Machine Learning Prediction of Non-Coding Variant Impact in Human Retinal cis-Regulatory Elements

Abstract

Purpose Prior studies have demonstrated the significance of specific cis -regulatory variants in retinal disease; however, determining the functional impact of regulatory variants remains a major challenge. In this study, we utilized a machine learning approach, trained on epigenomic data from the adult human retina, to systematically quantify the predicted impact of cis -regulatory variants. Methods We used human retinal DNA accessibility data (ATAC-seq) to determine a set of 18.9k high-confidence, putative cis -regulatory elements. Eighty percent of these elements were used to train a machine learning model utilizing a gapped k -mer support vector machine–based approach. In silico saturation mutagenesis and variant scoring was applied to predict the functional impact of all potential single nucleotide variants within cis -regulatory elements. Impact scores were tested in a 20% hold-out dataset and compared to allele population frequency, phylogenetic conservation, transcription factor (TF) binding motifs, and existing massively parallel reporter assay data. Results We generated a model that distinguishes between human retinal regulatory elements and negative test sequences with 95% accuracy. Among a hold-out test set of 3.7k human retinal CREs, all possible single nucleotide variants were scored. Variants with negative impact scores correlated with higher phylogenetic conservation of the reference allele, disruption of predicted TF binding motifs, and massively parallel reporter expression. Conclusions We demonstrated the utility of human retinal epigenomic data to train a machine learning model for the purpose of predicting the impact of non-coding regulatory sequence variants. Our model accurately scored sequences and predicted putative transcription factor binding motifs. This approach has the potential to expedite the characterization of pathogenic non-coding sequence variants in the context of unexplained retinal disease. Translational Relevance This workflow and resulting dataset serve as a promising genomic tool to facilitate the clinical prioritization of functionally disruptive non-coding mutations in the retina.

Characterization of sequence determinants of enhancer function using natural genetic variation

“…Therefore, we applied a gapped k-mer SVM approach (gkm-SVM) to our datasets that has been optimized to detect k-mers of similar length to typical TF-binding motifs ( Ghandi et al, 2016 ). Support vector machine (SVM) algorithms have been utilized in a variety of contexts to perform classification of DNA sequences in a supervised manner ( Barozzi et al, 2014 ; Ghandi et al, 2016 ; Van den Bosch et al, 2022 ). It should be noted that k-mers identified by gkm-SVM are simply DNA sequences of k length that can discriminate two sets of input sequences, and do not necessarily correspond to TF-binding sites per se.…”

Active learning of enhancer and silencer regulatory grammar in photoreceptors

“…Machine learning presents an opportunity to train better models of cis-regulatory grammars, due to its power to discover predictive features in high dimensional data. Deep neural network models trained on large epigenomic datasets often predict TF binding and chromatin accessibility with high accuracy, and these models have revealed important contextual features of local DNA sequence that determine TF binding (25)(26)(27)(28)(29)(30)(31)(32). However, models trained on massively parallel reporter gene assays (MPRAs) to predict CRE activity (20,(33)(34)(35)(36)(37)(38)(39)(40) often perform less well than binding models, likely because the cis-regulatory grammars that govern activity depend on additional higher-order interactions between bound TFs and their associated co-factors (2,(4)(5)(6)41).…”

Prioritizing Cardiovascular Disease-Associated Variants Altering NKX2-5 Binding through an Integrative Computational Approach

“…One approach to address this challenge is implementing predictive models to identify variants that create or disrupt TF binding sites (TFBS). [23][24][25] Large-scale gapped k-mer (LS-GKM) support vector machine (SVM) predictive models can be trained to identify TFBS by using in vitro or in vivo DNA-binding data, such as chromatin immunoprecipitation followed by sequencing (ChIP-seq). LS-GKM-SVM models outperform traditional approaches, such as position weight matrix (PWM)-based methods, by considering complex sequence features like dinucleotide interactions, longer/gapped k-mers, and intracellular patterns.…”

“…[26][27][28][29] LS-GKM-SVM predictive models can be trained with ChIP-seq data from specific cell lines or tissue to integrate relevant epigenomic and regulatory context. 23 In this work, we present an integrative approach to prioritize functional non-coding variants that can contribute to the biology of CVDs. Using publicly accessible data from the GWAS catalog 30 , GTEx Portal 31 , ENCODE 32 , ChIP-Atlas 33 , and Remap 34 , we compiled a list of CVD-associated SNPs linked with a differentially expressed gene in cardiac tissue.…”

“…26–29 LS-GKM-SVM predictive models can be trained with ChIP-seq data from specific cell lines or tissue to integrate relevant epigenomic and regulatory context. 23…”