Data-driven Clustering Algorithm Research Articles

AutoFocus: a hierarchical framework to explore multi-omic disease associations spanning multiple scales of biomolecular interaction

Recent advances in high-throughput measurement technologies have enabled the analysis of molecular perturbations associated with disease phenotypes at the multi-omic level. Such perturbations can range in scale from fluctuations of individual molecules to entire biological pathways. Data-driven clustering algorithms have long been used to group interactions into interpretable functional modules; however, these modules are typically constrained to a fixed size or statistical cutoff. Furthermore, modules are often analyzed independently of their broader biological context. Consequently, such clustering approaches limit the ability to explore functional module associations with disease phenotypes across multiple scales. Here, we introduce AutoFocus, a data-driven method that hierarchically organizes biomolecules and tests for phenotype enrichment at every level within the hierarchy. As a result, the method allows disease-associated modules to emerge at any scale. We evaluated this approach using two datasets: First, we explored associations of biomolecules from the multi-omic QMDiab dataset (n = 388) with the well-characterized type 2 diabetes phenotype. Secondly, we utilized the ROS/MAP Alzheimer’s disease dataset (n = 500), consisting of high-throughput measurements of brain tissue to explore modules associated with multiple Alzheimer’s Disease-related phenotypes. Our method identifies modules that are multi-omic, span multiple pathways, and vary in size. We provide an interactive tool to explore this hierarchy at different levels and probe enriched modules, empowering users to examine the full hierarchy, delve into biomolecular drivers of disease phenotype within a module, and incorporate functional annotations.

The path to global climate justice: From the perspective of regional discrepancy in embodied carbon emissions

Achieving climate justice is an important goal and basic principle of global climate governance. This paper evaluates the progress of achieving global climate justice by quantitatively depicting the evolution characteristics, group differences and driving factors of global embodied carbon emissions. The results indicate that the phenomenon of embodied carbon inequality is more obvious. Furthermore, a data-driven clustering algorithm is adopted to divide the world's major economies into four convergence clubs based on the embodied carbon emissions levels. We find that the embodied carbon emissions gap between different clubs has widened since 2008, which may be related to the re-industrialization of developed countries. The results of SDA analysis indicate that the embodied carbon emissions of developed countries, such as the United States, are more driven by per capita end-use. On the contrary, unclean consumption and investment structure is the main driving force for the increase of embodied carbon emissions in developing countries such as China. Therefore, efforts should be made in developing countries on the clean adjustment of consumption and investment structure to promote the convergence of embodied carbon emissions with developed countries.

Proteomic analysis identifies subgroups of patients with active systemic lupus erythematosus

ObjectiveSystemic lupus erythematosus (SLE) is a clinically and biologically heterogenous autoimmune disease. We aimed to investigate the plasma proteome of patients with active SLE to identify novel subgroups, or endotypes, of patients.MethodPlasma was collected from patients with active SLE who were enrolled in the British Isles Lupus Assessment Group Biologics Registry (BILAG-BR). The plasma proteome was analysed using a data-independent acquisition method, Sequential Window Acquisition of All theoretical mass spectra mass spectrometry (SWATH-MS). Unsupervised, data-driven clustering algorithms were used to delineate groups of patients with a shared proteomic profile.ResultsIn 223 patients, six clusters were identified based on quantification of 581 proteins. Between the clusters, there were significant differences in age (p = 0.012) and ethnicity (p = 0.003). There was increased musculoskeletal disease activity in cluster 1 (C1), 19/27 (70.4%) (p = 0.002) and renal activity in cluster 6 (C6) 15/24 (62.5%) (p = 0.051). Anti-SSa/Ro was the only autoantibody that significantly differed between clusters (p = 0.017). C1 was associated with p21-activated kinases (PAK) and Phospholipase C (PLC) signalling. Within C1 there were two sub-clusters (C1A and C1B) defined by 49 proteins related to cytoskeletal protein binding. C2 and C6 demonstrated opposite Rho family GTPase and Rho GDI signalling. Three proteins (MZB1, SND1 and AGL) identified in C6 increased the classification of active renal disease although this did not reach statistical significance (p = 0.0617).ConclusionsUnsupervised proteomic analysis identifies clusters of patients with active SLE, that are associated with clinical and serological features, which may facilitate biomarker discovery. The observed proteomic heterogeneity further supports the need for a personalised approach to treatment in SLE.

Electrical Stimulation of Temporal and Limbic Circuitry Produces Distinct Responses in Human Ventral Temporal Cortex.

The human ventral temporal cortex (VTC) is highly connected to integrate visual perceptual inputs with feedback from cognitive and emotional networks. In this study, we used electrical brain stimulation to understand how different inputs from multiple brain regions drive unique electrophysiological responses in the VTC. We recorded intracranial EEG data in 5 patients (3 female) implanted with intracranial electrodes for epilepsy surgery evaluation. Pairs of electrodes were stimulated with single-pulse electrical stimulation, and corticocortical evoked potential responses were measured at electrodes in the collateral sulcus and lateral occipitotemporal sulcus of the VTC. Using a novel unsupervised machine learning method, we uncovered 2-4 distinct response shapes, termed basis profile curves (BPCs), at each measurement electrode in the 11-500 ms after stimulation interval. Corticocortical evoked potentials of unique shape and high amplitude were elicited following stimulation of several regions and classified into a set of four consensus BPCs across subjects. One of the consensus BPCs was primarily elicited by stimulation of the hippocampus; another by stimulation of the amygdala; a third by stimulation of lateral cortical sites, such as the middle temporal gyrus; and the final one by stimulation of multiple distributed sites. Stimulation also produced sustained high-frequency power decreases and low-frequency power increases that spanned multiple BPC categories. Characterizing distinct shapes in stimulation responses provides a novel description of connectivity to the VTC and reveals significant differences in input from cortical and limbic structures.SIGNIFICANCE STATEMENT Disentangling the numerous input influences on highly connected areas in the brain is a critical step toward understanding how brain networks work together to coordinate human behavior. Single-pulse electrical stimulation is an effective tool to accomplish this goal because the shapes and amplitudes of signals recorded from electrodes are informative of the synaptic physiology of the stimulation-driven inputs. We focused on targets in the ventral temporal cortex, an area strongly implicated in visual object perception. By using a data-driven clustering algorithm, we identified anatomic regions with distinct input connectivity profiles to the ventral temporal cortex. Examining high-frequency power changes revealed possible modulation of excitability at the recording site induced by electrical stimulation of connected regions.

Data-Driven Analysis of Cycle-to-Cycle Variations, Scatter, and Furcation in the UH-60A Wind Tunnel Rotor Airloads Measurements

A data-driven clustering algorithm based on proper orthogonal decomposition was applied to assess the scatter found in the UH-60A wind tunnel airloads measurements. Upon verifying the capability of the algorithm, pushrod loads, blade surface pressures, sectional loads, and torsional moments were analyzed. Spatial eigenmodes resulting from the decomposition provided the optimal basis; projection of the individual cycles onto the high singular value modes allowed visualizing the statistical distribution of data over the entire azimuth. While not all cases showed furcation in the data, the bimodal distribution was found in the high-thrust cases, where statistically normal distribution is generally assumed. Consequent clustering of the measured cycles produced an excellent correlation among clusters found in the pushrod loads, blade surface pressure, and torsional moment that suggests a common source for furcation in the data. The cycles assigned to one group repeatedly showed distinguishable variations from the other group in terms of the presence/absence of a dynamic stall vortex, azimuthal occurrence of the stall, chordwise location of separation, reattachment, and so on. When one of the clusters is smaller in size compared to the other, the conventional phase average obscured all the intricate features even when the loads are substantially higher than the larger cluster. In general, clustering the dataset when warranted showed not only higher peak loads but also lower variance for both the clusters across the entire azimuth compared to the conventional simple phase-average results. Computational simulations were conducted using CREATETM -AV Helios towards understanding the underlying flow field. Misinterpreted earlier as under-/overpredictive when compared with the simple phase-average data, Helios results consistently showed significantly improved correlation with one of the two clusters. Combining the clustered results and the flow visualization provided by Helios, aperiodicity in the spatial location and the strength of both the trim tab vortices and tip vortices have been hypothesized as potential sources of furcation.

Maintaining a High-Quality Screening Collection: The GSK Experience.

The storage of screening collections in DMSO is commonplace in the pharmaceutical industry. To ensure a high-quality screening collection, and hence effective and efficient high-throughput screening, all compounds entering the GlaxoSmithKline (GSK) screening collection undergo a liquid chromatography-mass spectrometry (LC-MS) quality control (QC). It is generally accepted that even under optimal conditions, a small percentage of these compounds are unstable after prolonged storage in DMSO. This article presents how these QC data can be mined using a data-driven clustering algorithm to identify chemical substructures likely to cause degradation in DMSO. This knowledge provides new structural filters for use in excluding compounds with these undesirable substructures from the collection. This information also suggests an efficient, targeted approach to compound collection clean-up initiatives. Stability studies are also designed to maintain a high-quality screening collection. To define the best practice for the storage and handling of solution samples, GSK has undertaken stability experiments for two decades, initially to support the implementation of new automated liquid stores and, subsequently, to enhance storage and use of compounds in solution through an understanding of compound degradation under storage and assay conditions.

210. Vancomycin Infusion Frequency and Intensity: Analysis of Real-World Data Generated from Automated Infusion Devices

BackgroundAutomated infusion devices captures actual infused medication administration data in real-time. Vancomycin use is now recommended to be driven by AUC (area under the curve) dosing. We evaluated automated infusion device data to depict vancomycin administration practices in acute care hospitals.Figure 1. Distribution of vancomycin infusion dosing Figure 2. Distribution of time intervals between each vancomycin infusion session (mostly around 8 or 12 hours) MethodsWe analyzed archived vancomycin infusion data from 2,417 patients captured by automated infusion systems from 3 acute care hospitals. The infusion device informatics software recorded a variety of events during infusion – starting and stopping times, alarms and alerts, vancomycin dose, and other forms of timestamped usage information. We evaluated infusion session duration and dosing, using data-driven clustering algorithms.ResultsA total of 13,339 vancomycin infusion sessions from 2,417 unique adult patients were analyzed. Approximately 26.1% of patients had just one infusion of vancomycin. For the rest of the patients, the median number of infusion sessions per patient was 4; the interquartile range was 3 and 8. The most common dose was 1.0 gram (53.7%) or 1.5 gram (24.6%) (see Figure 1). The distribution of infusion session duration (hours) was 4.2% (≤1.0 hh); 40.1% (1.01–1.5 hh); 29.1% (1.51–2.0 hh); and 26.6% (>2.0 hh). The dosing frequency was 39.5% (q8 hh), 42.9% (q12 hh), 11.1% (q24 hh), and 6.5% (>q24 hh) (Figure 2), demonstrating clinical interpretability.ConclusionA considerable number of patients received just one vancomycin infusion during their hospital stay, suggesting a potential overuse of empiric vancomycin. The majority of infusion doses were between 1 to 1.5 grams and most infusion sessions were administered every 8 or 12 hours. The actual infusion duration for each dose often exceeds the prescribed 1- or 2-hour infusion orders, which may be due to known instances of infusion interruptions due to patient movement, procedures or IV access compromise. The data generated by infusion devices can augment insights on actual antimicrobial administration practices and duration. As vancomycin AUC dosing becomes more prevalent, real world infusion data may aid timely data-driven antimicrobial stewardship and patient safety interventions for vancomycin and other AUC dosed drugs.DisclosuresCynthia Yamaga, PharmD, BD (Employee) David L. Bostick, PhD, Becton, Dickinson and Co. (Employee) Ying P. Tabak, PhD, Becton, Dickinson and Co. (Employee) Ann Liu-Ferrara, PhD, Becton, Dickinson and Co. (Employee) Didier Morel, PhD, Becton, Dickinson and Co. (Employee) Kalvin Yu, MD, Becton, Dickinson and Company (Employee)GlaxoSmithKline plc. (Other Financial or Material Support, Funding)

Reducing Uncertainty in Dynamic Stall Measurements Through Data-Driven Clustering of Cycle-To-Cycle Variations

Pitching airfoil measurements are known to exhibit significant scatter near stall angles of attack that can make meaningful correlation with modeling or simulation difficult. Application of data-driven clustering algorithms to dynamic stall experiments, conducted at two different research facilities, revealed the presence of furcation within the data scatter. Such furcation render the statistical mean and standard deviation as inadequate to represent the observed cycle-to-cycle variations. After ruling out facility effects, an alternative approach to conventional statistical analysis is developed through the use of cluster averages, associated variances, and group probability. Several existing clustering techniques are tested; however, their shortcomings led to the development of two new data-driven algorithms that use proper orthogonal decomposition to cluster data based on flow phenomena that contribute the most energy to flow variations. Several test cases are used to show the physical mechanisms leading to cycle-to-cycle variations, such as differences in separation location, boundary layer reattachment, occurrence of leading-edge/trailing-edge stall, and presence of a dynamic stall vortex (or vortices). In all cases, these physical processes and their effects are obscured by conventional phase-averaging. Further analyses on the effects of the Mach number, reduced frequency, mean angle, and amplitude of oscillation reveal trends in the probability of a given flow behavior. An initial step towards using these results for advanced semiempirical models using Markov analysis is discussed.

Hierarchy of speech-driven spectrotemporal receptive fields in human auditory cortex

Existing data indicate that cortical speech processing is hierarchically organized. Numerous studies have shown that early auditory areas encode fine acoustic details while later areas encode abstracted speech patterns. However, it remains unclear precisely what speech information is encoded across these hierarchical levels. Estimation of speech-driven spectrotemporal receptive fields (STRFs) provides a means to explore cortical speech processing in terms of acoustic or linguistic information associated with characteristic spectrotemporal patterns. Here, we estimate STRFs from cortical responses to continuous speech in fMRI. Using a novel approach based on filtering randomly-selected spectrotemporal modulations (STMs) from aurally-presented sentences, STRFs were estimated for a group of listeners and categorized using a data-driven clustering algorithm. ‘Behavioral STRFs’ highlighting STMs crucial for speech recognition were derived from intelligibility judgments. Clustering revealed that STRFs in the supratemporal plane represented a broad range of STMs, while STRFs in the lateral temporal lobe represented circumscribed STM patterns important to intelligibility. Detailed analysis recovered a bilateral organization with posterior-lateral regions preferentially processing STMs associated with phonological information and anterior-lateral regions preferentially processing STMs associated with word- and phrase-level information. Regions in lateral Heschl's gyrus preferentially processed STMs associated with vocalic information (pitch).

Handling interpretability issues in ANFIS using rule base simplification and constrained learning

Adaptive neuro-fuzzy inference system (ANFIS) is a well-known neuro-fuzzy model for approximating highly complex non-linear systems. ANFIS uses precise fuzzy modelling concept that aims at accuracy of a fuzzy model being designed than on interpretability. But interpretability of a fuzzy system is an equally important aspect of fuzzy modelling as accuracy. So far the research based on ANFIS has been mostly application based due to which the various issues related to the interpretability of ANFIS have not been dealt with. The rule base of ANFIS is typically obtained using data driven clustering algorithms. But this process introduces redundancy in the system in terms of similar fuzzy sets and redundant fuzzy rules which unnecessarily increases system complexity. This in turn reduces both the interpretability as well as generalization capability of ANFIS. Additionally in case of ANFIS, unconstrained gradient descent based learning algorithms are used to fine-tune the membership function parameters which usually result in a rule base with inconsistent, excessively overlapping and indistinguishable membership functions for input variables and thus the interpretability of the final optimized system is not guaranteed. This paper is based on addressing the issue of rule base redundancy in ANFIS to reduce complexity and enforcing constraints during learning phase to ensure interpretability of the final optimized system. Rule base redundancy is removed using similarity analysis based rule base simplification in which similar fuzzy sets are merged and subsequently resulting fuzzy rules with equal premises are combined. Hybrid learning technique which is an efficient parameter tuning method for ANFIS is constrained to prevent inconsistency, excessive overlapping and inclusion of membership functions so that the final fuzzy partitions of inputs stay interpretable. The empirical analysis of the impact of rule base simplification and constrained learning on ANFIS is done by application to two well-known benchmark problems and a real world stock price prediction problem. The introduction of rule base simplification and constrained learning in ANFIS modeling has shown better results in terms of obtaining a desired accuracy-interpretability tradeoff than conventional ANFIS.

Data-Driven Subtyping of Executive Function–Related Behavioral Problems in Children

ObjectiveExecutive functions (EF) are cognitive skills that are important for regulating behavior and for achieving goals. Executive function deficits are common in children who struggle in school and are associated with multiple neurodevelopmental disorders. However, there is also considerable heterogeneity across children, even within diagnostic categories. This study took a data-driven approach to identify distinct clusters of children with common profiles of EF-related difficulties, and then identified patterns of brain organization that distinguish these data-driven groups.MethodThe sample consisted of 442 children identified by health and educational professionals as having difficulties in attention, learning, and/or memory. We applied community clustering, a data-driven clustering algorithm, to group children by similarities on a commonly used rating scale of EF-associated behavioral difficulties, the Conners 3 questionnaire. We then investigated whether the groups identified by the algorithm could be distinguished on white matter connectivity using a structural connectomics approach combined with partial least squares analysis.ResultsThe data-driven clustering yielded 3 distinct groups of children with symptoms of one of the following: (1) elevated inattention and hyperactivity/impulsivity, and poor EF; (2) learning problems; or (3) aggressive behavior and problems with peer relationships. These groups were associated with significant interindividual variation in white matter connectivity of the prefrontal and anterior cingulate cortices.ConclusionIn sum, data-driven classification of EF-related behavioral difficulties identified stable groups of children, provided a good account of interindividual differences, and aligned closely with underlying neurobiological substrates.

Confidence Set for Group Membership

We develop new procedures to quantify the statistical uncertainty of data-driven clustering algorithms. In our panel setting, each unit belongs to one of a finite number of latent groups with group-specific regression curves. We propose methods for computing unit-wise and joint confidence sets for group membership. The unit-wise sets give possible group memberships for a given unit and the joint sets give possible vectors of group memberships for all units. We also propose an algorithm that can improve the power of our procedures by detecting units that are easy to classify. The confidence sets invert a test for group membership that is based on a characterization of the true group memberships by a system of moment inequalities. To construct the joint confidence, we solve a high-dimensional testing problem that tests group membership simultaneously for all units. We justify this procedure under N, T → ∞ asymptotics where we allow T to be much smaller than N. As part of our theoretical arguments, we develop new simultaneous anti-concentration inequalities for the MAX and the QLR statistics. Monte Carlo results indicate that our confidence sets have adequate coverage and are informative. We illustrate the practical relevance of our confidence sets in two applications.

Burst detection in district metering areas using a data driven clustering algorithm

This paper describes a novel methodology for burst detection in a water distribution system. The proposed method has two stages. In the first stage, a clustering algorithm was employed for outlier detection, while the second stage identified the presence of bursts. An important feature of this method is that data analysis is carried out dependent on multiple flow meters whose measurements vary simultaneously in a district metering area (DMA). Moreover, the clustering-based method can automatically cope with non-stationary conditions in historical data; namely, the method has no prior data selection process. An example application of this method has been implemented to confirm that relatively large bursts (simulated by flushing) with short duration can be detected effectively. Noticeably, the method has a low false positive rate compared with previous studies and appearance of detected abnormal water usage consists with weather changes, showing great promise in real application to multi-inlet and multi-outlet DMAs.

Clustering of Resting State Networks

BackgroundThe goal of the study was to demonstrate a hierarchical structure of resting state activity in the healthy brain using a data-driven clustering algorithm.Methodology/Principal FindingsThe fuzzy-c-means clustering algorithm was applied to resting state fMRI data in cortical and subcortical gray matter from two groups acquired separately, one of 17 healthy individuals and the second of 21 healthy individuals. Different numbers of clusters and different starting conditions were used. A cluster dispersion measure determined the optimal numbers of clusters. An inner product metric provided a measure of similarity between different clusters. The two cluster result found the task-negative and task-positive systems. The cluster dispersion measure was minimized with seven and eleven clusters. Each of the clusters in the seven and eleven cluster result was associated with either the task-negative or task-positive system. Applying the algorithm to find seven clusters recovered previously described resting state networks, including the default mode network, frontoparietal control network, ventral and dorsal attention networks, somatomotor, visual, and language networks. The language and ventral attention networks had significant subcortical involvement. This parcellation was consistently found in a large majority of algorithm runs under different conditions and was robust to different methods of initialization.Conclusions/SignificanceThe clustering of resting state activity using different optimal numbers of clusters identified resting state networks comparable to previously obtained results. This work reinforces the observation that resting state networks are hierarchically organized.

Ellipsoidal support vector clustering for functional MRI analysis

As an exploratory approach, the clustering of fMRI time series has proved its effectiveness in analyzing the functional MRI, especially in the detection of activated regions. Due to the arbitrary distribution of fMRI time series in the temporal domain, imposing simple assumption on the data structure usually could be misleading and limit the detector's performance. Therefore, a true data-driven clustering algorithm that adapts to the data structure is preferred, and only high-level control over the clustering procedure is desired. Support vector clustering (SVC) is a suitable one in some extent because of its advantages, such as no cluster shape restriction, no need to explicitly specify the number of clusters, and the mechanism in outlier elimination. In this work, we propose an extension of the SVC to step further toward a data-sensitive detector. This approach is named as ellipsoidal support vector clustering (ESVC). To be robust to noise, the clustering is performed on features extracted from the fMRI time series via Fourier transform. Experimental results on simulated and real data sets demonstrate the effectiveness of incorporating data structure in clustering fMRI time series.

Study-level wavelet cluster analysis and data-driven signal models in pharmacological MRI

In pharmacological MRI (phMRI) studies tracking signal changes following the acute administration of a compound, the spatiotemporal pattern of response is often unknown a priori. Moreover, when analysed within a general linear model (GLM) framework, the experimental paradigm of a single injection point under-informs the construction of an appropriate signal model, and information from pharmacokinetics or ancillary in vivo studies may be unavailable or insufficient to accurately describe the dynamic signal changes observed following injection of the drug. Here, we extend the application of a data-driven clustering algorithm, wavelet cluster analysis (WCA), to phMRI data from one or more groups of subjects in a study. A WCA decomposition of spatially concatenated time series’ provides a compact overview of spatiotemporal response patterns across cohorts, highlighting typical temporal signatures, brain regions implicated in the response and inter-subject variability. Further, we demonstrate the use of regressors based on selected temporal components as suitable signal models in GLM-based analyses, resulting in a close fit to dynamic phMRI signal changes. This approach is illustrated with simulated data and two representative in vivo phMRI studies in the rat (nicotine and apomorphine challenges).

Oxidation of propylene to acrolein over Cu-on-Al 2O 3 catalysts modified by molybdenum oxides

Adaptive neuro-fuzzy inference system (ANFIS) is a well-known neuro-fuzzy model for approximating highly complex non-linear systems. ANFIS uses precise fuzzy modelling concept that aims at accuracy of a fuzzy model being designed than on interpretability. But interpretability of a fuzzy system is an equally important aspect of fuzzy modelling as accuracy. So far the research based on ANFIS has been mostly application based due to which the various issues related to the interpretability of ANFIS have not been dealt with. The rule base of ANFIS is typically obtained using data driven clustering algorithms. But this process introduces redundancy in the system in terms of similar fuzzy sets and redundant fuzzy rules which unnecessarily increases system complexity. This in turn reduces both the interpretability as well as generalization capability of ANFIS. Additionally in case of ANFIS, unconstrained gradient descent based learning algorithms are used to fine-tune the membership function parameters which usually result in a rule base with inconsistent, excessively overlapping and indistinguishable membership functions for input variables and thus the interpretability of the final optimized system is not guaranteed. This paper is based on addressing the issue of rule base redundancy in ANFIS to reduce complexity and enforcing constraints during learning phase to ensure interpretability of the final optimized system. Rule base redundancy is removed using similarity analysis based rule base simplification in which similar fuzzy sets are merged and subsequently resulting fuzzy rules with equal premises are combined. Hybrid learning technique which is an efficient parameter tuning method for ANFIS is constrained to prevent inconsistency, excessive overlapping and inclusion of membership functions so that the final fuzzy partitions of inputs stay interpretable. The empirical analysis of the impact of rule base simplification and constrained learning on ANFIS is done by application to two well-known benchmark problems and a real world stock price prediction problem. The introduction of rule base simplification and constrained learning in ANFIS modeling has shown better results in terms of obtaining a desired accuracy-interpretability tradeoff than conventional ANFIS.