Increasing Temporal Sensitivity of Omics Association Studies with Epigenome-Wide Distributed Lag Models.

Current methods for identifying temporal windows of effect for time-varying exposures in omics settings can control false discovery rates at the biomarker-level but cannot efficiently screen for timing-specific effects in high dimensions. Current approaches leverage separate models for site screening and identification of susceptible time windows, which miss associations that vary over time. We introduce the epigenome-wide distributed lag model (EWDLM), a novel approach that combines traditional false discovery rate methods with the distributed lag model (DLM) to screen for timing-specific effects in high dimensional settings. This is accomplished by marginalizing DLM effect estimates over time and correcting for multiple comparisons. In a simulation investigating timing-specific effects of ambient air pollution during pregnancy on DNA methylation across the epigenome at age 12 years, EWDLM achieved an increased sensitivity for associations limited to specific periods of time compared to traditional two-stage approaches. In a real-world EWDLM analysis, 353 CpG sites at which DNAm measured at age 12 was significantly associated with PM2.5 exposure during pregnancy were identified. EWDLM is a novel method that provides an efficient and sensitive way to screen epigenomic datasets for associations with exposures localized to specific time periods.

Epigenome-wide association of neonatal methylation and trimester-specific prenatal PM2.5 exposure.

Exposure to particulate matter with an aerodynamic diameter smaller than 2.5 microns (PM2.5) can affect birth outcomes through physiological pathways such as inflammation. One potential way PM2.5 affects physiology could be through altering DNA methylation (DNAm). Considering that exposures during specific windows of gestation may have unique effects on DNAm, we hypothesized a timing-specific association between PM2.5 exposure during pregnancy and DNAm in the neonatal epithelial-cell epigenome. Methods: After collecting salivary samples from a cohort of 91 neonates, DNAm was assessed at over 850,000 cytosine-guanine dinucleotide (CpG) methylation sites on the epigenome using the MethylationEPIC array. Daily ambient PM2.5 concentrations were estimated based on the mother's address of primary residence during pregnancy. PM2.5 was averaged over the first two trimesters, separately and combined, and tested for association with DNAm through an epigenome-wide association (EWA) analysis. For each EWA, false discovery rate (FDR)-corrected P < 0.05 constituted a significant finding and every CpG site with uncorrected P < 0.0001 was selected to undergo pathway and network analysis to identify molecular functions enriched by them. Results: Our analysis showed that cg18705808 was associated with the combined average of PM2.5. Pathway and network analysis revealed little similarity between the first two trimesters. Previous studies reported that TMEM184A, the gene regulated by cg18705808, has a putative role in inflammatory pathways. Conclusions: The differences in pathway and network analyses could potentially indicate trimester-specific effects of PM2.5 on DNAm. Further analysis with greater temporal resolution would be valuable to fully characterize the effect of PM2.5 on DNAm and child development.

Enhancing Diagnosis of Autism With Optimized Machine Learning Models and Personal Characteristic Data.

Autism spectrum disorder (ASD) is a developmental disorder, affecting about 1% of the global population. Currently, the only clinical method for diagnosing ASD are standardized ASD tests which require prolonged diagnostic time and increased medical costs. Our objective was to explore the predictive power of personal characteristic data (PCD) from a large well-characterized dataset to improve upon prior diagnostic models of ASD. We extracted six personal characteristics (age, sex, handedness, and three individual measures of IQ) from 851 subjects in the Autism Brain Imaging Data Exchange (ABIDE) database. ABIDE is an international collaborative project that collected data from a large number of ASD patients and typical non-ASD controls from 17 research and clinical institutes. We employed this publicly available database to test nine supervised machine learning models. We implemented a cross-validation strategy to train and test those machine learning models for classification between typical non-ASD controls and ASD patients. We assessed classification performance using accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve (AUC). Of the nine models we tested using six personal characteristics, the neural network model performed the best with a mean AUC (SD) of 0.646 (0.005), followed by k-nearest neighbor with a mean AUC (SD) of 0.641 (0.004). This study established an optimal ASD classification performance with PCD as features. With additional discriminative features (e.g., neuroimaging), machine learning models may ultimately enable automated clinical diagnosis of autism.