Do little interactions get lost in dark random forests?

Wright, Marvin N.; Ziegler, Andreas; König, Inke R.

doi:10.1186/s12859-016-0995-8

Cited by 127 publications

(93 citation statements)

References 39 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Network size was actually the most important variable in the wild type cognitive impairment model. Given that high Gini index can sometimes reflect interaction effects [52], we explored, post hoc, the joint variable importance [64] of all pairs of features in the wild type cognitive impairment final model. This revealed that larger tumor volume tended to be associated with smaller network size in patients with wild type tumor (r = −0.30, p = 0.09).…”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

The effect of IDH1 mutation on the structural connectome in malignant astrocytoma

et al. 2016

View full text Add to dashboard Cite

Mutation of the IDH1 gene is associated with differences in malignant astrocytoma growth characteristics that impact phenotypic severity, including cognitive impairment. We previously demonstrated greater cognitive impairment in patients with IDH1 wild type tumor compared to those with IDH1 mutant, and therefore we hypothesized that brain network organization would be lower in patients with wild type tumors. Volumetric, T1-weighted MRI scans were obtained retrospectively from 35 patients with IDH1 mutant and 32 patients with wild type malignant astrocytoma (mean age = 45 ± 14 years) and used to extract individual level, gray matter connectomes. Graph theoretical analysis was then applied to measure efficiency and other connectome properties for each patient. Cognitive performance was categorized as impaired or not and random forest classification was used to explore factors associated with cognitive impairment. Patients with wild type tumor demonstrated significantly lower network efficiency in several medial frontal, posterior parietal and subcortical regions (p < 0.05, corrected for multiple comparisons). Patients with wild type tumor also demonstrated significantly higher incidence of cognitive impairment (p = 0.03). Random forest analysis indicated that network efficiency was inversely, though nonlinearly associated with cognitive impairment in both groups (p < 0.0001). Cognitive reserve appeared to mediate this relationship in patients with mutant tumor suggesting greater neuroplasticity and/or benefit from neuroprotective factors. Tumor volume was the greatest contributor to cognitive impairment in patients with wild type tumor, supporting our hypothesis that greater lesion momentum between grades may cause more disconnection of core neurocircuitry and consequently lower efficiency of information processing.

show abstract

Section: Discussionmentioning

confidence: 99%

“…education level) with neural plasticity and repair [50, 51], we also examined a global efficiency*education interaction term. Variable importance was evaluated using mean decrease in Gini index [52]. Because classes (impaired vs. not impaired) were imbalanced in both groups, random minority over-sampling was employed [53, 54].…”

Section: Methodsmentioning

confidence: 99%

The effect of IDH1 mutation on the structural connectome in malignant astrocytoma

et al. 2016

View full text Add to dashboard Cite

show abstract

“…While our analyses did include epigenomic data, CpGs were not simulated to have main effects, as they only had interaction effects through corresponding SNPs. Interactions are particularly difficult to assess in RF when the interacting features lack main effects as they are unlikely to be selected and split on at all [16], which suggests that they are unlikely to rank highly in RF, and could explain why RF-RFE did not improve their rankings. Thus, we were unable to explicitly assess the performance of RF-RFE with this nongenetics omics data set.…”

Section: Discussionmentioning

confidence: 99%

“…The other three causal CpG sites ranked very poorly in RF and did not suggest interactions. However, this was not unexpected as permutation importance scores were not designed to detect interactions [16] and reportedly fail to do so in high-dimensional data with weak marginal effects [17]. Although it has been shown that RF is influenced by interactions, it is very difficult to specifically identify which variables are interacting with current variable importance methods [16].…”

Section: Discussionmentioning

confidence: 99%

Using recursive feature elimination in random forest to account for correlated variables in high dimensional data

2018

View full text Add to dashboard Cite

BackgroundRandom forest (RF) is a machine-learning method that generally works well with high-dimensional problems and allows for nonlinear relationships between predictors; however, the presence of correlated predictors has been shown to impact its ability to identify strong predictors. The Random Forest-Recursive Feature Elimination algorithm (RF-RFE) mitigates this problem in smaller data sets, but this approach has not been tested in high-dimensional omics data sets.ResultsWe integrated 202,919 genotypes and 153,422 methylation sites in 680 individuals, and compared the abilities of RF and RF-RFE to detect simulated causal associations, which included simulated genotype–methylation interactions, between these variables and triglyceride levels. Results show that RF was able to identify strong causal variables with a few highly correlated variables, but it did not detect other causal variables.ConclusionsAlthough RF-RFE decreased the importance of correlated variables, in the presence of many correlated variables, it also decreased the importance of causal variables, making both hard to detect. These findings suggest that RF-RFE may not scale to high-dimensional data.

show abstract

“…Both random forest and Logic Forest provide quantitative importance measures for individual predictors allowing them to be ranked according to their relative importance in determining an outcome [17,22]. However, predictor importance for each variable represents the marginal effect of a predictor and if a set of predictors is associated with the outcome only through interactions effects, these marginal importance measures may mask such interaction effects [23]. Unlike random forest, Logic Forest also provides a quantitative measure of importance for interactions identified by the forest, which is advantageous in complex disease settings where interactions among genetic and environmental factors rather than main effects lead to disease.…”

Section: Introductionmentioning

confidence: 99%

An Analytic Approach Using Candidate Gene Selection and Logic Forest to Identify Gene by Environment Interactions (G × E) for Systemic Lupus Erythematosus in African Americans

Wolf

Ramos

Hyer

et al. 2018

Genes

View full text Add to dashboard Cite

Development and progression of many human diseases, such as systemic lupus erythematosus (SLE), are hypothesized to result from interactions between genetic and environmental factors. Current approaches to identify and evaluate interactions are limited, most often focusing on main effects and two-way interactions. While higher order interactions associated with disease are documented, they are difficult to detect since expanding the search space to all possible interactions of p predictors means evaluating 2p − 1 terms. For example, data with 150 candidate predictors requires considering over 1045 main effects and interactions. In this study, we present an analytical approach involving selection of candidate single nucleotide polymorphisms (SNPs) and environmental and/or clinical factors and use of Logic Forest to identify predictors of disease, including higher order interactions, followed by confirmation of the association between those predictors and interactions identified with disease outcome using logistic regression. We applied this approach to a study investigating whether smoking and/or secondhand smoke exposure interacts with candidate SNPs resulting in elevated risk of SLE. The approach identified both genetic and environmental risk factors, with evidence suggesting potential interactions between exposure to secondhand smoke as a child and genetic variation in the ITGAM gene associated with increased risk of SLE.

show abstract

Do little interactions get lost in dark random forests?

Cited by 127 publications

References 39 publications

The effect of IDH1 mutation on the structural connectome in malignant astrocytoma

The effect of IDH1 mutation on the structural connectome in malignant astrocytoma

Using recursive feature elimination in random forest to account for correlated variables in high dimensional data

An Analytic Approach Using Candidate Gene Selection and Logic Forest to Identify Gene by Environment Interactions (G × E) for Systemic Lupus Erythematosus in African Americans

Contact Info

Product

Resources

About