Epigenome-wide three-way interaction study identifies a complex pattern between TRIM27, KIAA0226, and smoking associated with overall survival of early-stage NSCLC

BHCox: Bayesian heredity-constrained Cox proportional hazards models for detecting gene-environment interactions

BACKGROUND

Gene-environment (G × E) interactions play a critical role in understanding the etiology of diseases and exploring the factors that affect disease prognosis. There are several challenges in detecting G × E interactions for censored survival outcomes, such as the high dimensionality, complexity of environmental effects, and specificity of survival analysis. The effect heredity, which incorporates the dependence of the main effects and interactions in the analysis, has been widely applied in the study of interaction detection. However, it has not yet been applied to Bayesian Cox proportional hazards models for detecting interactions for censored survival outcomes.

RESULTS

In this study, we propose Bayesian heredity-constrained Cox proportional hazards (BHCox) models with novel spike-and-slab and regularized horseshoe priors that incorporate effect heredity to identify and estimate the main and interaction effects. The no-U-turn sampler (NUTS) algorithm, which has been implemented in the R package brms, was used to fit the proposed model. Extensive simulations were performed to evaluate and compare our proposed approaches with other alternative models. The simulation studies illustrated that BHCox models outperform other alternative models. We applied the proposed method to real data of non-small-cell lung cancer (NSCLC) and identified biologically plausible G × smoking interactions associated with the prognosis of patients with NSCLC.

CONCLUSIONS

In summary, BHCox can be used to detect the main effects and interactions and thus have significant implications for the discovery of high-dimensional interactions in censored survival outcome data.

A two-phase epigenome-wide four-way gene-smoking interaction study of overall survival for early-stage non-small cell lung cancer

High-order interactions associated with non-small cell lung cancer (NSCLC) survival may elucidate underlying molecular mechanisms and identify potential therapeutic targets. Our previous work has identified a three-way interaction among pack-year of smoking (the number of packs of cigarettes smoked per day multiplied by the number of years the person has smoked) and two DNA methylation probes (cg05293407TRIM27 and cg00060500KIAA0226). However, whether a four-way interaction exists remains unclear. Therefore, we adopted a two-phase design to identify the four-way gene-smoking interactions by a hill-climbing strategy on the basis of the previously detected three-way interaction. One CpG probe, cg16658473SHISA9, was identified with FDR-q ≤ 0.05 in the discovery phase and P ≤ 0.05 in the validation phase. Meanwhile, the four-way interaction improved the discrimination ability for the prognostic prediction model, as indicated by the area under the receiver operating characteristic curve (AUC) for both 3- and 5-year survival. In summary, we identified a four-way interaction associated with NSCLC survival among pack-year of smoking, cg05293407TRIM27, cg00060500KIAA0226 and g16658473SHISA9, providing novel insights into the complex mechanisms underlying NSCLC progression.

BHAFT: Bayesian heredity-constrained accelerated failure time models for detecting gene-environment interactions in survival analysis

In addition to considering the main effects, understanding gene-environment (G × E) interactions is imperative for determining the etiology of diseases and the factors that affect their prognosis. In the existing statistical framework for censored survival outcomes, there are several challenges in detecting G × E interactions, such as handling high-dimensional omics data, diverse environmental factors, and algorithmic complications in survival analysis. The effect heredity principle has widely been used in studies involving interaction identification because it incorporates the dependence of the main and interaction effects. However, Bayesian survival models that incorporate the assumption of this principle have not been developed. Therefore, we propose Bayesian heredity-constrained accelerated failure time (BHAFT) models for identifying main and interaction (M-I) effects with novel spike-and-slab or regularized horseshoe priors to incorporate the assumption of effect heredity principle. The R package rstan was used to fit the proposed models. Extensive simulations demonstrated that BHAFT models had outperformed other existing models in terms of signal identification, coefficient estimation, and prognosis prediction. Biologically plausible G × E interactions associated with the prognosis of lung adenocarcinoma were identified using our proposed model. Notably, BHAFT models incorporating the effect heredity principle could identify both main and interaction effects, which are highly useful in exploring G × E interactions in high-dimensional survival analysis. The code and data used in our paper are available at https://github.com/SunNa-bayesian/BHAFT.

ARTEMIS: An independently validated prognostic prediction model of breast cancer incorporating epigenetic biomarkers with main effects and gene-gene interactions

INTRODUCTION

Breast cancer, a heterogeneous disease, is influenced by multiple genetic and epigenetic factors. The majority of prognostic models for breast cancer focus merely on the main effects of predictors, disregarding the crucial impacts of gene-gene interactions on prognosis.

OBJECTIVES

Using DNA methylation data derived from nine independent breast cancer cohorts, we developed an independently validated prognostic prediction model of breast cancer incorporating epigenetic biomarkers with main effects and gene-gene interactions (ARTEMIS) with an innovative 3-D modeling strategy. ARTEMIS was evaluated for discrimination ability using area under the receiver operating characteristics curve (AUC), and calibration using expected and observed (E/O) ratio. Additionally, we conducted decision curve analysis to evaluate its clinical efficacy by net benefit (NB) and net reduction (NR). Furthermore, we conducted a systematic review to compare its performance with existing models.

RESULTS

ARTEMIS exhibited excellent risk stratification ability in identifying patients at high risk of mortality. Compared to those below the 25th percentile of ARTEMIS scores, patients with above the 90th percentile had significantly lower overall survival time (HR = 15.43, 95% CI: 9.57-24.88, P = 3.06 × 10-29). ARTEMIS demonstrated satisfactory discrimination ability across four independent populations, with pooled AUC3-year = 0.844 (95% CI: 0.805-0.883), AUC5-year = 0.816 (95% CI: 0.775-0.857), and C-index = 0.803 (95% CI: 0.776-0.830). Meanwhile, ARTEMIS had well calibration performance with pooled E/O ratio 1.060 (95% CI: 1.038-1.083) and 1.090 (95% CI: 1.057-1.122) for 3- and 5-year survival prediction, respectively. Additionally, ARTEMIS is a clinical instrument with acceptable cost-effectiveness for detecting breast cancer patients at high risk of mortality (Pt = 0.4: NB3-year = 19‰, NB5-year = 62‰; NR3-year = 69.21%, NR5-year = 56.01%). ARTEMIS has superior performance compared to existing models in terms of accuracy, extrapolation, and sample size, as indicated by the systematic review. ARTEMIS is implemented as an interactive online tool available at http://bigdata.njmu.edu.cn/ARTEMIS/.

CONCLUSION

ARTEMIS is an efficient and practical tool for breast cancer prognostic prediction.

ATHENA: an independently validated autophagy-related epigenetic prognostic prediction model of head and neck squamous cell carcinoma

The majority of these existing prognostic models of head and neck squamous cell carcinoma (HNSCC) have unsatisfactory prediction accuracy since they solely utilize demographic and clinical information. Leveraged by autophagy-related epigenetic biomarkers, we aim to develop a better prognostic prediction model of HNSCC incorporating CpG probes with either main effects or gene-gene interactions. Based on DNA methylation data from three independent cohorts, we applied a 3-D analysis strategy to develop An independently validated auTophagy-related epigenetic prognostic prediction model of HEad and Neck squamous cell carcinomA (ATHENA). Compared to prediction models with only demographic and clinical information, ATHENA has substantially improved discriminative ability, prediction accuracy and more clinical net benefits, and shows robustness in different subpopulations, as well as external populations. Besides, epigenetic score of ATHENA is significantly associated with tumor immune microenvironment, tumor-infiltrating immune cell abundances, immune checkpoints, somatic mutation and immunity-related drugs. Taken together these results, ATHENA has the demonstrated feasibility and utility of predicting HNSCC survival ( http://bigdata.njmu.edu.cn/ATHENA/ ).

A trans-omics assessment of gene-gene interaction in early-stage NSCLC

Epigenome-wide gene-gene (G × G) interactions associated with non-small-cell lung cancer (NSCLC) survival may provide insights into molecular mechanisms and therapeutic targets. Hence, we proposed a three-step analytic strategy to identify significant and robust G × G interactions that are relevant to NSCLC survival. In the first step, among 49 billion pairs of DNA methylation probes, we identified 175 775 G × G interactions with PBonferroni  ≤ 0.05 in the discovery phase of epigenomic analysis; among them, 15 534 were confirmed with P ≤ 0.05 in the validation phase. In the second step, we further performed a functional validation for these G × G interactions at the gene expression level by way of a two-phase (discovery and validation) transcriptomic analysis, and confirmed 25 significant G × G interactions enriched in the 6p21.33 and 6p22.1 regions. In the third step, we identified two G × G interactions using the trans-omics analysis, which had significant (P ≤ 0.05) epigenetic cis-regulation of transcription and robust G × G interactions at both the epigenetic and transcriptional levels. These interactions were cg14391855 × cg23937960 (βinteraction  = 0.018, P = 1.87 × 10-12 ), which mapped to RELA × HLA-G (βinteraction  = 0.218, P = 8.82 × 10-11 ) and cg08872738 × cg27077312 (βinteraction  = -0.010, P = 1.16 × 10-11 ), which mapped to TUBA1B × TOMM40 (βinteraction =-0.250, P = 3.83 × 10-10 ). A trans-omics mediation analysis revealed that 20.3% of epigenetic effects on NSCLC survival were significantly (P = 0.034) mediated through transcriptional expression. These statistically significant trans-omics G × G interactions can also discriminate patients with high risk of mortality. In summary, we identified two G × G interactions at both the epigenetic and transcriptional levels, and our findings may provide potential clues for precision treatment of NSCLC.

Epigenome-wide gene–age interaction study reveals reversed effects of MORN1 DNA methylation on survival between young and elderly oral squamous cell carcinoma patients

DNA methylation serves as a reversible and prognostic biomarker for oral squamous cell carcinoma (OSCC) patients. It is unclear whether the effect of DNA methylation on OSCC overall survival varies with age. As a result, we performed a two-phase gene–age interaction study of OSCC prognosis on an epigenome-wide scale using the Cox proportional hazards model. We identified one CpG probe, cg11676291_MORN1, whose effect was significantly modified by age (HR_discovery = 1.018, p = 4.07 × 10⁻⁰⁷, FDR-q = 3.67 × 10⁻⁰²; HR_validation = 1.058, p = 8.09 × 10⁻⁰³; HR_combined = 1.019, p = 7.36 × 10⁻¹⁰). Moreover, there was an antagonistic interaction between hypomethylation of cg11676291_MORN1 and age (HR_interaction = 0.284; 95% CI, 0.135–0.597; p = 9.04 × 10⁻⁰⁴). The prognosis of OSCC patients was well discriminated by the prognostic score incorporating cg11676291_MORN1–age interaction (HR_{high vs. low} = 3.66, 95% CI: 2.40–5.60, p = 1.93 × 10⁻⁰⁹). By adding 24 significant gene–age interactions using a looser criterion, we significantly improved the area under the receiver operating characteristic curve (AUC) of the model at 3- and 5-year prognostic prediction (AUC_3-year = 0.80, AUC_5-year = 0.79, C-index = 0.75). Our study identified a significant interaction between cg11676291_MORN1 and age on OSCC survival, providing a potential therapeutic target for OSCC patients.