Dataset Information

Deep embedded clustering generalisability and adaptation for integrating mixed datatypes: two critical care cohorts.

ABSTRACT: We validated a Deep Embedded Clustering (DEC) model and its adaptation for integrating mixed datatypes (in this study, numerical and categorical variables). Deep Embedded Clustering (DEC) is a promising technique capable of managing extensive sets of variables and non-linear relationships. Nevertheless, DEC cannot adequately handle mixed datatypes. Therefore, we adapted DEC by replacing the autoencoder with an X-shaped variational autoencoder (XVAE) and optimising hyperparameters for cluster stability. We call this model "X-DEC". We compared DEC and X-DEC by reproducing a previous study that used DEC to identify clusters in a population of intensive care patients. We assessed internal validity based on cluster stability on the development dataset. Since generalisability of clustering models has insufficiently been validated on external populations, we assessed external validity by investigating cluster generalisability onto an external validation dataset. We concluded that both DEC and X-DEC resulted in clinically recognisable and generalisable clusters, but X-DEC produced much more stable clusters.

SUBMITTER: de Kok JWTM

PROVIDER: S-EPMC10781731 | biostudies-literature | 2024 Jan

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Deep embedded clustering generalisability and adaptation for integrating mixed datatypes: two critical care cohorts.

de Kok Jip W T M JWTM van Rosmalen Frank F Koeze Jacqueline J Keus Frederik F van Kuijk Sander M J SMJ Castela Forte José J Schnabel Ronny M RM Driessen Rob G H RGH van Herpt Thijs T W TTW Sels Jan-Willem E M JEM Bergmans Dennis C J J DCJJ Lexis Chris P H CPH van Doorn William P T M WPTM Meex Steven J R SJR Xu Minnan M Borrat Xavier X Cavill Rachel R van der Horst Iwan C C ICC van Bussel Bas C T BCT

Scientific reports 20240110 1

We validated a Deep Embedded Clustering (DEC) model and its adaptation for integrating mixed datatypes (in this study, numerical and categorical variables). Deep Embedded Clustering (DEC) is a promising technique capable of managing extensive sets of variables and non-linear relationships. Nevertheless, DEC cannot adequately handle mixed datatypes. Therefore, we adapted DEC by replacing the autoencoder with an X-shaped variational autoencoder (XVAE) and optimising hyperparameters for cluster sta ...[more]

PMID: 38200252

Similar Datasets

Project description:BackgroundMobile transdiagnostic therapies offer a solution to the challenges of limited access to psychological care. However, it is unclear if individuals can actively synthesize and adopt concepts and skills via an app without clinician support.AimsThe present study measured comprehension of and engagement with a mobile acceptance and commitment therapy (ACT) intervention in two independent cohorts. Authors hypothesized that participants would recognize that behaviors can be flexible in form and function and respond in an ACT process-aligned manner.MethodsMixed-methods analyses were performed on open-ended responses collected from initial participants (n = 49) in two parallel micro-randomized trials with: 1) first-generation college students (FGCSs) (n = 25) from a four-year public research university and 2) individuals diagnosed with bipolar disorder (BP) (n = 24). Twice each day over six weeks, participants responded to questions about mood and behavior, after which they had a 50-50 chance of receiving an ACT-based intervention. Participants identified current behavior and categorized behavior as values-based or avoidant. Interventions were selected randomly from 84 possible prompts, each targeting one ACT process: engagement with values, openness to internal experiences, or self-awareness. Participants were randomly assigned to either exploratory (10 FGCS, 9 BP) or confirmatory (15 FGCS, 15 BP) groups for analyses. Responses from the exploratory group were used to inductively derive a qualitative coding system. This system was used to code responses in the confirmatory group. Coded confirmatory data were used for final analyses.ResultsOver 50% of participants in both cohorts submitted a non-blank response 100% of the time. For over 50% of participants, intervention responses aligned with the target ACT process for at least 96% of the time (FGCS) and 91% of the time (BP), and current behavior was labeled as values-based 70% (FGCS) and 85% (BP) of the time. Participants labeled similar behaviors flexibly as either values-based or avoidant in different contexts. Dominant themes were needs-based behaviors, interpersonal and family relationships, education, and time as a cost.ConclusionsBoth cohorts were engaged with the app, as demonstrated by responses that aligned with ACT processes. This suggests that participants had some level of understanding that behavior can be flexible in form and function.

Project description:BackgroundThe rationale for centre selection in randomised controlled trials (RCTs) is often unclear but may have important implications for the generalisability of trial results. The aims of this study were to evaluate the factors which currently influence centre selection in RCTs and consider how generalisability considerations inform current and optimal practice.Methods and findingsMixed methods approach consisting of a systematic review and meta-summary of centre selection criteria reported in RCT protocols funded by the UK National Institute of Health Research (NIHR) initiated between January 2005-January 2012; and an online survey on the topic of current and optimal centre selection, distributed to professionals in the 48 UK Clinical Trials Units and 10 NIHR Research Design Services. The survey design was informed by the systematic review and by two focus groups conducted with trialists at the Birmingham Centre for Clinical Trials. 129 trial protocols were included in the systematic review, with a total target sample size in excess of 317,000 participants. The meta-summary identified 53 unique centre selection criteria. 78 protocols (60%) provided at least one criterion for centre selection, but only 31 (24%) protocols explicitly acknowledged generalisability. This is consistent with the survey findings (n = 70), where less than a third of participants reported generalisability as a key driver of centre selection in current practice. This contrasts with trialists' views on optimal practice, where generalisability in terms of clinical practice, population characteristics and economic results were prime considerations for 60% (n = 42), 57% (n = 40) and 46% (n = 32) of respondents, respectively.ConclusionsCentres are rarely enrolled in RCTs with an explicit view to external validity, although trialists acknowledge that incorporating generalisability in centre selection should ideally be more prominent. There is a need to operationalize 'generalisability' and incorporate it at the design stage of RCTs so that results are readily transferable to 'real world' practice.

Project description:BackgroundSepsis is a heterogenous syndrome and individualized management strategy is the key to successful treatment. Genome wide expression profiling has been utilized for identifying subclasses of sepsis, but the clinical utility of these subclasses was limited because of the classification instability, and the lack of a robust class prediction model with extensive external validation. The study aimed to develop a parsimonious class model for the prediction of class membership and validate the model for its prognostic and predictive capability in external datasets.MethodsThe Gene Expression Omnibus (GEO) and ArrayExpress databases were searched from inception to April 2020. Datasets containing whole blood gene expression profiling in adult sepsis patients were included. Autoencoder was used to extract representative features for k-means clustering. Genetic algorithms (GA) were employed to derive a parsimonious 5-gene class prediction model. The class model was then applied to external datasets (n = 780) to evaluate its prognostic and predictive performance.FindingsA total of 12 datasets involving 1613 patients were included. Two classes were identified in the discovery cohort (n = 685). Class 1 was characterized by immunosuppression with higher mortality than class 2 (21.8% [70/321] vs. 12.1% [44/364]; p < 0.01 for Chi-square test). A 5-gene class model (C14orf159, AKNA, PILRA, STOM and USP4) was developed with GA. In external validation cohorts, the 5-gene class model (AUC: 0.707; 95% CI: 0.664 - 0.750) performed better in predicting mortality than sepsis response signature (SRS) endotypes (AUC: 0.610; 95% CI: 0.521 - 0.700), and performed equivalently to the APACHE II score (AUC: 0.681; 95% CI: 0.595 - 0.767). In the dataset E-MTAB-7581, the use of hydrocortisone was associated with increased risk of mortality (OR: 3.15 [1.13, 8.82]; p = 0.029) in class 2. The effect was not statistically significant in class 1 (OR: 1.88 [0.70, 5.09]; p = 0.211).InterpretationOur study identified two classes of sepsis that showed different mortality rates and responses to hydrocortisone therapy. Class 1 was characterized by immunosuppression with higher mortality rate than class 2. We further developed a 5-gene class model to predict class membership.FundingThe study was funded by the National Natural Science Foundation of China (Grant No. 81,901,929).

Dataset Information

Deep embedded clustering generalisability and adaptation for integrating mixed datatypes: two critical care cohorts.

Publications

Deep embedded clustering generalisability and adaptation for integrating mixed datatypes: two critical care cohorts.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets