Gene–gene interaction detection with deep learning

Cui, Tianyu; Mekkaoui, Khaoula El; Reinvall, Jaakko; Havulinna, Aki S.; Marttinen, Pekka; Kaski, Samuel

doi:10.1038/s42003-022-04186-y

Cited by 17 publications

(10 citation statements)

References 39 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Permutation tests are one possible solution. Permutation tests are generally used to determine the thresholds for IM and CIM (e.g., Broman et al 2003) and have been applied to detect gene-gene interactions based on the Shapley-based interaction score and neural networks (Cui et al 2022). However, further investigation is required to determine whether these tests work as expected.…”

Section: Discussionmentioning

confidence: 99%

Applying gradient tree boosting to QTL mapping with Shapley additive explanations

Ishibashi,

Onogi

2024

Preprint

View full text Add to dashboard Cite

Mapping quantitative trait loci (QTLs) is one of the major goals of quantitative genetics; however, identifying the interactions between QTLs remains challenging. Recently developed machine learning methods, such as deep learning and gradient boosting, are transforming the real world. These methods could advance QTL mapping methodologies because of their high capability for capturing complex relationships among features. One problem with applying such complex models to QTL mapping is evaluation of feature importance. In this study, XGBoost, a popular gradient tree boosting algorithm, was applied for QTL mapping in biparental populations with Shapley additive explanations (SHAPs). SHAP is a local (i.e., instance-wise) importance index with the desired properties as feature importance indices. The SHAP-assisted XGBoost (SHAP-XGB) was compared with conventional methods, including likelihood ratio tests (LRT), composite interval mapping (CIM), multiple interval mapping (MIM), and BayesC, using simulations and rice heading date data. SHAP-XGB performed comparable to CIM, MIM, and BayesC in mapping main QTL effects and was superior to conventional methods in mapping QTL interaction effects. As SHAP can evaluate local importance, interactions between markers can be visualized by plotting SHAP interaction values for each instance (plant/line). These results illustrated the strength of SHAP-XGB in detecting and interpreting epistatic QTLs and suggest the possibility that SHAP-XGB complements conventional methods.

show abstract

Section: Discussionmentioning

confidence: 99%

Applying gradient tree boosting to QTL mapping with Shapley additive explanations

Ishibashi,

Onogi

2024

Preprint

View full text Add to dashboard Cite

show abstract

“…Most approaches in computational genomics for uncovering interactions among genetic features typically involve built-in methods integrated within model architectures. These approaches include Gene‒ Gene Interaction Neural Networks (GGINNs) [14], the GenNet framework [4], Gene Network Inference (GNE) [15], Deep GONet [1], and multimodal deep learning models such as MDLCN [16]. Additionally, many of these approaches rely on preexisting gene interaction data to construct their networks, limiting their capacity to discover new phenotype-specific gene interactions or networks, namely, GenNet, GNE, MDLCN, and Deep GONet [1, 4, 15, 16].…”

Section: Introductionmentioning

confidence: 99%

A Model-agnostic Computational Method for Discovering Gene–Phenotype Relationships and Inferring Gene Networks viain silicoGene Perturbation

Stojšin,

Chen,

Zhao

2024

Preprint

View full text Add to dashboard Cite

BackgroundDeep learning architectures have advanced genotype‒phenotype mappings with precision but often obscure the roles of specific genes and their interactions. Our research introduces a model-agnostic computational methodology, capitalizing on the analytical strengths of deep learning models to serve as biological proxies, enabling interpretation of key gene interactions and their impact on phenotypic outcomes. The objective of this research is to refine the understanding of genetic networks in complex traits by leveraging the nuanced decision-making of advanced models.ResultsTesting was conducted across several computational models representing varying levels of complexity trained on gene expression datasets for the prediction of the Ki-67 biomarker, which is known for its prognostic value in breast cancer. The methodology is capable of using models as proxies to identify biologically significant genes and to infer relevant gene networks from an entirely data-driven analysis. Notably, the model-derived biomarkers (p-values of 0.013 and 0.003) outperformed the conventional Ki-67 biomarker (0.021) in terms of prognostic efficacy. Moreover, our analysis revealed high congruence between model precision and the biological relevance of the genes and gene relationships identified. Furthermore, we demonstrated that the complexity of the identified gene relationships was consistent with the decision-making intricacy of the model, with complex models capturing greater proportions of complex gene–gene interactions (61.2% and 31.1%) than simpler models (4.6%), reinforcing that the approach effectively captures biologically relevant in-model decision-making processes.ConclusionsThis methodology offers researchers a powerful tool to examine the decision-making processes within their genotype–phenotype mapping models. It accurately identifies critical genes and their interactions, revealing the biological rationale behind model decisions. It also enables comparisons of decision-making between different models. Furthermore, by discovering in-model critical gene networks, our approach helps bridge the gap between research and clinical applications. It facilitates the translation of complex, model-driven genetic discoveries into actionable clinical insights. This capability is pivotal for advancing personalized medicine, as it leverages the precision of deep learning models to uncover biologically relevant genes and gene networks and opens pathways for discovering new gene biomarker combinations and previously unknown gene interactions.

show abstract

“…In published studies, GG factors include gene expressions, methylation, SNPs, as well as several other types of omics measurements, for which the most widely used in the literature are genome-wide association studies (GWAS), successfully identifying alleles of risk for complex diseases by an association of SNPs with disease-related phenotypes. 1 , 2 , 3 , 4 , 5 …”

Section: Introductionmentioning

confidence: 99%

Role of gene interactions in the pathophysiology of skeletal dysplasias: A case report in Colombia

Madrid,

Giraldo

2024

Journal of Genetic Engineering and Biotechnology

View full text Add to dashboard Cite

Gene–gene interaction detection with deep learning

Cited by 17 publications

References 39 publications

Applying gradient tree boosting to QTL mapping with Shapley additive explanations

Applying gradient tree boosting to QTL mapping with Shapley additive explanations

A Model-agnostic Computational Method for Discovering Gene–Phenotype Relationships and Inferring Gene Networks viain silicoGene Perturbation

Role of gene interactions in the pathophysiology of skeletal dysplasias: A case report in Colombia

Contact Info

Product

Resources

About