Hierarchical Dirichlet Process Research Articles

A Dirichlet Process Model for Directional-linear Data with Application to Bloodstain Pattern Analysis

Directional data require specialized models because of the non-Euclidean nature of their domain. When a directional variable is observed jointly with linear variables, modeling their dependence adds an additional layer of complexity. A Bayesian nonparametric approach is introduced to analyze directional-linear data. Firstly, the projected normal distribution is extended to model the joint distribution of linear variables and a directional variable with arbitrary dimension projected from a higher-dimensional augmented multivariate normal distribution. The new distribution is called the semi-projected normal distribution (SPN) and can be used as the mixture distribution in a Dirichlet process model to obtain a more flexible class of models for directional-linear data. Then, a conditional inverse-Wishart distribution is proposed as part of the prior distribution to address an identifiability issue inherited from the projected normal and preserve conjugacy with the SPN. The SPN mixture model shows superior performance in clustering on synthetic data compared to the semi-wrapped Gaussian model. The experiments show the ability of the SPN mixture model to characterize bloodstain patterns. A hierarchical Dirichlet process model with the SPN distribution is built to estimate the likelihood of bloodstain patterns under a posited causal mechanism for use in a likelihood ratio approach to the analysis of forensic bloodstain pattern evidence.

Blocked Gibbs Sampler for Hierarchical Dirichlet Processes

Posterior computation in hierarchical Dirichlet process (HDP) mixture models is an active area of research in nonparametric Bayes inference of grouped data. Existing literature almost exclusively focuses on the Chinese restaurant franchise (CRF) analogy of the marginal distribution of the parameters, which can mix poorly and has a quadratic complexity with the sample size. A recently developed slice sampler allows for efficient blocked updates of the parameters, but is shown to be statistically unstable in our article. We develop a blocked Gibbs sampler that employs a truncated approximation of the underlying random measures to sample from the posterior distribution of HDP, which produces statistically stable results, is highly scalable with respect to sample size, and is shown to have good mixing. The heart of the construction is to endow the shared concentration parameter with an appropriately chosen gamma prior that allows us to break the dependence of the shared mixing proportions and permits independent updates of certain log-concave random variables in a block. En route, we develop an efficient rejection sampler for these random variables leveraging piece-wise tangent-line approximations. Supplementary materials, which include substantive additional details and code, are available online.

A Genome-Wide Association Study Suggests New Susceptibility Loci for Primary Antiphospholipid Syndrome.

Primary antiphospholipid syndrome (PAPS) is a rare autoimmune disease characterized by the presence of antiphospholipid antibodies and the occurrence of thrombotic events and pregnancy complications. Our study aimed to identify novel genetic susceptibility loci associated with PAPS. We performed a genome-wide association study comprising 5,485 individuals (482 affected individuals) of European ancestry. Significant and suggestive independent variants from a meta-analysis of approximately 7 million variants were evaluated for functional and biological process enrichment. The genetic risk variability for PAPS in different populations was also assessed. Hierarchical clustering, Mahalanobis distance, and Dirichlet Process Mixtures with uncertainty clustering methods were used to assess genetic similarities between PAPS and other immune-mediated diseases. We revealed genetic associations with PAPS in a regulatory locus within the HLA class II region near HLA-DRA and in STAT1-STAT4 with a genome-wide level of significance; 34 additional suggestive genetic susceptibility loci for PAPS were also identified. The disease risk allele near HLA-DRA is associated with overexpression of HLA-DRB6, HLA-DRB9, HLA-DQA2, and HLA-DQB2 in immune cells, vascular tissue, and nervous tissue. This association is independent of the association between PAPS and HLA-DRB1*1302. Functional analyses highlighted immune-related pathways in PAPS-associated loci. The comparison with other immune-mediated diseases revealed a close genetic relatedness to neuromyelitis optica, systemic sclerosis, and Sjögren syndrome, suggesting co-localized causal variations close to STAT1-STAT4, TNPO3, and BLK. This study represents a comprehensive large-scale genetic analysis for PAPS and provides new insights into the genetic basis and pathophysiology of this rare disease.

A Comparative Study of Topic Modelling Approaches for User-generated Point of Interest Data

This study aims to enhance urban planning and management by harnessing the power of machine learning (ML) andbig data. We focus on Urban Functional Zones (UFZs), the fundamental units for human socio-economic activities. Ourmethodology involves compiling Point of Interest (POI) data from various sources for comprehensive analysis. Weemploy various topic modeling approaches such as Latent Dirichlet Allocation (LDA), Latent Semantic Index (LSI),Hierarchical Dirichlet Process (HDP), and Top2Vec. Our principal results reveal significant differences in theperformance and coherence of these models on short text documents. Consequently, our major conclusion isidentifying the better-performing topic model for classifying UFZs from POI data. We also explore four textpreprocessing steps to optimize the performance of the topic models. This study contributes to the field by providinga nuanced understanding of UFZs, paving the way for future data-driven urban planning and management.

Bayesian hierarchical profile regression for binary covariates.

Dysphagia, a common result of other medical conditions, is caused by malfunctions in swallowing physiology resulting in difficulty eating and drinking. The Modified Barium Swallow Study (MBSS), the most commonly used diagnostic tool for evaluating dysphagia, can be assessed using the Modified Barium Swallow Impairment Profile (MBSImP™). The MBSImP assessment tool consists of a hierarchical grouped data structure with multiple domains, a set of components within each domain which characterize specific swallowing physiologies, and a set of tasks scored on a discrete scale within each component. We lack sophisticated approaches to extract patterns of physiologic swallowing impairment from the MBSImP task scores within a component while still recognizing the nested structure of components within a domain. We propose a Bayesian hierarchical profile regression model, which uses a Bayesian profile regression model in conjunction with a hierarchical Dirichlet process mixture model to (1) cluster subjects into impairment profile patterns while respecting the hierarchical grouped data structure of the MBSImP, and (2) simultaneously determine associations between latent profile cluster membership for all components and the outcome of dysphagia severity. We apply our approach to a cohort of patients referred for an MBSS and assessed using the MBSImP. Our research results can be used to inform appropriate intervention strategies, and provide tools for clinicians to make better multidimensional management and treatment decisions for patients with dysphagia.

MOSAIC: An Artificial Intelligence-Based Framework for Multimodal Analysis, Classification, and Personalized Prognostic Assessment in Rare Cancers.

Rare cancers constitute over 20% of human neoplasms, often affecting patients with unmet medical needs. The development of effective classification and prognostication systems is crucial to improve the decision-making process and drive innovative treatment strategies. We have created and implemented MOSAIC, an artificial intelligence (AI)-based framework designed for multimodal analysis, classification, and personalized prognostic assessment in rare cancers. Clinical validation was performed on myelodysplastic syndrome (MDS), a rare hematologic cancer with clinical and genomic heterogeneities. We analyzed 4,427 patients with MDS divided into training and validation cohorts. Deep learning methods were applied to integrate and impute clinical/genomic features. Clustering was performed by combining Uniform Manifold Approximation and Projection for Dimension Reduction + Hierarchical Density-Based Spatial Clustering of Applications with Noise (UMAP + HDBSCAN) methods, compared with the conventional Hierarchical Dirichlet Process (HDP). Linear and AI-based nonlinear approaches were compared for survival prediction. Explainable AI (Shapley Additive Explanations approach [SHAP]) and federated learning were used to improve the interpretation and the performance of the clinical models, integrating them into distributed infrastructure. UMAP + HDBSCAN clustering obtained a more granular patient stratification, achieving a higher average silhouette coefficient (0.16) with respect to HDP (0.01) and higher balanced accuracy in cluster classification by Random Forest (92.7% ± 1.3% and 85.8% ± 0.8%). AI methods for survival prediction outperform conventional statistical techniques and the reference prognostic tool for MDS. Nonlinear Gradient Boosting Survival stands in the internal (Concordance-Index [C-Index], 0.77; SD, 0.01) and external validation (C-Index, 0.74; SD, 0.02). SHAP analysis revealed that similar features drove patients' subgroups and outcomes in both training and validation cohorts. Federated implementation improved the accuracy of developed models. MOSAIC provides an explainable and robust framework to optimize classification and prognostic assessment of rare cancers. AI-based approaches demonstrated superior accuracy in capturing genomic similarities and providing individual prognostic information compared with conventional statistical methods. Its federated implementation ensures broad clinical application, guaranteeing high performance and data protection.

Topic Modeling for Evolving Textual Data Using LDA, HDP, NMF, BERTOPIC, and DTM With a Focus on Research Papers

As the volume of academic literature continues to burgeon, the necessity for advanced tools to decipher evolving research trends becomes increasingly apparent. This study delves into the utilization of topic modeling techniques—specifically Latent Dirichlet Allocation (LDA), Hierarchical Dirichlet Process (HDP), Non-negative Matrix Factorization (NMF), BERTopic, and Dynamic Topic Modeling (DTM)—applied to a dynamic corpus of research papers. Our research endeavors to confront the challenges posed by capturing temporal dynamics, evolving terminology, and interdisciplinary themes within academic literature. Through a comprehensive comparative investigation of these models, we assess their efficacy in extracting and tracking research topics over time. While DTM exhibited the highest term topic probability, its inclusion of non-meaningful words proved to be a hindrance to its suitability. Conversely, NMF, HDP, LDA, and BERTopic demonstrated comparable performance in topic extraction. Surprisingly, DTM emerged as the most effective model in our research, showcasing its prowess in navigating the intricacies of evolving research trends.

A dynamic driving-style analysis method based on drivers’ interaction with surrounding vehicles

The ability to recognize different driving styles of surrounding vehicles is crucial to determine the safest and most efficient driving decisions, prevent accidents, and analyze the causes of traffic accidents. Understanding if the surrounding vehicle is aggressive or cautious can greatly assist in the decision making of vehicles in terms of whether and when it is appropriate to make particular maneuvers. A driver’s driving style usually changes with the environment, which brings great challenges to the current research. To this end, a dynamic driving-style analysis framework, in which drivers’ interactions with other vehicles are considered, is proposed in this article. First, by analyzing common traffic scenarios, five surrounding vehicles are selected as the environmental vehicles to be considered. Time headway (THW) and time to collision (TTC) that can consider the relative speed and position with the ego vehicle are selected as the clustering indicators. Then, a Bayesian nonparametric learning method based on a hierarchical Dirichlet-process hidden semi-Markov model (HDP-HSMM) is introduced to extract primitive driving patterns from time-series driving data without prior knowledge of the number of these patterns. Then the driving pattern is scored according to the risk degree. A driver’s aggressiveness is scored and drivers are divided into different styles based on the frequency distribution of driving patterns. The effectiveness of the proposed method is demonstrated on a real-world vehicle trajectory data set where results show that driving pattern switches and more complex driving behaviors can be better captured and understood semantically.

Towards Attack Detection in Multimodal Cyber-Physical Systems with Sticky HDP-HMM based Time Series Analysis

Automatic detection of the precise occurrence and duration of an attack reflected in time-series logs generated by cyber-physical systems is a challenging problem. This problem is exacerbated when performing this analysis using logs with limited system information. In a realistic scenario, multiple and differing attack methods may be employed in rapid succession. Modern or legacy systems operate in multiple modes and contain multiple devices recording a variety of continuous and categorical data streams. This work presents a non-parametric Bayesian framework that addresses these challenges using the sticky Hierarchical Dirichlet Process Hidden Markov Model (sHDP-HMM) . Additionally, we explore metrics for measuring the accuracy of the detected events, their timings and durations, and compare the computational efficiency of different inference implementations of the model. The efficacy of attack detection is demonstrated in two settings: an avionics testbed and a consumer robot.

Scoping Review with Topic Modeling on the Diagnostic Criteria for Degenerative Cervical Myelopathy.

This study is a scoping review. There is a broad variability in the definition of degenerative cervical myelopathy (DCM) and no standardized set of diagnostic criteria to date. We interrogated the Myelopathy.org database, a hand-indexed database of primary clinical studies conducted exclusively on DCM in humans between 2005-2021. The DCM inclusion criteria used in these studies were inputted into 3 topic modeling algorithms: Hierarchical Dirichlet Process (HDP), Latent Dirichlet Allocation (LDA), and BERtopic. The emerging topics were subjected to manual labeling and interpretation. Of 1676 reports, 120 papers (7.16%) had well-defined inclusion criteria and were subjected to topic modeling. Four topics emerged from the HDP model: disturbance from extremity weakness and motor signs; fine-motor and sensory disturbance of upper extremity; a combination of imaging and clinical findings is required for the diagnosis; and "reinforcing" (or modifying) factors that can aid in the diagnosis in borderline cases. The LDA model showed the following topics: disturbance to the patient is required for the diagnosis; reinforcing factors can aid in the diagnosis in borderline cases; clinical findings from the extremities; and a combination of imaging and clinical findings is required for the diagnosis. BERTopic identified the following topics: imaging abnormality, typical clinical features, range of objective criteria, and presence of clinical findings. This review provides quantifiable data that only a minority of past studies in DCM provided meaningful inclusion criteria. The items and patterns found here are very useful for the development of diagnostic criteria for DCM.

An experimental study for identifying customer prominent viewpoints on different flight classes by topic modeling methods

The airline industry has a wide range of customers with different tastes and expectations of airline services. Therefore, knowing the different attitudes of customers and consequently addressing their opinions is paramount in this business. Airlines offer different flight classes to customers to provide passenger transportation services and use different methods to evaluate customer satisfaction. One of the methods used to analyze customer opinions is the use of text analysis with machine learning techniques. In this study, to analyze the textual data, we used topic modeling to identify passengers' attitudes. In the first stage, we evaluated the performance of three different topic modeling methods: Latent Semantic Indexing (LSI), Hierarchical Dirichlet Process (HDP), and Latent Dirichlet Allocation (LDA). As a result of the modeling process, LDA had the best performance with an average coherence value of 0.42. Eventually, after determining the optimum number of topics as 10 topics, the prominent topics were interpreted and explained. The results of this study demonstrate the common and distinct attitudes of customers of economy class, business class, and first class. As appeared in the results of the economy flight class, the most important topic that appeared in comments was "passenger care". However, in whole reviews and reviews of business and first flight classes, the topic of "time" was more repetitive. Given the wide application of the presented methodology, the findings of this research can provide airline managers with a broader insight into passenger preferences and guide them on how to provide quality service for their customers.

Shared Differential Clustering across Single-cell RNA Sequencing Datasets with the Hierarchical Dirichlet Process

Single-cell RNA sequencing (scRNA-seq) is a powerful technology that allows researchers to understand gene expression patterns at the single-cell level and uncover the heterogeneous nature of cells. Clustering is an important tool in scRNA-seq analysis to discover groups of cells with similar gene expression patterns and identify potential cell types. Integration of multiple scRNA-seq datasets is a pressing challenge, and in this direction, a novel model is developed to extend clustering methods to appropriately combine inference across multiple datasets. The model simultaneously addresses normalization to deal with the inherent noise and uncertainty in scRNA-seq, infers cell types, and integrates multiple datasets for shared clustering in principled manner through a hierarchical Bayesian framework. A Gibbs sampler is developed that copes with the high-dimensionality of scRNA-seq through consensus clustering. The methodological developments are driven by experimental data from embryonic cells, with the aim of understanding the role of PAX6 in prenatal development, and more specifically how cell-subtypes and their proportions change when knocking out this factor.

TEXT MINING: CLUSTERING USING BERT AND PROBABILISTIC TOPIC MODELING

In order to find significant patterns and fresh ideas, free-form content is transformed into structured format using a process known as text mining or text data mining. It enables businesses to easily locate important information in texts like emails, social media posts, support requests, chatbots, and other sorts of text. Text mining enables businesses to anticipate possible threats from rivals, react quickly to production or delivery problems, and provide more individualised customer service. Businesses employ text mining for a range of functions, including production, IT, marketing, sales, and customer service. By carefully examining the phrases used in the source texts, topic modelling aims to pinpoint the recurrent themes in a corpus. These concepts are known as “topics”. As a result, textual data may be measured and used in quantitative analysis. In this sector, there are several subject modelling kinds that differ from one another based on a few unique traits and criteria. In our paper we have represented mainly 3 types of topic modelling techniques namely Latent Semantic Analysis (LSA), Hierarchical Dirichlet Process (HDP), and Latent Dirichlet Analysis (LDA) and calculated the coherence score of each method and compared them. And we have infused the concept of BERT with this topic modelling models and proposed a new model called HDP BERT and calculated the coherence Score and clusters the topics. At the end the n-grams features are applied to all 4 models and compared among each other in bases of uni, bi and trigram rate percentage.

RHDP: An Aspect Sharing-Enhanced Hierarchical Topic Model for Multi-Domain Corpus

Learning topic hierarchies from a multi-domain corpus is crucial in topic modeling as it reveals valuable structural information embedded within documents. Despite the extensive literature on hierarchical topic models, effectively discovering inter-topic correlations and differences among subtopics at the same level in the topic hierarchy, obtained from multiple domains, remains an unresolved challenge. This article proposes an enhanced nested Chinese restaurant process (nCRP), nCRP+, by introducing an additional mechanism based on Chinese restaurant franchise (CRF) for aspect-sharing pattern extraction in the original nCRP. Subsequently, by employing the distribution extracted from nCRP+ as the prior distribution for topic hierarchy in the hierarchical Dirichlet processes (HDP), we develop a hierarchical topic model for multi-domain corpus, named rHDP. We describe the model with the analogy of Chinese restaurant franchise based on the central kitchen and propose a hierarchical Gibbs sampling scheme to infer the model. Our method effectively constructs well-established topic hierarchies, accurately reflecting diverse parent-child topic relationships, explicit topic aspect sharing correlations for inter-topics, and differences between these shared topics. To validate the efficacy of our approach, we conduct experiments using a renowned public dataset and an online collection of Chinese financial documents. The experimental results confirm the superiority of our method over the state-of-the-art techniques in identifying multi-domain topic hierarchies, according to multiple evaluation metrics.

Multiple maneuvering target tracking based on hierarchical Dirichlet process and hidden Markov model

This paper is concerned with multiple maneuvering target tracking problems. Although the interacting multiple model (IMM) algorithm can be used for maneuvering targets tracking, the performance is limited as the motion mode set of the target is fixed in advance. In this paper, we address the problem of multiple maneuvering targets with unknown maneuvering models in cluttered environments. A finite hidden Markov model (HMM) is developed with truncated Hierarchical Dirichlet process (HDP) to recognize the unknown and changing motion modes. Further, the HDP-HMM is introduced into multiple hypothesis tracking (MHT) and probabilistic multiple hypothesis tracking (PMHT) approaches for multiple maneuvering targets tracking with data association uncertainty. To joint estimate the HDP-HMM parameters and target states, the particle learning and variational inference approaches are used, and the solutions of HDP-HMM-MHT and HDP-HMM-PMHT algorithms are obtained based on Rao-Blackwellized particle filter and variational Bayesian methods, respectively. As the changes of maneuver models can be learned using HDP, the multi-target tracking accuracy of the proposed algorithm is improved. Simulation results show that the proposed algorithms outperform the traditional IMM-MHT method with fixed model sets.

Data-Driven Harmonization of 2022 Who and ICC Classifications of Myelodysplastic Syndromes/Neoplasms (MDS): A Study By the International Consortium for MDS (icMDS)

Background. The inclusion of gene mutations and chromosomal abnormalities in the 2022 WHO and ICC Classifications of MDS has enhanced diagnostic precision and is expected to improve clinical decision-making process. Although these two systems share similarities, clinically relevant discrepancies still exist and potentially cause inconsistency in their adoption in a clinical setting. In this study on behalf of the International Consortium for MDS (icMDS), we adopted a data-driven approach to provide a harmonization roadmap between the 2022 WHO and ICC classification for MDS. A modified Delphi Process consensus approach is currently ongoing among icMDS experts to finalize a harmonized MDS classification scheme. Methods. We analyzed retrospective international cohorts of patients with a diagnosis of MDS (n=7017) and AML (n=1002) according to WHO 2016 criteria. Hierarchical Dirichlet Processes were applied to define clusters capturing broad dependencies among all gene mutations and cytogenetic abnormalities. To investigate the features of importance and their impact on the clustering process, we employed the SHapley Additive exPlanations approach (SHAP). This allowed to define harmonized labels for each clinical entity. The clinical relevance of the unsupervised clustering was assessed through the analysis of phenotypic features and outcomes among each group. ( Blood 2022;140: 9828-9830) Results. Patients' characteristics are summarized in Table 1. We identified 9 clusters, defined by specific genomic features. The cluster of highest hierarchical importance was characterized by biallelic inactivation of TP53 (biTP53). According to SHAP analysis, bi TP53 was defined as 2 or more TP53 mutations, or 1 mutation with copy number loss or cnLOH. Most patients assigned to bi TP53 cluster had TP53 VAF>10% (77.9%) and complex karyotype (70.1%). Assignment to bi TP53 cluster was irrespective of blast count. Patients with monoallelic TP53 mutation segregated into other clusters. Hierarchically, the second cluster included patients with del(5q). SHAP analysis highlighted 5q deletion alone, or with one other chromosomal abnormality other than -7, and absence of bi TP53, as the most relevant features. Most of these patients had blast counts <5% (88.1%). The third distinct cluster included patients with SF3B1 mutations (in the absence of concurrent del(7q), abn3q26.2, complex karyotype or RUNX1 mutation). Most patients with MDS and SF3B1 mutation had <5% blasts (94.2%). Common co-mutated variants in the SF3B1 cluster included mutant DNMT3A (25.2%) and TET2 (38.3%). Morphologically defined MDS cases (i.e., not meeting criteria for bi TP53, del(5q) or SF3B1) were preferentially assigned to the following additional clusters: SF3B1 and concurrent higher-risk mutations (e.g., RUNX1 and ASXL1); SRSF2 and concomitant TET2 mutations; U2AF1 mutations with del(20q), del(7q) or -7; SRSF2 with TET2 mutations and co-mutational patterns including RUNX1 and ASXL1; and AML-like genomic signatures. Our analyses suggest that morphologically defined MDS is characterized by a large heterogeneity in terms of mutation profiles, not entirely captured by the presence of unilineage versus multilineage dysplasia, percentage of bone marrow blasts, and presence of hypocellularity and fibrosis. To better investigate the continuum between high risk MDS (i.e., patients with ≥10% blasts) and AML, an exploratory comparison was made using a cohort of AML (defined according to WHO 2016) patients analyzed using the same statistical methods. Only a partial overlap in genetic signatures was observed between MDS with ≥10% blasts and AML. However, similarities were observed between the AML-like MDS clusters (characterized by mutant NPM1, bZIP CEBPA, and Core Binding Factor abnormalities) and AML clusters defined by the same genetic signature, thus supporting the classification of these entities as AML, irrespective of blast count. Conclusion. Our study demonstrated the utility of a data-driven approach based on advanced statistical methods to generate a harmonized classification for MDS. Table 2 shows a provisional, hierarchical classification algorithm. Further refinement of entity labels and classification criteria is the subject of the ongoing modified Delphi Process consensus approach among icMDS experts.

Molecular Classification of Chronic Myelomonocytic Leukemia: Results of the Analysis of an International Cohort of 2,471 Patients

Background. Chronic myelomonocytic leukemia (CMML) is a rare myeloid neoplasm that shares both dysplastic and proliferative features and includes patients with highly heterogeneous clinical manifestations and prognosis. Despite known co-mutation patterns that are suggesting of a CMML diagnosis such as TET2 and SRSF2 mutations, the current diagnosis and classification schemes rely solely on morphological criteria that fail to capture the genomic heterogeneity of the disease. Here we perform an unsupervised analysis using only genomic features to generate a novel classification schema of CMML patients with unique clinical features. Methods. A cohort of 2,471 CMML patients defined according to WHO 2016 criteria from European and US centers was analyzed. Bayesian network analysis and hierarchical Dirichlet processes were used to identify genomic associations and subgroups as a basis to define a molecular classification of CMML. Bayesian networks allow to infer the structure of conditional dependencies among mutations. Dirichlet processes were applied to define clusters capturing broad dependencies among all gene mutations and cytogenetic abnormalities. Multivariate logistic regression analysis was applied to compare clinical and hematological characteristics among different groups. ( JCO 2021 39:11, 1223-1233) Results. Baseline clinical and molecular characteristics are summarized in Table 1. We identified 11 clusters defined according to specific genomic features (Table 2). Two clusters were primarily defined by the presence of mutations in splicing genes SRSF2 and ZRSR2, frequently associated with TET2 mutations. Patients assigned to these clusters showed myelodysplastic features according to French American British (FAB) criteria, had a low rate of peripheral blood cytopenias, and low blast counts. Additionally, two clusters were characterized by the same mutations in splicing genes (i.e., SRSF2 or ZRSR2 with associated TET2), together with co-mutations in genes that correlate with a more aggressive phenotype ( RUNX1, ASXL1, EZH2). These groups showed overlapping myelodysplastic and myeloproliferative features by FAB, intermediate-risk per prognostic scores, and decreased survival as compared with the first 2 clusters. A distinct cluster was characterized by SF3B1 gene mutations (frequently associated with TET2 or DNMT3A). The SF3B1 cluster was marked by a dysplastic phenotype and more severe anemia (with a high rate of transfusion-dependence) with respect to the other genomic groups. Four clusters (27.8% of the overall population) were characterized by the presence of mutations in signal transduction and tyrosine kinase pathways genes ( NRAS/KRAS, CBL, JAK2 and SETBP1 often associated with ASXL1/TET2, respectively). Patients within these groups showed a mostly proliferative phenotype and intermediate survival. The presence of JAK2 mutations appeared to confer a better prognosis compared with other proliferative clusters. A minority of patients (4.6%) were assigned to two high-risk clusters, defined by TP53 mutations (often biallelic) with complex karyotype, and AML-like mutations ( NPM1, IDH1/2, FLT3), respectively. Patients in these clusters were predominantly dysplastic by FAB, had a high prevalence of severe cytopenias, higher blast counts, and a dismal outcome. Finally, we identified a cluster of patients defined by the absence of gene mutations and chromosomal abnormalities (4.5%). The majority of patients assigned to this cluster were characterized by a dysplastic phenotype, had low absolute monocyte counts, and favorable outcome. A subset of patients (18.9%) could not be assigned to a specified cluster in this preliminary analysis. An independent validation of this molecular classification is currently being conducted on a US prospective cohort (n=650). Conclusion. Our results provide proof of concept for a molecular classification of CMML based on the identification of homogeneous genomic signatures and associated clinical features. 15% of patients showed clear overlapping features with other myeloid neoplasms, thus providing the rational to refine the boundaries among different clinical entities. A molecular classification of CMML may provide a basis to define a next-generation prognostic tool and improve clinical decision-making.

Bayesian prognosis analysis of human papillomavirus-associated head and neck cancer using hierarchical Dirichlet process mixture models

The incidence and prognosis of Head and Neck Cancer (HNC) depend heavily on patients’ Human PapillomaVirus (HPV) status. Prognosis analysis of HPV-associated HNC is of clinical importance because in-depth understanding of the survival distribution is valuable for designing more informed treatment strategies. In this article, we develop a novel Hierarchical Dirichlet Process Weibull Mixture Model (HDP-WMM) to study the prognosis of HNC patients given their HPV status. The HDP-WMM is capable of simultaneously characterizing the survival distributions of grouped data and capturing the dependence among different groups. Moreover, the HDP-WMM can identify clusters of patients based on their outcomes, providing additional information for exploring patient subtypes. Effective Markov chain Monte Carlo sampling algorithms are designed for model inference and function estimation. The clustering structure is identified by summarizing the posterior samples of the data random partition using the Bayesian cluster analysis tool. A simulation study is designed to validate the performance of the proposed inference methods. The practical utility of the proposed HDP-WMM is demonstrated by a case study on prognosis analysis of HPV-associated HNC. Our results show that the Bayesian HDP-WMM achieves satisfactory performance on estimating the survival functions and clustering patients based on their outcomes.

Web content topic modeling using LDA and HTML tags.

An immense volume of digital documents exists online and offline with content that can offer useful information and insights. Utilizing topic modeling enhances the analysis and understanding of digital documents. Topic modeling discovers latent semantic structures or topics within a set of digital textual documents. The Internet of Things, Blockchain, recommender system, and search engine optimization applications use topic modeling to handle data mining tasks, such as classification and clustering. The usefulness of topic models depends on the quality of resulting term patterns and topics with high quality. Topic coherence is the standard metric to measure the quality of topic models. Previous studies build topic models to generally work on conventional documents, and they are insufficient and underperform when applied to web content data due to differences in the structure of the conventional and HTML documents. Neglecting the unique structure of web content leads to missing otherwise coherent topics and, therefore, low topic quality. This study aims to propose an innovative topic model to learn coherence topics in web content data. We present the HTML Topic Model (HTM), a web content topic model that takes into consideration the HTML tags to understand the structure of web pages. We conducted two series of experiments to demonstrate the limitations of the existing topic models and examine the topic coherence of the HTM against the widely used Latent Dirichlet Allocation (LDA) model and its variants, namely the Correlated Topic Model, the Dirichlet Multinomial Regression, the Hierarchical Dirichlet Process, the Hierarchical Latent Dirichlet Allocation, the pseudo-document based Topic Model, and the Supervised Latent Dirichlet Allocation models. The first experiment demonstrates the limitations of the existing topic models when applied to web content data and, therefore, the essential need for a web content topic model. When applied to web data, the overall performance dropped an average of five times and, in some cases, up to approximately 20 times lower than when applied to conventional data. The second experiment then evaluates the effectiveness of the HTM model in discovering topics and term patterns of web content data. The HTM model achieved an overall 35% improvement in topic coherence compared to the LDA.

Modeling Contingent Decision Behavior: A Bayesian Nonparametric Preference-Learning Approach

We propose a preference-learning algorithm for uncovering Decision Makers’ (DMs’) contingent evaluation strategies in the context of multiple criteria sorting. We assume the preference information in the form of holistic assignment examples derived from the analysis of alternatives’ performance vectors and textual descriptions. We characterize the decision policies using a mixture of threshold-based, value-driven preference models and associated latent topics. The latter serve as the stimuli underlying the contingency in decision behavior. Such a probabilistic model is constructed by using a flexible and nonparametric Bayesian framework. The proposed method adopts a hierarchical Dirichlet process as the prior so that a group of DMs can share a countably infinite number of contingent models and topics. For all DMs, it automatically identifies the components representing their evaluation strategies adequately. The posterior is summarized by using the Hamiltonian Monte Carlo sampling method. We demonstrate the method’s practical usefulness in a real-world recruitment problem considered by a Chinese IT company. We also compare the approach with counterparts that use a single preference model, implement the parametric framework, or consider each DM’s preferences individually. The results indicate that our approach performs favorably in both interpreting DMs’ contingent decision behavior and recommending decisions on new alternatives. Furthermore, the approach’s performance and robustness are investigated through a computational experiment involving real-world data sets. History: Accepted by Ram Ramesh, Area Editor for Data Science and Machine Learning. Funding: J. Liu received financial support from the National Natural Science Foundation of China [Grants 72071155 and 71701160]. M. Kadziński received financial support from the Polish National Science Center under the SONATA BIS project [Grant DEC-2019/34/E/HS4/00045]. X. Liao received financial support from the National Natural Science Foundation of China [Grant 71872144]. Supplemental Material: The software that supports the findings of this study is available within the paper and its Supplemental Information ( https://pubsonline.informs.org/doi/suppl/10.1287/ijoc.2023.1292 ) as well as from the IJOC GitHub software repository ( https://github.com/INFORMSJoC/2021.0328 ) at ( http://dx.doi.org/10.5281/zenodo.7608750 ).