Taking promoters out of enhancers in sequence based predictions of tissue-specific mammalian enhancers

Abstract

BackgroundMany genetic diseases are caused by mutations in non-coding regions of the genome. These mutations are frequently found in enhancer sequences, causing disruption to the regulatory program of the cell. Enhancers are short regulatory sequences in the non-coding part of the genome that are essential for the proper regulation of transcription. While the experimental methods for identification of such sequences are improving every year, our understanding of the rules behind the enhancer activity has not progressed much in the last decade. This is especially true in case of tissue-specific enhancers, where there are clear problems in predicting specificity of enhancer activity.ResultsWe show a random-forest based machine learning approach capable of matching the performance of the current state-of-the-art methods for enhancer prediction. Then we show that it is, similarly to other published methods, frequently cross-predicting enhancers as active in different tissues, making it less useful for predicting tissue specific activity. Then we proceed to show that the problem is related to the fact that the enhancer predicting models exhibit a bias towards predicting gene promoters as active enhancers. Then we show that using a two-step classifier can lead to lower cross-prediction between tissues.ConclusionsWe provide whole-genome predictions of human heart and brain enhancers obtained with two-step classifier.Electronic supplementary materialThe online version of this article (doi:10.1186/s12920-017-0264-3) contains supplementary material, which is available to authorized users.

PrESOgenesis: A two-layer multi-label predictor for identifying fertility-related proteins using support vector machine and pseudo amino acid composition approach

“…In this study, a two-layer classifier was proposed for predicting fertility-related protein. The high efficiency of this method has been reported in previous studies such as predicting membrane proteins 92 , enhancer prediction 93 , remote protein homology detection 94 , identifying piwi-interacting RNAs 34 and miRNA Drosha processing site detection 95 .…”

K-mer Content Changes with Node Degree in Promoter–Enhancer Network of Mouse ES Cells

“…For straightforward comparison and interpretation of k-mer count vectors, we standardized the length of both types of regulatory regions at 1 kb, centered at the TSS or at the middle genomic position of the enhancer. Guided by our previous experience [ 8 ], we used k-mers of the length 4 because the shorter k-mers are less informative, while the longer are represented in too few data elements, given the data size. The steps of our sequence and graph analysis are illustrated in Figure S1 .…”

“…We assessed the sequence composition and the sequence similarity of interacting regions through analysis of their DNA sequence k-mer content, for k = 4, with selected results confirmed for k = 1 and k = 2. The k-mer content was used before for identifying enhancers, for distinguishing enhancers from promoters and distinguishing among enhancers with different tissue specificity [ 7 , 8 , 9 ]. K-mers, or a related operation of sequence convolution, were also used for prediction of individual promoter–enhancer interactions [ 10 , 11 , 12 ].…”

In Silico Analysis of Gene Expression Change Associated with Copy Number of Enhancers in Pancreatic Adenocarcinoma

“…Enhancers are DNA elements of up to 50–1500 base pairs (bp) [13]. They interact with their target promoters irrespective of their position to regulate downstream gene expression [14].…”