Predicting environmentally responsive transgenerational differential DNA methylated regions (epimutations) in the genome using a hybrid deep-machine learning approach

Abstract

Background Deep learning is an active bioinformatics artificial intelligence field that is useful in solving many biological problems, including predicting altered epigenetics such as DNA methylation regions. Deep learning (DL) can learn an informative representation that addresses the need for defining relevant features. However, deep learning models are computationally expensive, and they require large training datasets to achieve good classification performance. Results One approach to addressing these challenges is to use a less complex deep learning network for feature selection and Machine Learning (ML) for classification. In the current study, we introduce a hybrid DL-ML approach that uses a deep neural network for extracting molecular features and a non-DL classifier to predict environmentally responsive transgenerational differential DNA methylated regions (DMRs), termed epimutations, based on the extracted DL-based features. Various environmental toxicant induced epigenetic transgenerational inheritance sperm epimutations were used to train the model on the rat genome DNA sequence and use the model to predict transgenerational DMRs (epimutations) across the entire genome. Conclusion The approach was also used to predict potential DMRs in the human genome. Experimental results show that the hybrid DL-ML approach outperforms deep learning and traditional machine learning methods.

Identifying unique exposure-specific transgenerational differentially DNA methylated region epimutations in the genome using hybrid deep learning prediction models

“…These results indicate potential common mechanisms by which most toxicants affect the genome. Observations can be contrasted to those in previous work [ 18 ] that identify motifs in the features extracted from the DL network. Feature motifs do not necessarily represent common patterns in DMRs, but can also represent patterns in non-DMRs that are useful to discriminate them from DMRs.…”

“…A hybrid DL-ML approach that has previously shown success at predicting DMRs [ 18 ] was used to identify core sets of DMRs per exposure and then unique DMRs within these core sets. Analysis shows that there are unique DMRs associated with each exposure, and the exposure-specific models are a better solution to identify these unique DMRs.…”

“…Results in previous work show that the hybrid model has high accuracy on the data constructed from nine different exposures [ 18 ]. However, training only one model on DMRs from all nine exposures results in high variance and large numbers of predicted DMRs The actual number of DMRs is likely fewer than the number predicted.…”

“…Previous work [ 18 ] shows that a hybrid deep machine learning (DL-ML) model can accurately predict a DNA region’s likelihood to be differentially methylated (DMR) as a result of ancestral exposure to nine environmental toxicants: atrazine [ 11 ], DDT [ 19 ], glyphosate [ 20 ], vinclozolin [ 21 ], pesticides permethrin and N, N-diethyl-meta-toluamide [ 22 ], dioxin [ 23 ], jet fuel [ 24 ], methoxychlor [ 25 ], and plastics bisphenol A and phthalates [ 26 ]. The hybrid DL-ML model (see Fig.…”

“…However, learning a model to predict DMRs across all exposures can cause over-generalization [ 18 ]. One approach to address over-generalization is to determine a core set of predictions, which is the intersection of the predictions made by several trained models, each randomly initialized.…”