A deep recurrent neural network discovers complex biological rules to decipher RNA protein-coding potential

Hill, Steven T.; Kuintzle, Rachael; Teegarden, A.T.; Merrill, Erich; Danaee, Padideh; Hendrix, David A.

doi:10.1093/nar/gky567

Cited by 79 publications

(74 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…al. approach and mapped an input sequence of nucleotides to a sequence of vectors using an embedding layer (Hill et al, 2018). This sequence of vectors was then fed into a bidirectional recurrent neural network ( Fig.…”

Section: Classifiersmentioning

confidence: 99%

“…Convolutional neural networks apply small kernels in a sliding window across input sequences and have been shown to learn meaningful motif "features" from sequence data (Min et al, 2017). In fact, deep learning sequence based techniques have been applied to a variety of tasks including prediction of enhancers (Min et al, 2017), accessibility of DNA sequences (Kelley et al, 2016), the effect of non-coding variants (Zhou and Troyanskaya, 2015), sequence specificity of binding proteins (Alipanahi et al, 2015), and RNA protein coding potential (Hill et al, 2018). Often in these models, nucleotides are represented using one-hot encoded vectors.…”

Section: Introductionmentioning

confidence: 99%

“…Recurrent neural networks have also been shown to be able to learn from sequence data, owing to their ability to learn long term relationships and patterns. While it is possible to learn using the aforementioned one-hot encoding of input sequences, Hill et al (Hill et al, 2018) found that use of an embedding input layer mapping individual nucleotides to a learned dense vector representation can lead to increased prediction performance. Furthermore, they found that certain nucleotides such as C and U, or C and G had lower cosine distances in the embedding vector space than between other pairs of nucleotides, showing that the learned embeddings held some biologically relevant meaning.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

ATAC-seq signal processing and recurrent neural networks can identify RNA polymerase activity

Tripodi¹,

Chowdhury²,

Dowell

2019

Preprint

View full text Add to dashboard Cite

Nascent transcription assays are the current gold standard for identifying regions of active transcription, including markers for functional transcription factor (TF) binding. Here we present a signal processingbased model to determine regions of active transcription genome-wide using the simpler assay for transposase-accessible chromatin, followed by high-throughput sequencing (ATAC-seq). The focus of this study is twofold: First, we perform a frequency space analysis of the "signal" generated from ATAC-seq experiments' short reads, at a single-nucleotide resolution, using a discrete wavelet transform. Second, we explore different uses of neural networks to combine this signal with its underlying genome sequence in order to classify ATAC-seq peaks on the presence or absence of bidirectional transcription. We analyze the performance of different data encoding schemes and machine learning architectures, and show how a hybrid signal/sequence representation classified using recurrent neural networks (RNNs) yields the best performance across different cell types.

show abstract

Section: Classifiersmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

ATAC-seq signal processing and recurrent neural networks can identify RNA polymerase activity

Tripodi¹,

Chowdhury²,

Dowell

2019

Preprint

View full text Add to dashboard Cite

show abstract

“…develop a model of mRNN based on Recurrent Neural Networks (RNN) , and they predict the coding potential of sORF (Hill, Kuintzle, Teegarden, Merrill, Danaee and Hendrix 2018). In November 2019, Zhu et al also develop a tool specifically for predicting micropeptides, MiPepid, which is a tool developed by using K-mer features and machine learning (Zhu and Gribskov 2019).…”

Section: Introductionmentioning

confidence: 99%

CPPred-sORF: Coding Potential Prediction of sORF based on non-AUG

Tong

Xie

et al. 2020

Preprint

View full text Add to dashboard Cite

In recent years, researchers have discovered thousands of sORFs that can encode micropeptides, and more and more discoveries that non-AUG codons can be used as translation initiation sites for these micropeptides. On the basis of our previous tool CPPred, we develop CPPred-sORF by adding two features and using non-AUG as the starting codon, which makes a comprehensive evaluation of sORF. The database of CPPred-sORF are constructed by small coding RNA and lncRNA as positive and negative data, respectively. Compared to the small coding RNAs and small ncRNAs, lncRNAs and small coding RNAs are less distinguishable. This is because the longer the sequences, the easier to include open reading frames. We find that the sensitivity, specificity and MCC value of CPPred-sORF on the independent testing set can reach 88.22%, 88.84% and 0.768, respectively, which shows much better prediction performance than the other methods.The so-called ncRNAs must not be coding, but it is not the truth. A large number of research reports that ncRNAs may contain small open reading frames (small ORFs, sORFs) that can encode micropeptides (

show abstract

“…Deep learning allows computational models composed of multiple processing layers to learn representations of data with multiple levels of abstraction 22 . Deep learning models can identify dependencies and complex structures in very highdimensional data -such as CLIP data -and have been used, for example, for predicting the effects of mutations in non-coding DNA on gene expression and disease 1,23 , predicting DNA function 24 , mRNA coding potential 25 , and prediction of subcellular locations of proteins 26 . Starting with DeepBind 27 , which uses convolutional neural networks (CNN) to classify bound sequences and nonbound, neural networks have also been used to directly model RBP preferences.…”

Section: Introductionmentioning

confidence: 99%

DeepCLIP: Predicting the effect of mutations on protein-RNA binding with Deep Learning

Groenning

Doktor

Larsen

et al. 2019

Preprint

View full text Add to dashboard Cite

Nucleotide variants can cause functional changes by altering protein-RNA binding in various and subtle ways that are not easy to predict. This can affect processes such as splicing, nuclear shuttling, and stability of the transcript. Therefore, correct modelling of protein-RNA binding is critical when predicting the effects of sequence variations. Many RNA-binding proteins recognize a diverse set of motifs and binding is typically also dependent on the genomic context, making this task particularly challenging. Although existing protein binding site models incorporate various additional data sources to incorporate context, such as RNA structure and functional gene context, they still need improvement and they have not been developed to predict the effect of sequence variants. Here, we present DeepCLIP, the first method for context-aware modeling and predicting protein binding to nucleic acids using exclusively sequence data as input. We show that DeepCLIP outperforms existing methods for modelling RNA-protein binding. Importantly, we demonstrate that DeepCLIP is able to reliably predict the functional effects of contextually dependent nucleotide variants in independent wet lab experiments. Furthermore, we show how DeepCLIP binding profiles can be used in the design of therapeutically relevant antisense oligonucleotides, and to uncover possible position-dependent regulation in a tissue-specific manner. DeepCLIP can be freely used at http://deepclip.compbio.sdu.dk. Highlights We have designed DeepCLIP as a simple neural network that requires only CLIP binding sites as input. The architecture and parameter settings of DeepCLIP makes it an efficient classifier and robust to train, making high performing models easy to train and recreate. Using an extensive benchmark dataset, we demonstrate that DeepCLIP outperforms existing tools in classification. Furthermore, DeepCLIP provides direct information about the neural network's decision process through visualization of binding motifs and a binding profile that directly indicates sequence elements contributing to the classification. To show that DeepCLIP models generalize to different datasets we have demonstrated that predictions correlate with in vivo and in vitro experiments using quantitative binding assays and minigenes. Identifying the binding sites for regulatory RNA-binding proteins is fundamental for efficient design of (therapeutic) antisense oligonucleotides. Employing a reported disease associated mutation, we demonstrate that DeepCLIP can be used for design of therapeutic antisense oligonucleotides that block regions important for binding of regulatory proteins and correct aberrant splicing. Using DeepCLIP binding profiles, we uncovered a possible position-dependent mechanism behind the reported tissue-specificity of a group of TDP-43 repressed pseudoexons. We have made DeepCLIP available as an online tool for training and application of protein-RNA binding deep learning models and prediction of the potential effects of clinically detected sequence vari...

show abstract

A deep recurrent neural network discovers complex biological rules to decipher RNA protein-coding potential

Cited by 79 publications

References 22 publications

ATAC-seq signal processing and recurrent neural networks can identify RNA polymerase activity

ATAC-seq signal processing and recurrent neural networks can identify RNA polymerase activity

CPPred-sORF: Coding Potential Prediction of sORF based on non-AUG

DeepCLIP: Predicting the effect of mutations on protein-RNA binding with Deep Learning

Contact Info

Product

Resources

About