Semi-supervised clustering methods

Prevalence of Specific Mood Profile Clusters among Elite and Youth Athletes at a Brazilian Sports Club

Those responsible for elite and youth athletes are increasingly aware of the need to balance the quest for superior performance with the need to protect the physical and psychological wellbeing of athletes. As a result, regular assessment of risks to mental health is a common feature in sports organisations. In the present study, the Brazil Mood Scale (BRAMS) was administered to 898 athletes (387 female, 511 male, age range: 12-44 years) at a leading sports club in Rio de Janeiro using either "past week" or "right now" response timeframes. Using seeded k-means cluster analysis, six distinct mood profile clusters were identified, referred to as the iceberg, surface, submerged, shark fin, inverse iceberg, and inverse Everest profiles. The latter three profiles, which are associated with varying degrees of increased risk to mental health, were reported by 238 athletes (26.5%). The prevalence of these three mood clusters varied according to the response timeframe (past week > right now) and the sex of the athletes (female > male). The prevalence of the iceberg profile varied by athlete sex (male > female), and age (12-17 years > 18+ years). Findings supported use of the BRAMS as a screening tool for the risk of psychological issues among athletes in Brazilian sports organisations.

Model-based multifacet clustering with high-dimensional omics applications

High-dimensional omics data often contain intricate and multifaceted information, resulting in the coexistence of multiple plausible sample partitions based on different subsets of selected features. Conventional clustering methods typically yield only one clustering solution, limiting their capacity to fully capture all facets of cluster structures in high-dimensional data. To address this challenge, we propose a model-based multifacet clustering (MFClust) method based on a mixture of Gaussian mixture models, where the former mixture achieves facet assignment for gene features and the latter mixture determines cluster assignment of samples. We demonstrate superior facet and cluster assignment accuracy of MFClust through simulation studies. The proposed method is applied to three transcriptomic applications from postmortem brain and lung disease studies. The result captures multifacet clustering structures associated with critical clinical variables and provides intriguing biological insights for further hypothesis generation and discovery.

AnnoSpat annotates cell types and quantifies cellular arrangements from spatial proteomics

Cellular composition and anatomical organization influence normal and aberrant organ functions. Emerging spatial single-cell proteomic assays such as Image Mass Cytometry (IMC) and Co-Detection by Indexing (CODEX) have facilitated the study of cellular composition and organization by enabling high-throughput measurement of cells and their localization directly in intact tissues. However, annotation of cell types and quantification of their relative localization in tissues remain challenging. To address these unmet needs for atlas-scale datasets like Human Pancreas Analysis Program (HPAP), we develop AnnoSpat (Annotator and Spatial Pattern Finder) that uses neural network and point process algorithms to automatically identify cell types and quantify cell-cell proximity relationships. Our study of data from IMC and CODEX shows the higher performance of AnnoSpat in rapid and accurate annotation of cell types compared to alternative approaches. Moreover, the application of AnnoSpat to type 1 diabetic, non-diabetic autoantibody-positive, and non-diabetic organ donor cohorts recapitulates known islet pathobiology and shows differential dynamics of pancreatic polypeptide (PP) cell abundance and CD8+ T cells infiltration in islets during type 1 diabetes progression.

Clustering Research Proposal Submissions to Understand the Unmet Needs of Military Clinicians

INTRODUCTION

The Advanced Medical Technology Initiative (AMTI) program solicits research proposals for technology demonstrations and performance improvement projects in the domain of military medicine. Advanced Medical Technology Initiative is managed by the U.S. Army Telemedicine and Advanced Technology Research Center (TATRC). Advanced Medical Technology Initiative proposals span a wide range of topics, for example, treatment of musculoskeletal injury, application of virtual health technology, and demonstration of medical robots. The variety and distribution of central topics in these proposals (problems to be solved and technological solutions proposed) are not well characterized. Characterizing this content over time could highlight over- and under-served problem domains, inspire new technological applications, and inform future research solicitation efforts.

METHODS AND MATERIALS

This research sought to analyze and categorize historic AMTI proposals from 2010 to 2022 (n = 825). The analysis focused specifically on the "Problem to Be Solved" and "Technology to Demonstrated" sections of the proposals, whose categorizations are referred to as "Problem-Sets" and Solution-Sets" (PS and SS), respectively. A semi-supervised document clustering process was applied independently to the two sections. The process consisted of three stages: (1) Manual Document Annotation-a sample of proposals were manually labeled along each thematic axis; (2) Clustering-semi-supervised clustering, informed by the manually annotated sample, was applied to the proposals to produce document clusters; (3) Evaluation and Selection-quantitative and qualitative means were used to evaluate and select an optimal cluster solution. The results of the clustering were then summarized and presented descriptively.

RESULTS

The results of the clustering process identified 24 unique PS and 20 unique SS. The most prevalent PS were Musculoskeletal Injury (12%), Traumatic Injury (11%), and Healthcare Systems Optimization (11%). The most prevalent SS were Sensing and Imaging Technology (27%), Virtual Health (23%), and Physical and Virtual Simulation (11.5%). The most common problem-solution pair was Healthcare Systems Optimization-Virtual Health, followed by Musculoskeletal Injury-Sensing and Imaging Technology. The analysis revealed that problem-solution-set co-occurrences were well distributed throughout the domain space, demonstrating the variety of research conducted in this research domain.

CONCLUSIONS

A semi-supervised document clustering approach was applied to a repository of proposals to partially automate the process of document annotation. By applying this process, we successfully extracted thematic content from the proposals related to problems to be addressed and proposed technological solutions. This analysis provides a snapshot of the research supply in the domain of military medicine over the last 12 years. Future work should seek to replicate and improve the document clustering process used. Future efforts should also be made to compare these results to actual published work in the domain of military medicine, revealing differences in demand for research as determined by funding and publishing decision-makers and supply by researchers.

Exploring the Steps of Infrared (IR) Spectral Analysis: Pre-Processing, (Classical) Data Modelling, and Deep Learning

Infrared (IR) spectroscopy has greatly improved the ability to study biomedical samples because IR spectroscopy measures how molecules interact with infrared light, providing a measurement of the vibrational states of the molecules. Therefore, the resulting IR spectrum provides a unique vibrational fingerprint of the sample. This characteristic makes IR spectroscopy an invaluable and versatile technology for detecting a wide variety of chemicals and is widely used in biological, chemical, and medical scenarios. These include, but are not limited to, micro-organism identification, clinical diagnosis, and explosive detection. However, IR spectroscopy is susceptible to various interfering factors such as scattering, reflection, and interference, which manifest themselves as baseline, band distortion, and intensity changes in the measured IR spectra. Combined with the absorption information of the molecules of interest, these interferences prevent direct data interpretation based on the Beer-Lambert law. Instead, more advanced data analysis approaches, particularly artificial intelligence (AI)-based algorithms, are required to remove the interfering contributions and, more importantly, to translate the spectral signals into high-level biological/chemical information. This leads to the tasks of spectral pre-processing and data modeling, the main topics of this review. In particular, we will discuss recent developments in both tasks from the perspectives of classical machine learning and deep learning.

An overview of clustering methods with guidelines for application in mental health research

Cluster analyzes have been widely used in mental health research to decompose inter-individual heterogeneity by identifying more homogeneous subgroups of individuals. However, despite advances in new algorithms and increasing popularity, there is little guidance on model choice, analytical framework and reporting requirements. In this paper, we aimed to address this gap by introducing the philosophy, design, advantages/disadvantages and implementation of major algorithms that are particularly relevant in mental health research. Extensions of basic models, such as kernel methods, deep learning, semi-supervised clustering, and clustering ensembles are subsequently introduced. How to choose algorithms to address common issues as well as methods for pre-clustering data processing, clustering evaluation and validation are then discussed. Importantly, we also provide general guidance on clustering workflow and reporting requirements. To facilitate the implementation of different algorithms, we provide information on R functions and libraries.

Machine Learning Methods for Small Data Challenges in Molecular Science

Small data are often used in scientific and engineering research due to the presence of various constraints, such as time, cost, ethics, privacy, security, and technical limitations in data acquisition. However, big data have been the focus for the past decade, small data and their challenges have received little attention, even though they are technically more severe in machine learning (ML) and deep learning (DL) studies. Overall, the small data challenge is often compounded by issues, such as data diversity, imputation, noise, imbalance, and high-dimensionality. Fortunately, the current big data era is characterized by technological breakthroughs in ML, DL, and artificial intelligence (AI), which enable data-driven scientific discovery, and many advanced ML and DL technologies developed for big data have inadvertently provided solutions for small data problems. As a result, significant progress has been made in ML and DL for small data challenges in the past decade. In this review, we summarize and analyze several emerging potential solutions to small data challenges in molecular science, including chemical and biological sciences. We review both basic machine learning algorithms, such as linear regression, logistic regression (LR), k-nearest neighbor (KNN), support vector machine (SVM), kernel learning (KL), random forest (RF), and gradient boosting trees (GBT), and more advanced techniques, including artificial neural network (ANN), convolutional neural network (CNN), U-Net, graph neural network (GNN), Generative Adversarial Network (GAN), long short-term memory (LSTM), autoencoder, transformer, transfer learning, active learning, graph-based semi-supervised learning, combining deep learning with traditional machine learning, and physical model-based data augmentation. We also briefly discuss the latest advances in these methods. Finally, we conclude the survey with a discussion of promising trends in small data challenges in molecular science.

Semi-supervised sentiment clustering on natural language texts

In this paper, we propose a semi-supervised method to cluster unstructured textual data called semi-supervised sentiment clustering on natural language texts. The aim is to identify clusters homogeneous with respect to the overall sentiment of the texts analyzed. The method combines different techniques and methodologies: Sentiment Analysis, Threshold-based Naïve Bayes classifier, and Network-based Semi-supervised Clustering. It involves different steps. In the first step, the unstructured text is transformed into structured text, and it is categorized into positive or negative classes using a sentiment analysis algorithm. In the second step, the Threshold-based Naïve Bayes classifier is applied to identify the overall sentiment of the texts and to define a specific sentiment value for the topics. In the last step, Network-based Semi-supervised Clustering is applied to partition the instances into disjoint groups. The proposed algorithm is tested on a collection of reviews written by customers on Booking.com. The results have highlighted the capacity of the proposed algorithm to identify clusters that are distinct, non-overlapped, and homogeneous with respect to the overall sentiment. Results are also easily interpretable thanks to the network representation of the instances that helps to understand the relationship between them.

HETEROGENEITY ANALYSIS VIA INTEGRATING MULTI-SOURCES HIGH-DIMENSIONAL DATA WITH APPLICATIONS TO CANCER STUDIES

This study has been motivated by cancer research, in which heterogeneity analysis plays an important role and can be roughly classified as unsupervised or supervised. In supervised heterogeneity analysis, the finite mixture of regression (FMR) technique is used extensively, under which the covariates affect the response differently in subgroups. High-dimensional molecular and, very recently, histopathological imaging features have been analyzed separately and shown to be effective for heterogeneity analysis. For simpler analysis, they have been shown to contain overlapping, but also independent information. In this article, our goal is to conduct the first and more effective FMR-based cancer heterogeneity analysis by integrating high-dimensional molecular and histopathological imaging features. A penalization approach is developed to regularize estimation, select relevant variables, and, equally importantly, promote the identification of independent information. Consistency properties are rigorously established. An effective computational algorithm is developed. A simulation and an analysis of The Cancer Genome Atlas (TCGA) lung cancer data demonstrate the practical effectiveness of the proposed approach. Overall, this study provides a practical and useful new way of conducting supervised cancer heterogeneity analysis.

Comparing Clustering Methods Applied to Tinnitus within a Bootstrapped and Diagnostic-Driven Semi-Supervised Framework

The understanding of tinnitus has always been elusive and is largely prevented by its intrinsic heterogeneity. To address this issue, scientific research has aimed at defining stable and easily identifiable subphenotypes of tinnitus. This would allow better disentangling the multiple underlying pathophysiological mechanisms of tinnitus. In this study, three-dimensionality reduction techniques and two clustering methods were benchmarked on a database of 2772 tinnitus patients in order to obtain a reliable segmentation of subphenotypes. In this database, tinnitus patients’ endotypes (i.e., parts of a population with a condition with distinct underlying mechanisms) are reported when diagnosed by an ENT expert in tinnitus management. This partial labeling of the dataset enabled the design of an original semi-supervised framework. The objective was to perform a benchmark of different clustering methods to get as close as possible to the initial ENT expert endotypes. To do so, two metrics were used: a primary one, the quality of the separation of the endotypes already identified in the database, as well as a secondary one, the stability of the obtained clusterings. The relevance of the results was finally reviewed by two ENT experts in tinnitus management. A 20-cluster clustering was selected as the best-performing, the most-clinically relevant, and the most-stable through bootstrapping. This clustering used a T-SNE method as the dimensionality reduction technique and a k-means algorithm as the clustering method. The characteristics of this clustering are presented in this article.

A new method based on ensemble time series for fast and accurate clustering

Purpose

The common methods for clustering time series are the use of specific distance criteria or the use of standard clustering algorithms. Ensemble clustering is one of the common techniques used in data mining to increase the accuracy of clustering. In this study, based on segmentation, selecting the best segments, and using ensemble clustering for selected segments, a multistep approach has been developed for the whole clustering of time series data.

Design/methodology/approach

First, this approach divides the time series dataset into equal segments. In the next step, using one or more internal clustering criteria, the best segments are selected, and then the selected segments are combined for final clustering. By using a loop and how to select the best segments for the final clustering (using one criterion or several criteria simultaneously), two algorithms have been developed in different settings. A logarithmic relationship limits the number of segments created in the loop.

Finding

According to Rand's external criteria and statistical tests, at first, the best setting of the two developed algorithms has been selected. Then this setting has been compared to different algorithms in the literature on clustering accuracy and execution time. The obtained results indicate more accuracy and less execution time for the proposed approach.

Originality/value

This paper proposed a fast and accurate approach for time series clustering in three main steps. This is the first work that uses a combination of segmentation and ensemble clustering. More accuracy and less execution time are the remarkable achievements of this study.

AnnoSpat annotates cell types and quantifies cellular arrangements from spatial proteomics

Cellular composition and anatomical organization influence normal and aberrant organ functions. Emerging spatial single-cell proteomic assays such as Image Mass Cytometry (IMC) and Co-Detection by Indexing (CODEX) have facilitated the study of cellular composition and organization by enabling high-throughput measurement of cells and their localization directly in intact tissues. However, annotation of cell types and quantification of their relative localization in tissues remain challenging. To address these unmet needs, we developed AnnoSpat (Annotator and Spatial Pattern Finder) that uses neural network and point process algorithms to automatically identify cell types and quantify cell-cell proximity relationships. Our study of data from IMC and CODEX show the superior performance of AnnoSpat in rapid and accurate annotation of cell types compared to alternative approaches. Moreover, the application of AnnoSpat to type 1 diabetic, non-diabetic autoantibody-positive, and non-diabetic organ donor cohorts recapitulated known islet pathobiology and showed differential dynamics of pancreatic polypeptide (PP) cell abundance and CD8+ T cells infiltration in islets during type 1 diabetes progression.

Research proposal content extraction using natural language processing and semi-supervised clustering: A demonstration and comparative analysis

Funding institutions often solicit text-based research proposals to evaluate potential recipients. Leveraging the information contained in these documents could help institutions understand the supply of research within their domain. In this work, an end-to-end methodology for semi-supervised document clustering is introduced to partially automate classification of research proposals based on thematic areas of interest. The methodology consists of three stages: (1) manual annotation of a document sample; (2) semi-supervised clustering of documents; (3) evaluation of cluster results using quantitative metrics and qualitative ratings (coherence, relevance, distinctiveness) by experts. The methodology is described in detail to encourage replication and is demonstrated on a real-world data set. This demonstration sought to categorize proposals submitted to the US Army Telemedicine and Advanced Technology Research Center (TATRC) related to technological innovations in military medicine. A comparative analysis of method features was performed, including unsupervised vs. semi-supervised clustering, several document vectorization techniques, and several cluster result selection strategies. Outcomes suggest that pretrained Bidirectional Encoder Representations from Transformers (BERT) embeddings were better suited for the task than older text embedding techniques. When comparing expert ratings between algorithms, semi-supervised clustering produced coherence ratings ~ 25% better on average compared to standard unsupervised clustering with negligible differences in cluster distinctiveness. Last, it was shown that a cluster result selection strategy that balances internal and external validity produced ideal results. With further refinement, this methodological framework shows promise as a useful analytical tool for institutions to unlock hidden insights from untapped archives and similar administrative document repositories.

Supplementary Information

The online version contains supplementary material available at 10.1007/s11192-023-04689-3.

A systematic literature review of cyber-security data repositories and performance assessment metrics for semi-supervised learning

In Machine Learning, the datasets used to build models are one of the main factors limiting what these models can achieve and how good their predictive performance is. Machine Learning applications for cyber-security or computer security are numerous including cyber threat mitigation and security infrastructure enhancement through pattern recognition, real-time attack detection, and in-depth penetration testing. Therefore, for these applications in particular, the datasets used to build the models must be carefully thought to be representative of real-world data. However, because of the scarcity of labelled data and the cost of manually labelling positive examples, there is a growing corpus of literature utilizing Semi-Supervised Learning with cyber-security data repositories. In this work, we provide a comprehensive overview of publicly available data repositories and datasets used for building computer security or cyber-security systems based on Semi-Supervised Learning, where only a few labels are necessary or available for building strong models. We highlight the strengths and limitations of the data repositories and sets and provide an analysis of the performance assessment metrics used to evaluate the built models. Finally, we discuss open challenges and provide future research directions for using cyber-security datasets and evaluating models built upon them.

Multiobjective semisupervised learning with a right-censored endpoint adapted to the multiple imputation framework

Semisupervised learning aims to use additional knowledge in the search for data structure. In clinical applications, including predictive information in the construction of a data-driven classification is of major importance. This work was motivated by a study that aimed to identify different patterns of immune parameters that would be associated with relapse-free survival in a cohort of breast cancer patients. Supervised and unsupervised objectives can be concomitantly optimized using multiobjective optimization. We propose such a procedure that addresses two challenges in the semisupervised approach, that is, missing data and additional knowledge based on survival time. The former was handled by using multiple imputation and consensus clustering. Survival information was incorporated in the supervised objective through the estimation of a cross-validation error of a Cox regression. A simulation study was performed to assess the performance of the proposed procedure. On complete datasets, the performances were compared to those of an existing modified multiobjective semisupervised learning method. The added value of including the survival data in the learning process was assessed by comparing the procedure to unsupervised learning. The proposed procedure showed better performance than the existing method, notably in the selection of the number of clusters. On incomplete datasets, the procedure showed little sensitivity to most of its parameters, even though a high number of imputations and partition initialization seeds improved the performance. The performance was degraded with a high proportion of missing data (40%) and with more ambiguous data structures. Simulation results and application on real data support the conclusion that our procedure enables the construction of a classification associated with a right-censored endpoint on a possibly incomplete dataset.

Defining biologically relevant and hierarchically nested population units to inform wildlife management

Wildlife populations are increasingly affected by natural and anthropogenic changes that negatively alter biotic and abiotic processes at multiple spatiotemporal scales and therefore require increased wildlife management and conservation efforts. However, wildlife management boundaries frequently lack biological context and mechanisms to assess demographic data across the multiple spatiotemporal scales influencing populations. To address these limitations, we developed a novel approach to define biologically relevant subpopulations of hierarchically nested population levels that could facilitate managing and conserving wildlife populations and habitats. Our approach relied on the Spatial "K"luster Analysis by Tree Edge Removal clustering algorithm, which we applied in an agglomerative manner (bottom-to-top). We modified the clustering algorithm using a workflow and population structure tiers from least-cost paths, which captured biological inferences of habitat conditions (functional connectivity), dispersal capabilities (potential connectivity), genetic information, and functional processes affecting movements. The approach uniquely included context of habitat resources (biotic and abiotic) summarized at multiple spatial scales surrounding locations with breeding site fidelity and constraint-based rules (number of sites grouped and population structure tiers). We applied our approach to greater sage-grouse (Centrocercus urophasianus), a species of conservation concern, across their range within the western United States. This case study produced 13 hierarchically nested population levels (akin to cluster levels, each representing a collection of subpopulations of an increasing number of breeding sites). These closely approximated population closure at finer ecological scales (smaller subpopulation extents with fewer breeding sites; cluster levels ≥2), where >92% of individual sage-grouse's time occurred within their home cluster. With available population monitoring data, our approaches can support the investigation of factors affecting population dynamics at multiple scales and assist managers with making informed, targeted, and cost-effective decisions within an adaptive management framework. Importantly, our approach provides the flexibility of including species-relevant context, thereby supporting other wildlife characterized by site fidelity.

Trajectories of imitation skills in preschoolers with autism spectrum disorders

Background

Imitation skills play a crucial role in social cognitive development from early childhood. Many studies have shown a deficit in imitation skills in children with autism spectrum disorders (ASD). Little is known about the development of imitation behaviors in children with ASD. This study aims to measure the trajectories of early imitation skills in preschoolers with ASD and how these skills impact other areas of early development.

Methods

For this purpose, we assessed imitation, language, and cognition skills in 177 children with ASD and 43 typically developing children (TD) aged 2 to 5 years old, 126 of which were followed longitudinally, yielding a total of 396 time points.

Results

Our results confirmed the presence of an early imitation deficit in toddlers with ASD compared to TD children. The study of the trajectories showed that these difficulties were marked at the age of 2 years and gradually decreased until the age of 5 years old. Imitation skills were strongly linked with cognitive and language skills and level of symptoms in our ASD group at baseline. Moreover, the imitation skills at baseline were predictive of the language gains a year later in our ASD group. Using a data-driven clustering method, we delineated different developmental trajectories of imitation skills within the ASD group.

Conclusions

The clinical implications of the findings are discussed, particularly the impact of an early imitation deficit on other areas of competence of the young child.

Supplementary Information

The online version contains supplementary material available at 10.1186/s11689-021-09412-y.

Stability estimation for unsupervised clustering: A review

Cluster analysis remains one of the most challenging yet fundamental tasks in unsupervised learning. This is due in part to the fact that there are no labels or gold standards by which performance can be measured. Moreover, the wide range of clustering methods available is governed by different objective functions, different parameters, and dissimilarity measures. The purpose of clustering is versatile, often playing critical roles in the early stages of exploratory data analysis and as an endpoint for knowledge and discovery. Thus, understanding the quality of a clustering is of critical importance. The concept of stability has emerged as a strategy for assessing the performance and reproducibility of data clustering. The key idea is to produce perturbed data sets that are very close to the original, and cluster them. If the clustering is stable, then the clusters from the original data will be preserved in the perturbed data clustering. The nature of the perturbation, and the methods for quantifying similarity between clusterings, are nontrivial, and ultimately what distinguishes many of the stability estimation methods apart. In this review, we provide an overview of the very active research area of cluster stability estimation and discuss some of the open questions and challenges that remain in the field. This article is categorized under:Statistical Learning and Exploratory Methods of the Data Sciences > Clustering and Classification.

Towards precision cardiometabolic prevention: results from a machine learning, semi-supervised clustering approach in the nationwide population-based ORISCAV-LUX 2 study

Given the rapid increase in the incidence of cardiometabolic conditions, there is an urgent need for better approaches to prevent as many cases as possible and move from a one-size-fits-all approach to a precision cardiometabolic prevention strategy in the general population. We used data from ORISCAV-LUX 2, a nationwide, cross-sectional, population-based study. On the 1356 participants, we used a machine learning semi-supervised cluster method guided by body mass index (BMI) and glycated hemoglobin (HbA1c), and a set of 29 cardiometabolic variables, to identify subgroups of interest for cardiometabolic health. Cluster stability was assessed with the Jaccard similarity index. We have observed 4 clusters with a very high stability (ranging between 92 and 100%). Based on distinctive features that deviate from the overall population distribution, we have labeled Cluster 1 (N = 729, 53.76%) as "Healthy", Cluster 2 (N = 508, 37.46%) as "Family history-Overweight-High Cholesterol ", Cluster 3 (N = 91, 6.71%) as "Severe Obesity-Prediabetes-Inflammation" and Cluster 4 (N = 28, 2.06%) as "Diabetes-Hypertension-Poor CV Health". Our work provides an in-depth characterization and thus, a better understanding of cardiometabolic health in the general population. Our data suggest that such a clustering approach could now be used to define more targeted and tailored strategies for the prevention of cardiometabolic diseases at a population level. This study provides a first step towards precision cardiometabolic prevention and should be externally validated in other contexts.

Semi-supervised clustering of time-dependent categorical sequences with application to discovering education-based life patterns

A new approach to the analysis of heterogeneous categorical sequences is proposed. The first-order Markov model is employed in a finite mixture setting with initial state and transition probabilities being expressed as functions of time. The expectation–maximization algorithm approach to parameter estimation is implemented in the presence of positive equivalence constraints that determine which observations must be placed in the same class in the solution. The proposed model is applied to a dataset from the British Household Panel Survey to evaluate the association between the education background and life outcomes of study participants. The analysis of the survey data reveals many interesting relationships between the level of education and major life events.

Integrating Prior Knowledge in Mixed-Initiative Social Network Clustering

We propose a new approach-called PK-clustering-to help social scientists create meaningful clusters in social networks. Many clustering algorithms exist but most social scientists find them difficult to understand, and tools do not provide any guidance to choose algorithms, or to evaluate results taking into account the prior knowledge of the scientists. Our work introduces a new clustering approach and a visual analytics user interface that address this issue. It is based on a process that 1) captures the prior knowledge of the scientists as a set of incomplete clusters, 2) runs multiple clustering algorithms (similarly to clustering ensemble methods), 3) visualizes the results of all the algorithms ranked and summarized by how well each algorithm matches the prior knowledge, 4) evaluates the consensus between user-selected algorithms and 5) allows users to review details and iteratively update the acquired knowledge. We describe our approach using an initial functional prototype, then provide two examples of use and early feedback from social scientists. We believe our clustering approach offers a novel constructive method to iteratively build knowledge while avoiding being overly influenced by the results of often randomly selected black-box clustering algorithms.

Brain Asymmetry Detection and Machine Learning Classification for Diagnosis of Early Dementia

Early identification of degenerative processes in the human brain is considered essential for providing proper care and treatment. This may involve detecting structural and functional cerebral changes such as changes in the degree of asymmetry between the left and right hemispheres. Changes can be detected by computational algorithms and used for the early diagnosis of dementia and its stages (amnestic early mild cognitive impairment (EMCI), Alzheimer’s Disease (AD)), and can help to monitor the progress of the disease. In this vein, the paper proposes a data processing pipeline that can be implemented on commodity hardware. It uses features of brain asymmetries, extracted from MRI of the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database, for the analysis of structural changes, and machine learning classification of the pathology. The experiments provide promising results, distinguishing between subjects with normal cognition (NC) and patients with early or progressive dementia. Supervised machine learning algorithms and convolutional neural networks tested are reaching an accuracy of 92.5% and 75.0% for NC vs. EMCI, and 93.0% and 90.5% for NC vs. AD, respectively. The proposed pipeline offers a promising low-cost alternative for the classification of dementia and can be potentially useful to other brain degenerative disorders that are accompanied by changes in the brain asymmetries.

Association of Chronic Condition Special Needs Plan With Hospitalization and Mortality Among Patients With End-Stage Kidney Disease

IMPORTANCE

While several studies have demonstrated the benefit of enrollment in chronic condition special needs plans (C-SNPs) for other chronic diseases (eg, diabetes), there is no evaluation of the association of C-SNPs with outcomes among patients with end-stage kidney disease (ESKD).

OBJECTIVE

To examine whether and to what degree C-SNP enrollment was associated with improved clinical outcomes and quality of life in patients with ESKD.

DESIGN, SETTING, AND PARTICIPANTS

This multicenter cohort study included 2718 patients who were newly enrolled in an ESKD C-SNP between January 1, 2013, and September 30, 2017, and receiving dialysis from DaVita Kidney Care. Patients were followed up until death, loss to follow-up, or end of study (ie, December 31, 2018). Enrollees in C-SNP were matched via multiple clinical and demographic characteristics with 2 different control populations, as follows: (1) those in the same facilities (n = 2545) or (2) those in similar counties (n = 1986). Patients enrolled in CareMore C-SNPs (n = 206) were excluded from the study. Data analysis was conducted June to December 2019.

EXPOSURES

Standard ESKD care with dialysis plus access to an integrated care team who worked with the patient and the dialysis team, comprehensive health assessments done by the integrated care team, and access to select benefits (such as vision and dental care) as a C-SNP enrollee.

MAIN OUTCOMES AND MEASURES

Hospitalizations, mortality, laboratory values indicative of metabolic control, and Kidney Disease Quality of Life 36-item (KDQOL-36) survey scores.

RESULTS

The 2545 C-SNP enrollees in the facility-matched analysis had a mean (SD) age of 57.2 (12.9) years, and included 968 (38.0%) women, 1328 (52.2%) Hispanic individuals, and 553 (21.7%) African American individuals. The 1986 C-SNP enrollees in the county-matched analysis had a mean (SD) age of 57.8 (12.2) years, with 705 (35.5%) women, 1085 (54.6%) Hispanic individuals, and 472 (23.8%) African American individuals. Compared with patients not enrolled in C-SNP, enrollees had lower hospitalization rates, with incidence rate ratios of 0.90 (95% CI, 0.84-0.97; P = .006) in the facility-matched analysis and 0.76 (95% CI, 0.70-0.83; P < .001) in the county-matched analysis. Compared with patients not enrolled in C-SNP, enrollees had decreased mortality risk in the same facilities (hazard ratio, 0.77; 95% CI, 0.68-0.88; P < .001) and in the same counties (hazard ratio, 0.77; 95% CI, 0.66-0.88; P < .001). No significant differences were observed between C-SNP enrollees and matched patients in metabolic laboratory values or KDQOL-36 survey scores.

CONCLUSIONS AND RELEVANCE

This cohort study found a positive association of C-SNP enrollment with lower rates of hospitalization and mortality. The findings suggest that the additional services and benefits C-SNPs provide may improve outcomes compared with standard of care for patients with ESKD.

Challenges of Clustering Multimodal Clinical Data: Review of Applications in Asthma Subtyping

Background

In the current era of personalized medicine, there is increasing interest in understanding the heterogeneity in disease populations. Cluster analysis is a method commonly used to identify subtypes in heterogeneous disease populations. The clinical data used in such applications are typically multimodal, which can make the application of traditional cluster analysis methods challenging.

Objective

This study aimed to review the research literature on the application of clustering multimodal clinical data to identify asthma subtypes. We assessed common problems and shortcomings in the application of cluster analysis methods in determining asthma subtypes, such that they can be brought to the attention of the research community and avoided in future studies.

Methods

We searched PubMed and Scopus bibliographic databases with terms related to cluster analysis and asthma to identify studies that applied dissimilarity-based cluster analysis methods. We recorded the analytic methods used in each study at each step of the cluster analysis process.

Results

Our literature search identified 63 studies that applied cluster analysis to multimodal clinical data to identify asthma subtypes. The features fed into the cluster algorithms were of a mixed type in 47 (75%) studies and continuous in 12 (19%), and the feature type was unclear in the remaining 4 (6%) studies. A total of 23 (37%) studies used hierarchical clustering with Ward linkage, and 22 (35%) studies used k-means clustering. Of these 45 studies, 39 had mixed-type features, but only 5 specified dissimilarity measures that could handle mixed-type features. A further 9 (14%) studies used a preclustering step to create small clusters to feed on a hierarchical method. The original sample sizes in these 9 studies ranged from 84 to 349. The remaining studies used hierarchical clustering with other linkages (n=3), medoid-based methods (n=3), spectral clustering (n=1), and multiple kernel k-means clustering (n=1), and in 1 study, the methods were unclear. Of 63 studies, 54 (86%) explained the methods used to determine the number of clusters, 24 (38%) studies tested the quality of their cluster solution, and 11 (17%) studies tested the stability of their solution. Reporting of the cluster analysis was generally poor in terms of the methods employed and their justification.

Conclusions

This review highlights common issues in the application of cluster analysis to multimodal clinical data to identify asthma subtypes. Some of these issues were related to the multimodal nature of the data, but many were more general issues in the application of cluster analysis. Although cluster analysis may be a useful tool for investigating disease subtypes, we recommend that future studies carefully consider the implications of clustering multimodal data, the cluster analysis process itself, and the reporting of methods to facilitate replication and interpretation of findings.

Characterizing the weight-glycemia phenotypes of type 1 diabetes in youth and young adulthood

INTRODUCTION

Individuals with type 1 diabetes (T1D) present with diverse body weight status and degrees of glycemic control, which may warrant different treatment approaches. We sought to identify subgroups sharing phenotypes based on both weight and glycemia and compare characteristics across subgroups.

RESEARCH DESIGN AND METHODS

Participants with T1D in the SEARCH study cohort (n=1817, 6.0-30.4 years) were seen at a follow-up visit >5 years after diagnosis. Hierarchical agglomerative clustering was used to group participants based on five measures summarizing the joint distribution of body mass index z-score (BMIz) and hemoglobin A1c (HbA1c) which were estimated by reinforcement learning tree predictions from 28 covariates. Interpretation of cluster weight status and glycemic control was based on mean BMIz and HbA1c, respectively.

RESULTS

The sample was 49.5% female and 55.5% non-Hispanic white (NHW); mean±SD age=17.6±4.5 years, T1D duration=7.8±1.9 years, BMIz=0.61±0.94, and HbA1c=76±21 mmol/mol (9.1±1.9)%. Six weight-glycemia clusters were identified, including four normal weight, one overweight, and one subgroup with obesity. No cluster had a mean HbA1c <58 mmol/mol (7.5%). Cluster 1 (34.0%) was normal weight with the lowest HbA1c and comprised 85% NHW participants with the highest socioeconomic position, insulin pump use, dietary quality, and physical activity. Subgroups with very poor glycemic control (ie, ≥108 mmol/mol (≥12.0%); cluster 4, 4.4%, and cluster 5, 7.5%) and obesity (cluster 6, 15.4%) had a lower proportion of NHW youth, lower socioeconomic position, and reported decreased pump use and poorer health behaviors (overall p<0.01). The overweight subgroup with very poor glycemic control (cluster 5) showed the highest lipids and blood pressure (p<0.01).

CONCLUSIONS

There are distinct subgroups of youth and young adults with T1D that share weight-glycemia phenotypes. Subgroups may benefit from tailored interventions addressing differences in clinical care, health behaviors, and underlying health inequity.

Data Stream Classification by Dynamic Incremental Semi-Supervised Fuzzy Clustering

A data stream classification method called DISSFCM (Dynamic Incremental Semi-Supervised FCM) is presented, which is based on an incremental semi-supervised fuzzy clustering algorithm. The method assumes that partially labeled data belonging to different classes are continuously available during time in form of chunks. Each chunk is processed by semi-supervised fuzzy clustering leading to a cluster-based classification model. The proposed DISSFCM is capable of dynamically adapting the number of clusters to data streams, by splitting low-quality clusters so as to improve classification quality. Experimental results on both synthetic and real-world data show the effectiveness of the proposed method in data stream classification.

A survey on semi-supervised learning

Semi-supervised learning is the branch of machine learning concerned with using labelled as well as unlabelled data to perform certain learning tasks. Conceptually situated between supervised and unsupervised learning, it permits harnessing the large amounts of unlabelled data available in many use cases in combination with typically smaller sets of labelled data. In recent years, research in this area has followed the general trends observed in machine learning, with much attention directed at neural network-based models and generative learning. The literature on the topic has also expanded in volume and scope, now encompassing a broad spectrum of theory, algorithms and applications. However, no recent surveys exist to collect and organize this knowledge, impeding the ability of researchers and engineers alike to utilize it. Filling this void, we present an up-to-date overview of semi-supervised learning methods, covering earlier work as well as more recent advances. We focus primarily on semi-supervised classification, where the large majority of semi-supervised learning research takes place. Our survey aims to provide researchers and practitioners new to the field as well as more advanced readers with a solid understanding of the main approaches and algorithms developed over the past two decades, with an emphasis on the most prominent and currently relevant work. Furthermore, we propose a new taxonomy of semi-supervised classification algorithms, which sheds light on the different conceptual and methodological approaches for incorporating unlabelled data into the training process. Lastly, we show how the fundamental assumptions underlying most semi-supervised learning algorithms are closely connected to each other, and how they relate to the well-known semi-supervised clustering assumption.

Location Privacy Protection in Distributed IoT Environments Based on Dynamic Sensor Node Clustering

One of the most significant challenges in Internet of Things (IoT) environments is the protection of privacy. Failing to guarantee the privacy of sensitive data collected and shared over IoT infrastructures is a critical barrier that delays the wide penetration of IoT technologies in several user-centric application domains. Location information is the most common dynamic information monitored and lies among the most sensitive ones from a privacy perspective. This article introduces a novel mechanism that aims to protect the privacy of location information across Data Centric Sensor Networks (DCSNs) that monitor the location of mobile objects in IoT systems. The respective data dissemination protocols proposed enhance the security of DCSNs rendering them less vulnerable to intruders interested in obtaining the location information monitored. In this respect, a dynamic clustering algorithm is that clusters the DCSN nodes not only based on the network topology, but also considering the current location of the objects monitored. The proposed techniques do not focus on the prevention of attacks, but on enhancing the privacy of sensitive location information once IoT nodes have been compromised. They have been extensively assessed via series of experiments conducted over the IoT infrastructure of FIT IoT-LAB and the respective evaluation results indicate that the dynamic clustering algorithm proposed significantly outperforms existing solutions focusing on enhancing the privacy of location information in IoT.

Cross-Cultural Validation of Mood Profile Clusters in a Sport and Exercise Context

Mood profiling has a long history in the field of sport and exercise. Several novel mood profile clusters were identified and described in the literature recently (Parsons-Smith et al., 2017). In the present study, we investigated whether the same clusters were evident in an Italian-language, sport and exercise context. The Italian Mood Scale (ITAMS; Quartiroli et al., 2017) was administered to 950 Italian-speaking sport participants (659 females, 284 males, 7 unspecified; age range = 16-63 year, M = 25.03, SD = 7.62) and seeded k-means clustering methodology applied to the responses. Six distinct mood profiles were identified, termed the iceberg, inverse iceberg, inverse Everest, shark fin, surface, and submerged profiles, which closely resembled those reported among English-speaking participants (Parsons-Smith et al., 2017). Significant differences were found in the distribution of specific mood profiles across gender and age groups. Findings supported the cross-cultural generalizability of the six mood profiles and offer new research avenues into their antecedents, correlates and behavioral consequences in Italian-language contexts.

No wisdom in the crowd: genome annotation in the era of big data - current status and future prospects

Science and engineering rely on the accumulation and dissemination of knowledge to make discoveries and create new designs. Discovery-driven genome research rests on knowledge passed on via gene annotations. In response to the deluge of sequencing big data, standard annotation practice employs automated procedures that rely on majority rules. We argue this hinders progress through the generation and propagation of errors, leading investigators into blind alleys. More subtly, this inductive process discourages the discovery of novelty, which remains essential in biological research and reflects the nature of biology itself. Annotation systems, rather than being repositories of facts, should be tools that support multiple modes of inference. By combining deduction, induction and abduction, investigators can generate hypotheses when accurate knowledge is extracted from model databases. A key stance is to depart from 'the sequence tells the structure tells the function' fallacy, placing function first. We illustrate our approach with examples of critical or unexpected pathways, using MicroScope to demonstrate how tools can be implemented following the principles we advocate. We end with a challenge to the reader.

Development of Structural Covariance From Childhood to Adolescence: A Longitudinal Study in 22q11.2DS

Background: Schizophrenia is currently considered a neurodevelopmental disorder of connectivity. Still few studies have investigated how brain networks develop in children and adolescents who are at risk for developing psychosis. 22q11.2 Deletion Syndrome (22q11DS) offers a unique opportunity to investigate the pathogenesis of schizophrenia from a neurodevelopmental perspective. Structural covariance (SC) is a powerful approach to explore morphometric relations between brain regions that can furthermore detect biomarkers of psychosis, both in 22q11DS and in the general population.

Methods: Here we implement a state-of-the-art sliding-window approach to characterize maturation of SC network architecture in a large longitudinal cohort of patients with 22q11DS (110 with 221 visits) and healthy controls (117 with 211 visits). We furthermore propose a new clustering-based approach to group regions according to trajectories of structural connectivity maturation. We correlate measures of SC with development of working memory, a core executive function that is highly affected in both idiopathic psychosis and 22q11DS. Finally, in 22q11DS we explore correlations between SC dysconnectivity and severity of internalizing psychopathology.

Results: In HCs network architecture underwent a quadratic developmental trajectory maturing up to mid-adolescence. Late-childhood maturation was particularly evident for fronto-parietal cortices, while Default-Mode-Network-related regions showed a more protracted linear development. Working memory performance was positively correlated with network segregation and fronto-parietal connectivity. In 22q11DS, we demonstrate aberrant maturation of SC with disturbed architecture selectively emerging during adolescence and correlating more severe internalizing psychopathology. Patients also presented a lack of typical network development during late-childhood, that was particularly prominent for frontal connectivity.

Conclusions: Our results suggest that SC maturation may underlie critical cognitive development occurring during late-childhood in healthy controls. Aberrant trajectories of SC maturation may reflect core developmental features of 22q11DS, including disturbed cognitive maturation during childhood and predisposition to internalizing psychopathology and psychosis during adolescence.

Identification of relevant subtypes via preweighted sparse clustering

Cluster analysis methods are used to identify homogeneous subgroups in a data set. In biomedical applications, one frequently applies cluster analysis in order to identify biologically interesting subgroups. In particular, one may wish to identify subgroups that are associated with a particular outcome of interest. Conventional clustering methods generally do not identify such subgroups, particularly when there are a large number of high-variance features in the data set. Conventional methods may identify clusters associated with these high-variance features when one wishes to obtain secondary clusters that are more interesting biologically or more strongly associated with a particular outcome of interest. A modification of sparse clustering can be used to identify such secondary clusters or clusters associated with an outcome of interest. This method correctly identifies such clusters of interest in several simulation scenarios. The method is also applied to a large prospective cohort study of temporomandibular disorders and a leukemia microarray data set.

Structured Sparse Subspace Clustering: A Joint Affinity Learning and Subspace Clustering Framework

Subspace clustering refers to the problem of segmenting data drawn from a union of subspaces. State-of-the-art approaches for solving this problem follow a two-stage approach. In the first step, an affinity matrix is learned from the data using sparse or low-rank minimization techniques. In the second step, the segmentation is found by applying spectral clustering to this affinity. While this approach has led to the state-of-the-art results in many applications, it is suboptimal, because it does not exploit the fact that the affinity and the segmentation depend on each other. In this paper, we propose a joint optimization framework - Structured Sparse Subspace Clustering (S³C) - for learning both the affinity and the segmentation. The proposed S³C framework is based on expressing each data point as a structured sparse linear combination of all other data points, where the structure is induced by a norm that depends on the unknown segmentation. Moreover, we extend the proposed S³C framework into Constrained S³C (CS³C) in which available partial side-information is incorporated into the stage of learning the affinity. We show that both the structured sparse representation and the segmentation can be found via a combination of an alternating direction method of multipliers with spectral clustering. Experiments on a synthetic data set, the Extended Yale B face data set, the Hopkins 155 motion segmentation database, and three cancer data sets demonstrate the effectiveness of our approach.