Identification of the human DPR core promoter element using machine learning

Abstract

The RNA polymerase II (Pol II) core promoter is the strategic site of convergence of the signals that lead to transcription initiation 1 - 5 , but the downstream core promoter in humans has been difficult to decipher 1 - 3 . Here, we analyze the human Pol II core promoter and use machine learning to generate predictive models for the downstream core promoter region (DPR) and the TATA box. We developed a method termed HARPE (high-throughput analysis of randomized promoter elements) to create hundreds of thousands of DPR (or TATA box) variants that are each of known transcriptional strength. We then analyzed the HARPE data by support vector regression (SVR) to provide comprehensive models for the sequence motifs, and found that the SVR-based approach is more effective than a consensus-based method for predicting transcriptional activity. These studies revealed that the DPR is a functionally important core promoter element that is widely used in human promoters. Importantly, there appears to be a duality between the DPR and TATA box, as many promoters contain one or the other element. More broadly, these findings show that functional DNA motifs can be identified by machine learning analysis of a comprehensive set of sequence variants.

Conserved DNA sequence features underlie pervasive RNA polymerase pausing

“…The latter was critical to minimize sequence biases originating from the high GC content in promoter-proximal gene regions ( 85 ). Interestingly, the promoter-proximal pause motif shows similarities to the consensus sequence of the following core promoter elements located downstream of the TSS ( 86 , 87 ): the downstream core promoter element (DPE) ( 88 ), the downstream core element (DCE) ( 89 , 90 ) and the recently uncovered human DPR core promoter element ( 91 ) ( Supplementary Figure S6B ). The region where these cis -elements are located strongly overlap with the region where promoter-proximal pausing of Pol II usually occurs.…”

Computational identification and experimental characterization of preferred downstream positions in human core promoters

“…Moreover, the experiments were performed in two different cell lines, using different readout as the outcome, either the indirect luciferase reporter activity (this study) or the RNA output itself, using either RNAseq or primer extension analysis [55]. Surprisingly, the two independent approaches identified functional downstream positions/region within the ANP32E promoter.…”

“…During the preparation of the manuscript, we became aware of a comprehensive work from the Kadonaga lab [55], which used machine learning to generate predictive models to analyze human Pol II core promoters and identified a downstream promoter region (DPR) spanning from +17 to +35, which contributes to the transcriptional output of a fraction of human promoters. Reassuringly, the positions identified in our study highly match specific positions within the DPR identified by the Kadonaga lab, which supports the concept of a single functional downstream unit [30, 52].…”

“…Moreover, different approaches to identify the important downstream positions were taken; while we started from bioinformatics analysis and then tested naturally-occurring minimal promoters, the Kadonaga lab has first used massively parallel reporter assays (MPRA) of an extensive library composed of randomized version of the downstream region, using a specific promoter backbone. Moreover, the experiments were performed in two different cell lines, using different readout as the outcome, either the indirect luciferase reporter activity (this study) or the RNA output itself, using either RNAseq or primer extension analysis [55]. Surprisingly, the two independent approaches identified functional downstream positions/region within the ANP32E promoter.…”

“…Surprisingly, the two independent approaches identified functional downstream positions/region within the ANP32E promoter. Moreover, we ran the support vector regression (SVRb) model that was generated using in vitro transcription [55] on the +17 to +35 sequences of the wt and mutant promoters identified using the EM algorithm (S2 Table ). Overall, our computational model was successful in making similar predictions (correlation coefficient ~0.75) as the SVRb model that used experimentally-based training data.…”