Comethyl: A network-based methylome approach to investigate the multivariate nature of health and disease

Mordaunt, Charles E.; Mouat, Julia S.; Schmidt, Rebecca Jean; LaSalle, Janine M.

doi:10.1101/2021.07.14.452385

Cited by 3 publications

(2 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We therefore decided to split our dataset chromosome-wise and train individual models on each chromosome. This assumes that there are no long range inter-chromosomal interactions of CpGsalthough this topic is underrepresented in literature, there is evidence of strong short range correlations between CpGs and methods focusing on co-methylation patterns assume short windows for interaction [55,58]. While the reconstruction accuracies we achieved are close to the theoretical maximum valueas indicated through models with the same number of latent dimensions as input features -they range around a Pearson correlation of 0.65.…”

Section: Limitationsmentioning

confidence: 51%

mEthAE: an Explainable AutoEncoder for methylation data

Katz

Santos

Saccenti

et al. 2023

Preprint

View full text Add to dashboard Cite

Despite the wealth of knowledge generated through epigenome-wide association studies our understanding of the relationships of CpG sites is still limited, as analysis of DNA methylation data remains difficult due its high dimensionality. To combat this problem, deep learning algorithms, such as autoencoders, are increasingly applied to capture the complex patterns and reduce dimensionality into latent space. We believe that the way an autoencoder groups together CpGs in its latent dimensions has biological meaning and might reveal novel insights regarding the relationship of CpGs. Therefore, in this work, we propose a chromosome-wise autoencoder for interpretable dimensionality reduction of methylation data (mEthAE). Our framework shows an impressive reduction in dimensions of up to 400-fold compared to the provided input, without compromising on reconstruction accuracy or predictive power in the latent space. Through our perturbation-based interpretability approach we revealed groups of CpGs which are highly connected across all latent dimensions (global CpGs) and were significantly more often reported in EWAS studies, indicating our interpretability method can successfully identify CpGs with biological relevance. In an attempt to gain a deeper understanding of the relationship between individual CpG sites, we focused on interpreting individual latent features and found that CpGs connected to a common feature do not share biological associations, correlation patterns, or are located in close proximity on the chromosome. We conclude that while there is evidence that the autoencoder does not group CpGs randomly, the logic behind the observed CpG relationships can not be delineated easily. With regards to the analyses done in this work, we believe that the autoencoder groups CpGs according to long range non-linear interaction patterns that lack characterisation in the current epigenetic research landscape.

show abstract

Section: Limitationsmentioning

confidence: 51%

mEthAE: an Explainable AutoEncoder for methylation data

Katz

Santos

Saccenti

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…After alignment to a reference genome, percent methylation is calculated for individual CpGs or clusters of CpGs in differentially methylated regions (DMRs). Low pass WGBS varies from 1x-10x coverage genome-wide, which is sufficient for DMR-based analyses, network analyses of comethylated regions [152], as well as global methylation analyses). High coverage WGBS varies from 30x-50x coverage genome-wide, which is sufficient for single CpG resolution [153].…”

Section: Sequencing-based Methodsmentioning

confidence: 99%

Epigenomic signatures reveal mechanistic clues and predictive markers for autism spectrum disorder

LaSalle

2023

Mol Psychiatry

Self Cite

View full text Add to dashboard Cite

Autism spectrum disorder (ASD) comprises a heterogeneous group of neurodevelopmental outcomes in children with a commonality in deficits in social communication and language combined with repetitive behaviors and interests. The etiology of ASD is heterogeneous, as several hundred genes have been implicated as well as multiple in utero environmental exposures. Over the past two decades, epigenetic investigations, including DNA methylation, have emerged as a novel way to capture the complex interface of multivariate ASD etiologies. More recently, epigenome-wide association studies using human brain and surrogate accessible tissues have revealed some convergent genes that are epigenetically altered in ASD, many of which overlap with known genetic risk factors. Unlike transcriptomes, epigenomic signatures defined by DNA methylation from surrogate tissues such as placenta and cord blood can reflect past differences in fetal brain gene transcription, transcription factor binding, and chromatin. For example, the discovery of NHIP (neuronal hypoxia inducible, placenta associated) through an epigenome-wide association in placenta, identified a common genetic risk for ASD that was modified by prenatal vitamin use. While epigenomic signatures are distinct between different genetic syndromic causes of ASD, bivalent chromatin and some convergent gene pathways are consistently epigenetically altered in both syndromic and idiopathic ASD, as well as some environmental exposures. Together, these epigenomic signatures hold promising clues towards improved early prediction and prevention of ASD as well genes and gene pathways to target for pharmacological interventions. Future advancements in single cell and multi-omic technologies, machine learning, as well as non-invasive screening of epigenomic signatures during pregnancy or newborn periods are expected to continue to impact the translatability of the recent discoveries in epigenomics to precision public health.

show abstract

Networks of placental DNA methylation correlate with maternal serum PCB concentrations and child neurodevelopment

Mouat

Neier

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Background: Gestational exposure to polychlorinated biphenyls (PCBs) has been associated with elevated risk for neurodevelopmental disorders. The mechanism of risk is unclear but may involve placental epigenetics. Prior studies have associated differential placental DNA methylation with maternal PCB exposure or with increased risk of autism spectrum disorder (ASD). However, sequencing-based placental methylomes have not previously been tested for simultaneous associations with maternal PCB levels and child neurodevelopmental outcomes. Objectives: We aimed to identify placental DNA methylation patterns associated with maternal PCB levels and child neurodevelopmental outcomes in the high-risk ASD MARBLES cohort. Methods: We measured 209 PCB congeners in 104 maternal serum samples collected at delivery. We identified networks of DNA methylation from 147 placenta samples using the Comethyl R package, which performs weighted gene correlation network analysis for whole genome bisulfite sequencing data. We tested placental DNA methylation modules for association with maternal serum PCB levels, child neurodevelopment, and other participant traits. Results: PCBs 153 + 168, 170, 180 + 193, and 187 were detected in over 50% of maternal serum samples and were highly correlated with one another. Consistent with previous findings, maternal age was the strongest predictor of serum PCB levels, alongside year of sample collection, pre-pregnancy BMI, and polyunsaturated fatty acid levels. Twenty seven modules of placental DNA methylation were identified, including five which significantly correlated with one or more PCBs, and four which correlated with child neurodevelopment. Two modules associated with maternal PCB levels as well as child neurodevelopment, and mapped to CSMD1 and AUTS2, genes previously implicated in ASD and identified as differentially methylated regions in mouse brain and placenta following gestational PCB exposure. Conclusions: Placental DNA co-methylation modules were associated with maternal PCBs and child neurodevelopment. Methylation of CSMD1 and AUTS2 could potentially be mechanistically involved in ASD risk following maternal PCB exposure.

show abstract

Comethyl: A network-based methylome approach to investigate the multivariate nature of health and disease

Cited by 3 publications

References 32 publications

mEthAE: an Explainable AutoEncoder for methylation data

mEthAE: an Explainable AutoEncoder for methylation data

Epigenomic signatures reveal mechanistic clues and predictive markers for autism spectrum disorder

Networks of placental DNA methylation correlate with maternal serum PCB concentrations and child neurodevelopment

Contact Info

Product

Resources

About