Comparison of strategies for scalable causal discovery of latent variable models from mixed data

Raghu, Vineet K.; Ramsey, Joseph D.; Morris, Alison; Manatakis, Dimitrios V.; Sprites, Peter; Chrysanthis, Panos K.; Glymour, Clark; Benos, Panayiotis V.

doi:10.1007/s41060-018-0104-3

Cited by 46 publications

(44 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For modelling the dependencies in the data, we used MGM-FCI-MAX,25 which learns a directed graph over mixed data types in the presence of latent confounders. This is important because medical and biomedical data usually contain variables of mixed types (continuous and discrete) and unmeasured confounders (due to lack of knowledge or measurement inability).…”

Section: Methodsmentioning

confidence: 99%

Feasibility of lung cancer prediction from low-dose CT scan and smoking factors using causal models

Raghu

Zhao

et al. 2019

Thorax

View full text Add to dashboard Cite

IntroductionLow-dose CT (LDCT) is currently used in lung cancer screening of high-risk populations for early lung cancer diagnosis. However, 96% of individuals with detected nodules are false positives.MethodsIn order to develop an efficient early lung cancer predictor from clinical, demographic and LDCT features, we studied a total of 218 subjects with lung cancer or benign nodules. Probabilistic graphical models (PGMs) were used to integrate demographics, clinical data and LDCT features from 92 subjects (training cohort) from the Pittsburgh Lung Screening Study cohort.ResultsLearnt PGMs identified three variables directly (causally) linked to malignant nodules and the largest benign nodule and used them to build the Lung Cancer Causal Model (LCCM), which was validated in a separate cohort of 126 subjects. Nodule and vessel numbers and years since the subject quit smoking were sufficient to discriminate malignant from benign nodules. Comparison with existing predictors in the training and validation cohorts showed that (1) incorporating LDCT scan features greatly enhances predictive accuracy; and (2) LCCM improves cancer detection over existing methods, including the Brock parsimonious model (p<0.001). Notably, the number of surrounding vessels, a feature not previously used in predictive models, significantly improves predictive efficiency. Based on the validation cohort results, LCCM is able to identify 30% of the benign nodules without risk of misclassifying cancer nodules.DiscussionLCCM shows promise as a lung cancer predictor as it is significantly improved over existing models. Validated in a larger, prospective study, it may help reduce unnecessary follow-up visits and procedures.

show abstract

Section: Methodsmentioning

confidence: 99%

Feasibility of lung cancer prediction from low-dose CT scan and smoking factors using causal models

Raghu

Zhao

et al. 2019

Thorax

View full text Add to dashboard Cite

show abstract

“…PGMs can estimate and graphically represent the complex relationships of large numbers of variables that interact with each other, allowing for the discovery of direct links between variables based on their conditional dependencies. We used the CausalMGM (Causal Mixed Graphical Model) R package 1 , a novel algorithm that can accurately identify the underlying graphical model structure over mixed data types (continuous and discrete) ( Sedgewick et al, 2016 ; Raghu et al, 2017 ).…”

Section: Methodsmentioning

confidence: 99%

Respiratory Microbiome Profiling for Etiologic Diagnosis of Pneumonia in Mechanically Ventilated Patients

et al. 2018

Self Cite

View full text Add to dashboard Cite

Etiologic diagnosis of bacterial pneumonia relies on identification of causative pathogens by cultures, which require extended incubation periods and have limited sensitivity. Next-generation sequencing of microbial DNA directly from patient samples may improve diagnostic accuracy for guiding antibiotic prescriptions. In this study, we hypothesized that enhanced pathogen detection using sequencing can improve upon culture-based diagnosis and that certain sequencing profiles correlate with host response. We prospectively collected endotracheal aspirates and plasma within 72 h of intubation from patients with acute respiratory failure. We performed 16S rRNA gene sequencing to determine pathogen abundance in lung samples and measured plasma biomarkers to assess host responses to detected pathogens. Among 56 patients, 12 patients (21%) had positive respiratory cultures. Sequencing revealed lung communities with low diversity (p < 0.02) dominated by taxa (>50% relative abundance) corresponding to clinically isolated pathogens (concordance p = 0.009). Importantly, sequencing detected dominant pathogens in 20% of the culture-negative patients exposed to broad-spectrum empiric antibiotics. Regardless of culture results, pathogen dominance correlated with increased plasma markers of host injury (receptor of advanced glycation end-products-RAGE) and inflammation (interleukin-6, tumor necrosis factor receptor 1-TNFR1) (p < 0.05), compared to subjects without dominant pathogens in their lung communities. Machine-learning algorithms identified pathogen abundance by sequencing as the most informative predictor of culture positivity. Thus, enhanced detection of pathogenic bacteria by sequencing improves etiologic diagnosis of pneumonia, correlates with host responses, and offers substantial opportunity for individualized therapeutic targeting and antimicrobial stewardship. Clinical translation will require validation with rapid whole meta-genome sequencing approaches to guide real-time antibiotic prescriptions.

show abstract

“…Casual inference was performed by using the on-line CausalMGM 46 and the command-line tool for FCI 47 .…”

Section: Casual Inference Analysismentioning

confidence: 99%

Lipidomic Signatures Align with Inflammatory Patterns and Outcomes in Critical Illness

Wu¹,

Cyr²,

Gruen³

et al. 2021

Preprint

View full text Add to dashboard Cite

Alterations in lipid metabolism have the potential to be markers as well as drivers of the pathobiology of acute critical illness. Here, we took advantage of the temporal precision offered by trauma as a common cause of critical illness to identify the dynamic patterns in the circulating lipidome in critically ill humans. The major findings include an early loss of all classes of circulating lipids followed by a delayed and selective lipogenesis in patients destined to remain critically ill. Early in the clinical course, Fresh Frozen Plasma administration led to improved survival in association with preserved lipid levels that related to favorable changes in coagulation and inflammation biomarkers. Late over-representation of phosphatidylethanolamines with critical illness led to the validation of a Lipid Reprogramming Score that was prognostic not only in trauma but also severe COVID-19 patients. Our lipidomic findings provide a new paradigm for the lipid response underlying critical illness.

show abstract

Comparison of strategies for scalable causal discovery of latent variable models from mixed data

Cited by 46 publications

References 19 publications

Feasibility of lung cancer prediction from low-dose CT scan and smoking factors using causal models

Feasibility of lung cancer prediction from low-dose CT scan and smoking factors using causal models

Respiratory Microbiome Profiling for Etiologic Diagnosis of Pneumonia in Mechanically Ventilated Patients

Lipidomic Signatures Align with Inflammatory Patterns and Outcomes in Critical Illness

Contact Info

Product

Resources

About