Deciphering regulatory DNA sequences and noncoding genetic variants using neural network models of massively parallel reporter assays

Abstract

The relationship between noncoding DNA sequence and gene expression is not well-understood. Massively parallel reporter assays (MPRAs), which quantify the regulatory activity of large libraries of DNA sequences in parallel, are a powerful approach to characterize this relationship. We present MPRA-DragoNN, a convolutional neural network (CNN)-based framework to predict and interpret the regulatory activity of DNA sequences as measured by MPRAs. While our method is generally applicable to a variety of MPRA designs, here we trained our model on the Sharpr-MPRA dataset that measures the activity of ∼500,000 constructs tiling 15,720 regulatory regions in human K562 and HepG2 cell lines. MPRA-DragoNN predictions were moderately correlated (Spearman ρ = 0.28) with measured activity and were within range of replicate concordance of the assay. State-of-the-art model interpretation methods revealed high-resolution predictive regulatory sequence features that overlapped transcription factor (TF) binding motifs. We used the model to investigate the cell type and chromatin state preferences of predictive TF motifs. We explored the ability of our model to predict the allelic effects of regulatory variants in an independent MPRA experiment and fine map putative functional SNPs in loci associated with lipid traits. Our results suggest that interpretable deep learning models trained on MPRA data have the potential to reveal meaningful patterns in regulatory DNA sequences and prioritize regulatory genetic variants, especially as larger, higher-quality datasets are produced.

Cis and trans effects differentially contribute to the evolution of promoters and enhancers

“…Thus, while cis effects can generally be attributed to the disruption of strong activating motifs, in rarer cases, cis effects are due to the disruption of weak repressive motifs. While this may reflect real biological effects, it may also be due to the fact that MPRAs are more powered to detect activators over repressors [ 29 , 32 ].

Fig.

…”

Fully interpretable deep learning model of transcriptional control

“…Deep learning has been widely deployed in genomics and systems biology over the last few years ( Alipanahi et al , 2015 ; Avsec et al , 2019 ; Celesti et al , 2017 ; Cuperus et al , 2017 ; Greenside et al , 2018 ; Koh et al , 2017 ; Libbrecht and Noble, 2015 ; Movva et al , 2019 ; Nair et al , 2019 ; Pouladi et al , 2015 ; Rui et al , 2007 ; Shen et al , 2018 ). Many of the developed tools have been highly successful in classification problems such as the identification of binding sites, open regions of chromatin and the location of enhancers.…”

Sequence-to-function deep learning frameworks for engineered riboregulators

“…While recent work such as soft explainable decision trees 34 have enabled researchers to look inside this "black-box" by using a neural network to train a decision tree, we chose to visualize weights and activations of our trained model directly 35 . Taking inspiration from work in the fields of image recognition and genomics 23,[36][37][38] , we "un-boxed" the first convolutional layer to visualize the features our model deemed important by interpreting the filter weights learned from input sequences as sequence logos ( Fig. 2a).…”