Meta‐analysis of massively parallel reporter assays enables prediction of regulatory function across cell types

Abstract

Deciphering the potential of noncoding loci to influence gene regulation has been the subject of intense research, with important implications in understanding genetic underpinnings of human diseases. Massively parallel reporter assays (MPRAs) can measure regulatory activity of thousands of DNA sequences and their variants in a single experiment. With increasing number of publically available MPRA data sets, one can now develop data‐driven models which, given a DNA sequence, predict its regulatory activity. Here, we performed a comprehensive meta‐analysis of several MPRA data sets in a variety of cellular contexts. We first applied an ensemble of methods to predict MPRA output in each context and observed that the most predictive features are consistent across data sets. We then demonstrate that predictive models trained in one cellular context can be used to predict MPRA output in another, with loss of accuracy attributed to cell‐type‐specific features. Finally, we show that our approach achieves top performance in the Fifth Critical Assessment of Genome Interpretation “Regulation Saturation” Challenge for predicting effects of single‐nucleotide variants. Overall, our analysis provides insights into how MPRA data can be leveraged to highlight functional regulatory regions throughout the genome and can guide effective design of future experiments by better prioritizing regions of interest.

Identification and Massively Parallel Characterization of Regulatory Elements Driving Neural Induction

“…The selected oligonucleotides were generated and cloned upstream of a minimal promoter (mP) and EGFP reporter gene into a lentivirus-based enhancer assay vector (Figure 2B) as previously described (Inoue et al, 2017). Although the sequences are assayed in the context of enhancer activity in this assay, previous work has shown that it also provides a good indication for promoter activity (Kreimer et al, 2017(Kreimer et al, , 2019Melnikov et al,2014). Each individual CRS was designed to be associated with 90 different 15-bp barcodes, thus allowing robust evaluation of the pertaining expression output and to correct for site of integration biases (Ashuach et al, 2019;Inoue et al, 2017).…”

Reports from the fifth edition of CAGI: The Critical Assessment of Genome Interpretation

“…There had also been only one previous expression regulatory variant challenge, in CAGI4 (Kreimer et al, ). CAGI5 has a new expression regulatory challenge (Shigaki et al, ), and there are also two participant papers (Dong & Boyle, ; Kreimer, Yan, Ahituv, & Yosef, ).…”

Transcriptional enhancers: from prediction to functional assessment on a genome-wide scale

“…Although H3K27ac has been the most widely used chromatin mark to identify enhancers, this feature was outperformed by fifteen different experimental and sequence features included in the meta-analysis. Nevertheless, overlap with H3K27ac signal demonstrated greater predictive value for enhancer activity than all other histone modification profiles tested, with the exception of H3K4me2 and pan-histone H2 acetylation (Kreimer et al 2019).…”

“…MPRA data can also be used to determine the sequence or chromatin features that best identify active enhancers. A recent meta-analysis combined the quantitative transcript data from five different MPRAs that each tested ~2000 putative enhancer sequences in a total of 6 different cell types (Kreimer et al 2019). This analysis compared 56 different chromatin features in overall predictive capacity and found DNase I signal and the number of different TFs bound by ChIP-seq within the sequence most accurately predicts enhancer activity (Kreimer et al 2019).…”

“…A recent meta-analysis combined the quantitative transcript data from five different MPRAs that each tested ~2000 putative enhancer sequences in a total of 6 different cell types (Kreimer et al 2019). This analysis compared 56 different chromatin features in overall predictive capacity and found DNase I signal and the number of different TFs bound by ChIP-seq within the sequence most accurately predicts enhancer activity (Kreimer et al 2019). Although H3K27ac has been the most widely used chromatin mark to identify enhancers, this feature was outperformed by fifteen different experimental and sequence features included in the meta-analysis.…”