels from fixation and saccadic eye movement using machine learning

BackgroundDyslexia is a neurological disorder that causes poor reading and comprehension skills. Dyslexics experience problems in understanding phonemes of languages hence they show less ability to relate letters to form words and sentences. Dyslexia is not directly related to vision but it has been observed that most dyslexics have a magnocellular deficit which causes abnormal eye movements. From the literature, it has been observed that atypical eye movement could be a strong indicator of dyslexia. MethodsThis research work identifies the severity levels of dyslexia from fixations and saccade eye movement events based on eye gaze points. Fixations and saccades are computed using a velocity threshold algorithm. Features related to fixations and saccades are used to divide dyslexics into high and low based on their severity levels. Unsupervised learning approach K-Means clustering is adopted to group the dyslexics into high and low groups. Cluster evaluation measures such as Silhouette analysis, Elbow Method, and Davies Bouldin Index are used to find the optimal value of K. The quality of the clusters is further reinforced by Multivariate analysis of variance (MANOVA) analysis. Data in the clusters are visualized to understand the data distribution and range of different features. ResultsK-means clustering gave two clusters – high and low dyslexics. The Silhouette score was high and the Davies Bouldin Index was low showing that the clusters are of good quality. MANOVA analysis has given a p-value less than 0.01 which further reinforces the quality of clusters to be good. The data in clusters are visualized, analyzed, and compared to existing eye movement findings on dyslexia in literature. ConclusionIt has been observed that features such as fixation duration, number of fixations, saccade duration, and number of saccades act as key indicators to identify the severity levels of dyslexia. The characteristics of data in the clusters generated using these features match with the characteristics of existing eye movement findings, hence the data value ranges can be used to identify the severity levels of dyslexia.

The Impacts of Air Quality on Vegetation Health in Dense Urban Environments: A Ground-Based Hyperspectral Imaging Approach

We examine the impact of changes in ozone (O3), particulate matter (PM2.5), temperature, and humidity on the health of vegetation in dense urban environments, using a very high-resolution, ground-based Visible and Near-Infrared (VNIR, 0.4–1.0 μm with a spectral resolution of 0.75 nm) hyperspectral camera deployed by the Urban Observatory (UO) in New York City. Images were captured at 15 min intervals from 08h00 to 18h00 for 30 days between 3 May and 6 June 2016 with each image containing a mix of dense built structures, sky, and vegetation. Vegetation pixels were identified using unsupervised k-means clustering of the pixel spectra and the time dependence of the reflection spectrum of a patch of vegetation at roughly 1 km from the sensor that was measured across the study period. To avoid illumination and atmospheric variability, we introduce a method that measures the ratio of vegetation pixel spectra to the spectrum of a nearby building surface at each time step relative to that ratio at a fixed time. This “Compound Ratio” exploits the (assumed) static nature of the building reflectance to isolate the variability of vegetation reflectance. Two approaches are used to quantify the health of vegetation at each time step: (a) a solar-induced fluorescence indicator (SIFi) calculated as the simple ratio of the amplitude of the Compound Ratio at 0.75 μm and 0.9 μm, and (b) Principal Component Analysis (PCA) decomposition designed to capture more global spectral features. The time dependence of these vegetation health indicators is compared to that of O3, PM2.5, temperature, and humidity values from a distributed and publicly available in situ air quality sensor network. Assuming a linear relationship between vegetation health indicators and air quality indicators, we find that changes in both SIF indicator values and PC amplitudes show a strong correlation (r2 value of 40% and 47%, respectively) with changes in air quality, especially in comparison with nearby buildings used as controls (r2 value of 1% and 4%, respectively, and with all molecular correlations consistent with zero to within 3σ uncertainty). Using the SIF indicator, O3 and temperature exhibit a positive correlation with changes in photosynthetic rate in vegetation, while PM2.5 and humidity exhibit a negative correlation. We estimate full covariant uncertainties on the coefficients using a Markov Chain Monte Carlo (MCMC) approach and demonstrate that these correlations remain statistically significant even when controlling for the effects of diurnal sun-sensor geometry and temperature variability. This work highlights the importance of quantifying the effects of various air quality parameters on vegetation health in urban environments in order to uncover the complexity, covariance, and interdependence of the numerous factors involved.

Identification of pyroptosis-related subtypes and establishment of prognostic model and immune characteristics in asthma.

BackgroundAlthough studies have shown that cell pyroptosis is involved in the progression of asthma, a systematic analysis of the clinical significance of pyroptosis-related genes (PRGs) cooperating with immune cells in asthma patients is still lacking.MethodsTranscriptome sequencing datasets from patients with different disease courses were used to screen pyroptosis-related differentially expressed genes and perform biological function analysis. Clustering based on K-means unsupervised clustering method is performed to identify pyroptosis-related subtypes in asthma and explore biological functional characteristics of poorly controlled subtypes. Diagnostic markers between subtypes were screened and validated using an asthma mouse model. The infiltration of immune cells in airway epithelium was evaluated based on CIBERSORT, and the correlation between diagnostic markers and immune cells was analyzed. Finally, a risk prediction model was established and experimentally verified using differentially expressed genes between pyroptosis subtypes in combination with asthma control. The cMAP database and molecular docking were utilized to predict potential therapeutic drugs.ResultsNineteen differentially expressed PRGs and two subtypes were identified between patients with mild-to-moderate and severe asthma conditions. Significant differences were observed in asthma control and FEV1 reversibility between the two subtypes. Poor control subtypes were closely related to glucocorticoid resistance and airway remodeling. BNIP3 was identified as a diagnostic marker and associated with immune cell infiltration such as, M2 macrophages. The risk prediction model containing four genes has accurate classification efficiency and prediction value. Small molecules obtained from the cMAP database that may have therapeutic effects on asthma are mainly DPP4 inhibitors.ConclusionPyroptosis and its mediated immune phenotype are crucial in the occurrence, development, and prognosis of asthma. The predictive models and drugs developed on the basis of PRGs may provide new solutions for the management of asthma.

Multi-scale visualization of incipient CaCO3 scaling on the polyamide layer of reverse osmosis membranes

Although robust models on water transport across reverse osmosis (RO) membranes exist, detailed mechanistic insights of mineral scaling in the active polyamide (PA) layer, especially through visual investigations, are still lacking. Herein, a holistic approach of imaging the onset of mineral scales in the PA-layer is detailed at different length scales. BW30 RO membranes were scaled by running CaCO3-rich feedwater in a narrow, spacer-filled crossflow channel for 120 min. Then, microscopy and spectroscopy-based characterization tools were applied to investigate the CaCO3 scaling in the PA-layer. The vaterite phase of CaCO3 dominated at the incipient stage and was amplified at higher flux. TEM imaging, followed by digital staining with unsupervised K-means clustering, demonstrated incipient scaling primarily on top of the intricate folds of the PA-layer. Scales with a characteristic size between 1 and 6 nm were also identified deep inside the randomly distributed microvoids formed by those folds, hinting at heterogeneous crystallization of CaCO3. Thus, membranes designed to limit the complex folding of the PA-layer could show more scale resistance. The improved visualization of PA-layer via TEM–K-means and resulting insights can be broadly used in designing novel scale-resistant membranes and evaluating the efficacy of membrane cleaning operations.