Deep learning architectures for prediction of nucleosome positioning from sequences data

Abstract

BackgroundNucleosomes are DNA-histone complex, each wrapping about 150 pairs of double-stranded DNA. Their function is fundamental for one of the primary functions of Chromatin i.e. packing the DNA into the nucleus of the Eukaryote cells. Several biological studies have shown that the nucleosome positioning influences the regulation of cell type-specific gene activities. Moreover, computational studies have shown evidence of sequence specificity concerning the DNA fragment wrapped into nucleosomes, clearly underlined by the organization of particular DNA substrings. As the main consequence, the identification of nucleosomes on a genomic scale has been successfully performed by computational methods using a sequence features representation.ResultsIn this work, we propose a deep learning model for nucleosome identification. Our model stacks convolutional layers and Long Short-term Memories to automatically extract features from short- and long-range dependencies in a sequence. Using this model we are able to avoid the feature extraction and selection steps while improving the classification performances.ConclusionsResults computed on eleven data sets of five different organisms, from Yeast to Human, show the superiority of the proposed method with respect to the state of the art recently presented in the literature.

CORENup: a combination of convolutional and recurrent deep neural networks for nucleosome positioning identification

“…The output of the network is calculated by using the following equation: where the symbol ⊙ indicates the multiplication element by element. In a preceding work [ 26 ] we found that substituting a convolutional layer with an LSTM layer can improve the performance in sequence classification tasks, and the resulting network is in Fig. 3 .…”

“…We also calculated the Roc Curve to compare the prediction performance of the four methods and we reported the Area Under the Curve (AUC). The first set of results is related to the HS, DM, E and Y datasets, and in Table 4 the results obtained from two models used in our recent works, indicated as simple LSTM [ 26 ] and ConvNet [ 27 ] are reported, together with the CORENup and the LeNup networks. The two simpler networks have almost always worst performance than the more complex networks, confirming that with more parameters it is possible to improve the performance of the classifier.…”

“…NucMap is an online database which includes 798 experimental data from 477 samples across 15 species, also supplying a set of very useful tools to visualize and compare the data. These high-resolution data leads to the development of many computational methodologies able to successfully process sequence information to predict the nucleosome presence[ 17 , 23 – 26 ]. Taking into consideration these biological studies, and differently from other studies proposed so far, this work intends to try to understand at which extent the DNA sequence is solely responsible for nucleosome positioning.…”

“…This investigation is carried out by a machine learning model, in particular a deep neural network, which processes only sequence information. In the past years, we gave several contributions to the study of deep learning networks for sequence classification [ 26 – 30 ]. In this work we present CORENup , an extension of our previous models by the integration of two different neural networks, each specialized in extracting specific features from sequences, i.e.…”

“…the representation for the input data involve a specific preprocessing stage. Conversely, other authors have used the simple one hot representation obtaining greater accuracy, using deep neural network classifiers [ 25 , 26 , 28 ]. Actually the model called Le-Nup [ 25 ] is the top performer.…”