A new method for exploring gene–gene and gene–environment interactions in GWAS with tree ensemble methods and SHAP values

Abstract

Background The identification of gene–gene and gene–environment interactions in genome-wide association studies is challenging due to the unknown nature of the interactions and the overwhelmingly large number of possible combinations. Parametric regression models are suitable to look for prespecified interactions. Nonparametric models such as tree ensemble models, with the ability to detect any unspecified interaction, have previously been difficult to interpret. However, with the development of methods for model explainability, it is now possible to interpret tree ensemble models efficiently and with a strong theoretical basis. Results We propose a tree ensemble- and SHAP-based method for identifying as well as interpreting potential gene–gene and gene–environment interactions on large-scale biobank data. A set of independent cross-validation runs are used to implicitly investigate the whole genome. We apply and evaluate the method using data from the UK Biobank with obesity as the phenotype. The results are in line with previous research on obesity as we identify top SNPs previously associated with obesity. We further demonstrate how to interpret and visualize interaction candidates. Conclusions The new method identifies interaction candidates otherwise not detected with parametric regression models. However, further research is needed to evaluate the uncertainties of these candidates. The method can be applied to large-scale biobanks with high-dimensional data.

Comparing the XGBoost machine learning algorithm to polygenic scoring for the prediction of intelligence based on genotype data

“…For example, a SNP may achieve a higher score in XGBoost than in the GWAS because its effect is dependent on an interaction with another SNP, which would not be captured in the GWAS. A recent study (Johnsen et al, 2021) used SHapley Additive exPlanation (SHAP) values to score the pairwise contribution of SNPs to XGBoost predictions (rather than using gain values to score the individual contribution of each SNP). However, more work needs to be done to assess the validity of SHap-values in this context (Johnsen et al, 2021).…”

“…A recent study (Johnsen et al, 2021) used SHapley Additive exPlanation (SHAP) values to score the pairwise contribution of SNPs to XGBoost predictions (rather than using gain values to score the individual contribution of each SNP). However, more work needs to be done to assess the validity of SHap-values in this context (Johnsen et al, 2021).…”

Compressive Strength Evaluation of Ultra-High-Strength Concrete by Machine Learning

“…A better model was then proposed based on the performance of several statistical parameters to predict UHSC outcomes. Furthermore, a post hoc model-agnostic technique, i.e., SHapley Additive exPlanations (SHAP), was also implemented to give ML model insight [ 28 , 29 ]. The integration of SHAP with ML algorithms was performed in the current research to provide a comprehensive understanding of the mix design of concrete, regarding its strength parameters through its non-linear complex behavior, and to describe the contribution of input parameters by assigning a weight factor to each input parameter.…”

Interpretable XGBoost-SHAP model predicts the nanoparticles delivery and reveals its interaction with tumor genomic profiles

“…[18,19] SHAP explores and illustrates the importance of and correlations between input features based on the game-theoretic approach, which has been applied to explore the gene-gene and gene-environment interactions very recently. [20] To capture the NP-tumor interaction complexities and enable the accurate prediction of delivery efficacy of NPs in different tumors models, we developed ML models by integrating NPs properties of 162 samples from literature data-mining and genomic profiles of 23 cell lines from the CCLE database. We observed that the genomic mutations of tumors have a dominating influence on the delivery efficacy, which was comparable to the influence of tumor weight and NPs properties.…”

“…[18,19] SHAP explores and illustrates the importance of and correlations between input features based on the game-theoretic approach, which has been applied to explore the gene-gene and gene-environment interactions very recently. [20]…”