Binary Clustering Research Articles

An Application of Hyperspectral Image Clustering Based on Texture-Aware Superpixel Technique in Deep Sea

This paper aims to study the application of hyperspectral technology in the classification of deep-sea manganese nodules. Considering the spectral spatial variation of hyperspectral images, the difficulty of label acquisition, and the inability to guarantee stable illumination in deep-sea environments. This paper proposes a local binary pattern manifold superpixel-based fuzzy clustering method (LMSLIC-FCM). Firstly, we introduce a uniform local binary pattern (ULBP) to design a superpixel algorithm (LMSLIC) that is insensitive to illumination and has texture perception. Secondly, the weighted feature and the mean feature are fused as the representative features of superpixels. Finally, it is fused with fuzzy clustering method (FCM) to obtain a superpixel-based clustering algorithm LMSLIC-FCM. To verify the feasibility of LMSLIC-FCM on deep-sea manganese nodule data, the experiments were conducted on three different types of manganese nodule data. The average identification rate of LMSLIC-FCM reached 83.8%, and the average true positive rate reached 93.3%, which was preferable to the previous algorithms. Therefore, LMSLIC-FCM is effective in the classification of manganese nodules.

Multi-view clustering with orthogonal mapping and binary graph

The multi-view clustering problem has attracted considerable attention over recent years for the remarkable clustering performance due to exploiting complementary information from multiple views. Most existing related research work processes data in the decimal real value space that is not the most compatible space for computers. Binary code learning, also known as hashing technology, is well-known for fast Hamming distance computation, less storage requirement and accurate calculation results. The Hamming space is most enjoyed by computers because of binary/hash codes. Several studies combine multi-view clustering with binary code learning for improving clustering performance. However, there is much redundant information contained in the learned binary codes, which negatively affects the clustering performance, but these studies ignore eliminating redundant information for learning compact codes. In addition, they don’t give a unified (one-step) clustering framework with binary graph structure, which doesn’t lead to the optimal clustering result due to the information loss during the two-step process. In this paper, to cope with the two issues, we propose an orthogonal mapping binary graph method (OMBG) for the multi-view clustering problem, which makes the mapping matrix of every view orthogonalize for eliminating redundant information and embeds a binary graph structure into the unified binary multi-view clustering framework for extracting local geometric structure information of binary codes and achieving the optimal clustering result. Furthermore, we design an effective optimization algorithm based on alternating direction minimization to solve the model of OMBG. Extensive experiments performed on four frequently-used benchmark multi-view datasets illustrate the superiority of OMBG which is compared with some state-of-the-art clustering baselines.

P15.02.B Spatial habitat analysis in high-grade gliomas combining perfusion, diffusion, and hypoxia features derived from multiparametric MRI and 18F-FAZA PET

Abstract Background Heterogeneity represents one of the main issues in high-grade gliomas (HGGs) management and presumably is the key to understanding treatment failure. Spatial habitat imaging embodies a novel, non-invasive method to cluster different tumor biological microenvironments and characterize intratumoral heterogeneity through a quantitative radiomic approach. So far, habitat imaging has been chiefly explored on morphological magnetic resonance imaging (MRI): the aim of our study was to evaluate this technique on advanced imaging techniques as perfusion (PWI) and diffusion (dMRI) MRI, as well as 18F-FAZA PET, to subdivide the tumor into different biological compartments. Materials and Methods A preoperative PWI, dMRI, and 18F-FAZA PET acquisition was obtained in 17 HGG patients to assess intratumoral vascularization, cellularity, and hypoxia, respectively. A tumor mask was created including contrast-enhancing/necrotic regions on T1 imaging and surrounding FLAIR hyperintensity, and each parametric map (plasma volume from dynamic contrast-enhanced PWI, mean diffusivity from diffusion tensor imaging, and standardized uptake value from 18F-FAZA PET) underwent OTSU’s binary clustering by grouping together voxels with high and low-intensity values. Cluster intersections were then generated by combining all possible high and low-intensity regions of each map, and the final habitat map was composed. Habitats’ quantitative features, spatial distribution, and histopathological characteristics were analyzed. Results The eight habitats generated were carefully inspected, and their distribution pattern, in relation to morphological landmarks (as contrast-enhancing, necrotic, infiltrative and edematous areas), showed an excellent reproducibility among different HGGs, offering complementary data to conventional imaging. A consistent correlation was noted between each habitat’s awaited microenvironment and the corresponding histopathological characteristics among the ten patients for whom precise coordinates of stereotactic biopsies were available. Clusters considered more aggressive were more represented in WHO grade IV (vs. grade III) tumors. The habitat possibly representing edematous areas was significantly larger in IDH-1 wild-type (vs. IDH-1 mutated) tumors. Conclusions: PET and MRI hypoxia, perfusion, and tissue diffusion habitat maps proved an excellent reproducibility among HGGs and correlation with expected histopathological features, making it a way to characterize intratumoral heterogeneity. However, further study is needed to elucidate clinical implications and confirm the spatio-temporal evolution of the identified habitats. Funding Italian Ministry of Health, grant number GR-2018-12365670; Italian Association for Cancer Research (grant IG 2014 Id.1524).

A novel combination of methods identifies priority conservation areas for an endemic California Current seabird

There are growing pressures on marine biodiversity. Seabirds in particular are one the most-threatened groups. The black-vented shearwater (Puffinus opisthomelas) is endemic to Mexican islands and the only shearwater living its entire life cycle in the California Current System, one of the most productive large marine ecosystems in the world. Marine Protected Areas (MPAs) in this region, however, were designed without consideration of accurate data on seabird distributions. Here, 57 black-vented shearwaters were GPS-tracked from their main breeding colony (95% of the global population) over four seasons (2016–2019) to estimate their at-sea distribution. Two methods were applied to identify priority conservation areas: the approach developed by BirdLife International to identify marine Important Bird and Biodiversity Areas and a method using expectation-maximization binary clustering to identify core foraging areas. One potential marine Important Bird and Biodiversity Area close to the breeding colony and five core foraging areas were identified. These priority conservation areas were largely beyond the bounds of the current MPA network in the region. Our results detail opportunities for improving the implementation of conservation and management measures in the California Current System region with respect to seabirds. The approach of combining site identification methods can be applied to other seabird species for which high-resolution tracking data are available and can help guide conservation action plans and MPA design.

An efficient Berkeley’s wavelet convolutional transfer learning and local binary Gabor fuzzy C-means clustering for brain tumour detection

ABSTRACT Manual diagnosis of brain tumour tissues is particularly labourintensive as well as operator-dependent due to the intricacy of brain tissue. Traditional approaches are ineffective in the presence of these effects, necessitating the assessment of the photographs by professionals who can identify them. This research proposes a novel technique in brain tumour detection based on segmentation with classification utilizing DL architectures. Here, input has been collected as various brain slice image datasets. Initially, this image has been processed for resizing and smoothening and this image has been segmented. The segmentation has been carried out using local binary Gabor fuzzy C-means clustering. Then the segmented image has been classified for spotting the tumour using Berkeley’s wavelet convolutional transfer learning. Based on the accuracy, sensitivity, specificity, Jaccard’s coefficient, spatial overlap, AVME, and FoM, the creative outcome of the approach used was evaluated.

SimplifyEnrichment: A Bioconductor Package for Clustering and Visualizing Functional Enrichment Results

Functional enrichment analysis or gene set enrichment analysis is a basic bioinformatics method that evaluates the biological importance of a list of genes of interest. However, it may produce a long list of significant terms with highly redundant information that is difficult to summarize. Current tools to simplify enrichment results by clustering them into groups either still produce redundancy between clusters or do not retain consistent term similarities within clusters. We propose a new method named binary cut for clustering similarity matrices of functional terms. Through comprehensive benchmarks on both simulated and real-world datasets, we demonstrated that binary cut could efficiently cluster functional terms into groups where terms showed consistent similarities within groups and were mutually exclusive between groups. We compared binary cut clustering on the similarity matrices obtained from different similarity measures and found that semantic similarity worked well with binary cut, while similarity matrices based on gene overlap showed less consistent patterns. We implemented the binary cut algorithm in the R package simplifyEnrichment, which additionally provides functionalities for visualizing, summarizing, and comparing the clustering. The simplifyEnrichment package and the documentation are available at https://bioconductor.org/packages/simplifyEnrichment/.

Single-cell RNA sequencing and binary hierarchical clustering define lung interstitial macrophage heterogeneity in response to hypoxia.

Few studies have examined lung interstitial macrophage (IM) molecular phenotypes after being exposed to hypoxia in vivo at the single-cell level, even though macrophages contribute to hypoxic pulmonary hypertension (PH). We aimed to determine IM diversity and its association with hypoxia-induced PH. We hypothesized that integrating single-cell RNA sequencing (scRNAseq) and binary hierarchal clustering (BHC) could resolve IM heterogeneity under normal homeostatic conditions and changes induced by hypoxia exposure. Cx3cr1GFP/+ reporter mice were exposed to normoxic conditions (∼21% ) or exposed to 1 day (D1) or 7 days (D7) of hypoxia (∼10% ). We used flow cytometry to isolate Cx3cr1+ IMs and the 10X Genomics platform for scRNAseq, Cell Ranger, Seurat, ClusterMap, monocle, ingenuity pathway analysis, and Fisher’s exact test (q value < 0.05) for functional investigations. n = 374 (normoxia), n = 2,526 (D1), and n = 1,211 (D7) IMs were included in the analyses. We identified three normoxia-related cell types, five hypoxia-associated cell types that emerged at D1, and three that appeared at D7. We describe the existence of a putative resident trained innate IM, which is present in normoxia, transiently depleted at D1, and recovered after 7 days of sustained hypoxia. We also define a rare putative pathogenic population associated with transcripts implicated in PH development that emerges at D7. In closing, we describe the successful integration of BHC with scRNAseq to determine IM heterogeneity and its association with PH. These results shed light on how resident-trained innate IMs become more heterogeneous but ultimately accustomed to hypoxia.

Decoding atopic dermatitis: patient heterogeneity and genetic variation contributing to immune-mediated dermatologic disease in the UK Biobank

Abstract Background Dupilumab is the only approved biologic used to treat moderate to severe atopic dermatitis (AD) but has a non-response rate of 55–60%. This suggests underlying heterogeneity in AD etiology, requiring alternate therapies for patients. Study Objective Identify biomarkers and disease conditions in the UK Biobank that stratify AD patients into homogenous subgroups for genetic characterization. Methods We identified 13,137 AD cases and 13,136 age and sex-matched controls using hospital ICD9/10 and primary care codes in the UK Biobank. A total of 94 biomarkers and 1,596 disease PheCodes were used to isolate traits that vary in AD patients relative to controls; K-means (quantitative) and K-modes (binary) clustering was performed to identify subgroups within AD patients. Genetic analysis was performed for subgroups vs. controls using REGENIE. Results We observed enrichment of dermatologic, gastrointestinal, and rheumatologic conditions in AD patients, as well as differences in eosinophil percentage, Cystatin C, C-reactive protein, and erythrocyte distribution width relative to controls. There were eight subgroups of AD patients including groups with co-occurring psoriasis, asthma and gastrointestinal problems, penicillin allergy, and AD with no co-morbidities. Genetic analyses revealed novel HLA and non-HLA associations with subgroups, while also replicating known associations with FLG. Conclusion There were eight distinct AD subgroups based on co-occurring disease states. Genetic analyses show subgroup-specific factors can contribute to disease pathology—signaling potential utility for differential therapies and highlight mechanisms for targeted therapeutic development in the treatment of AD.

Pamlr: A toolbox for analysing animal behaviour using pressure, acceleration, temperature, magnetic or light data in R.

Light‐level geolocators have revolutionised the study of animal behaviour. However, lacking spatial precision, their usage has been primary targeted towards the analysis of large‐scale movements. Recent technological developments have allowed the integration of magnetometers and accelerometers into geolocator tags in addition to barometers and thermometers, offering new behavioural insights.Here, we introduce an R toolbox for identifying behavioural patterns from multisensor geolocator tags, with functions specifically designed for data visualisation, calibration, classification and error estimation. More specifically, the package allows for the flexible analysis of any combination of sensor data using k‐means clustering, expectation maximisation binary clustering, hidden Markov models and changepoint analyses. Furthermore, the package integrates tailored algorithms for identifying periods of prolonged high activity (most commonly used for identifying migratory flapping flight), and pressure changes (most commonly used for identifying dive or flight events).Finally, we highlight some of the limitations, implications and opportunities of using these methods.

Sleep staging using semi-unsupervised clustering of EEG: Application to REM sleep behavior disorder

ObjectiveTo develop a semi-unsupervised automatic sleep staging method capable of detecting novel sleep patterns beyond standard sleep stages in the EEG, as may be seen in clinical populations. MethodsWe employed a two-step approach that utilized prior knowledge extracted from labeled data to cluster unlabeled data into standard sleep stages and potential emergent EEG patterns. In the first step, 62 standard EEG features per channel from 30 s labelled epochs are obtained. Subsequently, the subset of features that best discriminate each standard sleep stage from all other stages (the most discriminative features -- MDF) are determined. In the second step, applied to unlabeled data, iterative binary clustering performed in the stage-specific MDFs, with the staging order and initial cluster centers that are obtained in the first step. In the first experiment, we tested the performance of the method on EEG data from 20 healthy subjects and compared its performance with state-of-the-art methods on the same dataset. In the second experiment, we applied the method to the data from 20 subjects with rapid eye movement (REM) sleep behavior disorder (RBD) utilizing the prior knowledge derived from 9 healthy subjects. ResultsResults of the first experiment showed that the proposed method could provide comparable performance to other semi-supervised methods across all sleep stages. Results of the second experiment showed that the prior knowledge inferred from healthy participants were transferable to RBD populations with a minimal performance drop. In addition, the step-wise binary clustering beyond standard sleep stages resulted in the discovery of a novel EEG characteristic in subjects with RBD. This was predominately seen in NREM2-3 stages and was characterized by significantly lower power in the delta band and significantly higher power in alpha, beta, theta, and sigma bands compared to normal NREM2-3. ConclusionOur results suggest that the proposed approach may fill an important gap in the situation where labels of the target data are not readily available for a fully supervised approach, but some prior knowledge is still available from related data. SignificanceLabels from healthy data can be used to still allow for investigation of clinical populations, with possible discovery of novel sleep patterns.

Multiobjective multitasking optimization assisted by multidirectional prediction method

Multiobjective multitasking optimization (MTO) is an emerging research topic in the field of evolutionary computation, which has attracted extensive attention, and many evolutionary multitasking (EMT) algorithms have been proposed. One of the core issues, designing an efficient transfer strategy, has been scarcely explored. Keeping this in mind, this paper is the first attempt to design an efficient transfer strategy based on multidirectional prediction method. Specifically, the population is divided into multiple classes by the binary clustering method, and the representative point of each class is calculated. Then, an effective prediction direction method is developed to generate multiple prediction directions by representative points. Afterward, a mutation strength adaptation method is proposed according to the improvement degree of each class. Finally, the predictive transferred solutions are generated as transfer knowledge by the prediction directions and mutation strengths. By the above process, a multiobjective EMT algorithm based on multidirectional prediction method is presented. Experiments on two MTO test suits indicate that the proposed algorithm is effective and competitive to other state-of-the-art EMT algorithms.

Unsupervised Binary Protocol Clustering Based on Maximum Sequential Patterns

With the rapid development of the Internet, a large number of private protocols emerge on the network. However, some of them are constructed by attackers to avoid being analyzed, posing a threat to computer network security. The blockchain uses the P2P protocol to implement various functions across the network. Furthermore, the P2P protocol format of blockchain may differ from the standard format specification, which leads to sniffing tools such as Wireshark and Fiddler not being able to recognize them. Therefore, the ability to distinguish different types of unknown network protocols is vital for network security. In this paper, we propose an unsupervised clustering algorithm based on maximum frequent sequences for binary protocols, which can distinguish various unknown protocols to provide support for analyzing unknown protocol formats. We mine the maximum frequent sequences of protocol message sets in bytes. And we calculate the fuzzy membership of the protocol message to each maximum frequent sequence, which is based on fuzzy set theory. Then we construct the fuzzy membership vector for each protocol message. Finally, we adopt K-means++ to split different types of protocol messages into several clusters and evaluate the performance by calculating homogeneity, integrity, and Fowlkes and Mallows Index (FMI). Besides, the clustering algorithms based on Needleman–Wunsch and the fixed-length prefix are compared with the algorithm presented in this paper. Compared with these traditional clustering methods, we demonstrate a certain improvement in the clustering performance of our work.

Adaptive Binary Bat and Markov Clustering Algorithms for Optimal Text Feature Selection in News Events Detection Model

Adaptive Binary Bat and Markov Clustering Algorithms for Optimal Text Feature Selection in News Events Detection Model

Association of childhood metabolic syndrome and metabolic phenotypes with the carotid intima-media thickness (CIMT) in early adulthood: Tehran lipid and glucose study

BackgroundThe clinical significance of metabolic syndrome (MetS) during adolescence and its association with adulthood adverse outcomes is a controversial issue. This study aimed to investigate the association of MetS and metabolic phenotypes with a high carotid intima-media thickness (CIMT) in early adulthood. MethodsMetS was defined as proposed by Cook, de Ferranti, the National Cholesterol Education Program (NCEP), and the pediatric International Diabetes Federation (IDF). Metabolic phenotypes were defined based on the binary clustering of cardiovascular risk factors. The participants were adolescent's individual selected from a large cohort study and followed for 18.2 years. Multivariate-adjusted odds ratios (ORs) were calculated for a high CIMT incidence (≥95th percentile). ResultsIn this study, 862 adolescents (52.3% males), with the mean age of 13.4 ± 2.2 years, were included. The presence of MetS, based on the definitions proposed by Cook (OR = 1.90, 95% CI: 1.01–3.57, P = 0.046) and de Ferranti (OR = 1.74; 95% CI: 1.04–2.90, P = 0.033), was associated with the increased risk of a high CIMT in early adulthood. Metabolic phenotypes, including high waist circumference (WC)/hypertension (HTN), high WC/low high-density lipoprotein-cholesterol (HDL-C), and high triglyceride (TG)/high WC, showed higher risks of a high CIMT. However, these positive associations become insignificant after adjusting for the adulthood BMI, except for the high WC/low HDL-C phenotype (OR = 2.04, 95% CI: 1.02–4.11, P = 0.044). ConclusionThe high WC/low HDL-C phenotype had a better predictive value and could be used as a simpler alternative for MetS to identify adolescents with a higher risk of high CIMT during early adulthood.

IhybCNV: An intra-hybrid approach for CNV detection from next-generation sequencing data

Copy number variation (CNV) is a common phenomenon in the human genome, accounting for a large proportion of structural variations. Accurate detection of CNVs has been regarded as a regular approach for studying cancer mechanisms and evolution and seeking targeted drugs for cancer diagnosis. Next-generation sequencing (NGS) data has provided an unprecedented opportunity for this task. Currently, a large number of methods have been developed for CNV detection by using NGS data. Each of them has its own viewpoint on the characteristics of CNVs associated with the features of NGS data. Due to the intrinsic complexity of CNVs, new methods with a comprehensive viewpoint on the characteristics of CNVs combined with NGS data are required for the detection of various forms of CNVs. In this paper, we propose an intra-hybrid approach, IhybCNV, for the detection of CNVs from NGS data. The central idea of the IhybCNV is that it regards CNVs as outlier events from various viewpoints on the read depth (RD) profile to be analyzed. More precisely, the IhybCNV integrates five individual detectors to calculate five types of outlier scores for each genome segment and then combines the five types of outlier scores into an anomaly score via locally selective combination in parallel outlier ensembles (LSCP). With the anomaly scores of the genome segments, the IhybCNV determines abnormal segments (CNVs) or normal segments by designing a binary clustering model (BCM). The performance of the IhybCNV is evaluated by carrying out simulation studies and real data applications and is compared with five peer methods. The experimental results demonstrate that the IhybCNV obtains a higher F1-score value than the other methods. Thus, the IhybCNV could be expected as a supplementary to current methods and might become a promising tool for the detection of genome mutations.

Identifying latent behavioural states in animal movement with M4, a nonparametric Bayesian method

Abstract Understanding animal movement often relies upon telemetry and biologging devices. These data are frequently used to estimate latent behavioural states to help understand why animals move across the landscape. While there are a variety of methods that make behavioural inferences from biotelemetry data, some features of these methods (e.g. analysis of a single data stream, use of parametric distributions) may limit their generality to reliably discriminate among behavioural states. To address some of the limitations of existing behavioural state estimation models, we introduce a nonparametric Bayesian framework called the mixed‐membership method for movement (M4), which is available within the open‐source bayesmove R package. This framework can analyse multiple data streams (e.g. step length, turning angle, acceleration) without relying on parametric distributions, which may capture complex behaviours more successfully than current methods. We tested our Bayesian framework using simulated trajectories and compared model performance against two segmentation methods (behavioural change point analysis (BCPA) and segclust2d), one machine learning method [expectation‐maximization binary clustering (EMbC)] and one type of state‐space model [hidden Markov model (HMM)]. We also illustrated this Bayesian framework using movements of juvenile snail kites Rostrhamus sociabilis in Florida, USA. The Bayesian framework estimated breakpoints more accurately than the other segmentation methods for tracks of different lengths. Likewise, the Bayesian framework provided more accurate estimates of behaviour than the other state estimation methods when simulations were generated from less frequently considered distributions (e.g. truncated normal, beta, uniform). Three behavioural states were estimated from snail kite movements, which were labelled as ‘encamped’, ‘area‐restricted search’ and ‘transit’. Changes in these behaviours over time were associated with known dispersal events from the nest site, as well as movements to and from possible breeding locations. Our nonparametric Bayesian framework estimated behavioural states with comparable or superior accuracy compared to the other methods when step lengths and turning angles of simulations were generated from less frequently considered distributions. Since the most appropriate parametric distributions may not be obvious a priori, methods (such as M4) that are agnostic to the underlying distributions can provide powerful alternatives to address questions in movement ecology.

Classification based on dynamic mode decomposition applied to brain recognition of context

Local Field Potentials (LFPs) are easy to access electrical signals of the brain that represent the summation in the extracellular space, of currents originated within the neurons. As such, LFPs could contain information about ongoing computations in neuronal circuits and could potentially be used to design brain machine interface algorithms. However how brain computations could be decoded from LFPs is not clear. Within this context, a methodology for signal classification is proposed in this study, particularly based on the Dynamic Mode Decomposition method, in conjunction with binary clustering routines based on supervised learning. Note that, although the classification methodology is presented here in the context of a biological problem, it can be applied to a broad range of applications. Then, as a case-study, the proposed method is validated with the classification of LFP-based brain cognitive states. All the analysis, signals, and results shown in this study consider real data measured in the hippocampus, in rats performing exploration tasks. Consequently, it is shown that, using the measured LFP, the method infers which context was the animal exploring. Thus, evidence on the spatial codification in LFP signals is consequently provided, which still is an open question in neuroscience.

Aperture: alignment-free detection of structural variations and viral integrations in circulating tumor DNA.

The identification of structural variations (SVs) and viral integrations in circulating tumor DNA (ctDNA) is a key step in precision oncology that may assist clinicians in treatment selection and monitoring. However, due to the short fragment size of ctDNA, it is challenging to accurately detect low-frequency SVs or SVs involving complex junctions in ctDNA sequencing data. Here, we describe Aperture, a new fast SV caller that applies a unique strategy of $k$-mer-based searching, binary label-based breakpoint detection and candidate clustering to detect SVs and viral integrations with high sensitivity, especially when junctions span repetitive regions. Aperture also employs a barcode-based filter to ensure specificity. Compared with existing methods, Aperture exhibits superior sensitivity and specificity in simulated, reference and real data tests, especially at low dilutions. Additionally, Aperture is able to predict sites of viral integration and identify complex SVs involving novel insertions and repetitive sequences in real patient data. Aperture is freely available at https://github.com/liuhc8/Aperture.

TAM: Targeted Analysis Model With Reinforcement Learning on Short Texts.

Mining topics on social media (e.g., Twitter and Facebook) is an important task for various applications, such as hot topic discovery, advertising, and promotion activities. Topic modeling techniques are helpful to find out topics that people are talking about. However, current full-analysis models cannot perform well on a focused analysis task-find out all topics related to one particular area in short documents. One reason is that the targeted topic is usually sparse in the corpus of short texts. Another one is, during clustering, even minor errors may compound and render the model useless. This article studies these problems and proposes a targeted analysis model (TAM) with reinforcement learning (RL) to extract any specific topic in a given corpus and perform fine-grained topic generation. In this work, we design a reward function of RL to prevent the false propagation problem induced by Gibbs sampling during the clustering. We amend the targeted topic modeling techniques to the case of RL and use policy search combined with the Gibbs EM algorithm for parameter estimation. Metrics of F1 score and the proposed normalized mutual information-F1 are exploited for the evaluation of clustering and topic generation, respectively. Our experiments have demonstrated that TAM can outperform state-of-the-art models-specifically achieving 25.7% improvement on the F1 score for binary clustering on average.

Metabolomics patterns of breast cancer tumors using mass spectrometry imaging.

Intraoperative assessment of surgical margins is important for reducing the rate of revisions in breast conserving surgery for palpable malignant tumors. The hypothesis was that metabolomics methods, based on mass spectrometry, could find patterns of relative abundances of molecules that distinguish clusters of benign tissue and cancer in surgical resections. Excisions from 8 patients were used to acquire 112,317 mass spectrometry signals by desorption electrospray ionization. A process of nonnegative matrix factorization and graph decomposition produced clusters that were approximated as affine spaces. Each signal's distance to the affine space of a cluster was used to visualize the clustering. The distance maps were superior to binary clustering in identifying cancer regions. They were particularly effective at finding cancer regions that were discontinuously distributed within benign tissue. Desorption electrospray ionization mass spectrometry, which has been shown to be useful intraoperatively, can acquire signals that distinguish malignant from benign breast tissue in surgically excised tumors. The method may be suitable for real-time surgical decisions based on cancer margins.