Prediction of the RBP binding sites on lncRNAs using the high-order nucleotide encoding convolutional neural network

Zhang, Shaowu; Ya, Wang; Zhang, Xi-Xi; Wang, Jiaqi

doi:10.1016/j.ab.2019.113364

Cited by 20 publications

(18 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…iDeepS uses a combination of CNN and biLSTM to predict RBPbinding sites from RNA sequences and their estimated secondary structures (Pan et al, 2018). HOCNNLB is another method for training CNNs to predict RBP binding sites while taking k-mer representations of RNA sequences (Zhang et al, 2019). In addition to the above models, the baseline BERT model (BERT-baseline), whose parameters were randomly initialized instead of transferring parameters from DNABERT, was also trained.…”

Section: Baseline Modelsmentioning

confidence: 99%

“…iDeepS uses a combination of CNN and biLSTM to predict RBP-binding sites from RNA sequences and their estimated secondary structures (Pan et al, 2018). HOCNNLB is another method for training CNNs to predict RBP binding sites while taking k-mer representations of RNA sequences (Zhang et al, 2019). DeepCLIP consists of one convolutional layer that extracts sequence features and following biLSTM to detect RBP binding sites (Grønning et al, 2020).…”

Section: Baseline Modelsmentioning

confidence: 99%

“…Since Alipanahi et al (2015) demonstrated the applicability of convolutional neural networks (CNNs) for predicting RNA-protein and DNA-protein interactions, several deep learning models have been developed. While some models incorporate a single CNN with some modifications (Pan and Shen, 2018;Zhang et al, 2019;Tahir et al, 2021), others use a different neural network model (Uhl et al, 2020) or a combination of several neural network architectures (Ben-Bassat et al, 2018;Pan et al, 2018;Yan et al, 2020;Deng et al, 2020;Grønning et al, 2020). For instance, HOCNNLB uses high-order encodings of RNA sequences as inputs for CNN (Zhang et al, 2019), and iDeepS uses stacked CNN and bidirectional long short-term memory (biLSTM) and takes both RNA sequences and their estimated secondary structures as inputs (Pan et al, 2018).…”

Section: Introductionmentioning

confidence: 99%

“…While some models incorporate a single CNN with some modifications (Pan and Shen, 2018;Zhang et al, 2019;Tahir et al, 2021), others use a different neural network model (Uhl et al, 2020) or a combination of several neural network architectures (Ben-Bassat et al, 2018;Pan et al, 2018;Yan et al, 2020;Deng et al, 2020;Grønning et al, 2020). For instance, HOCNNLB uses high-order encodings of RNA sequences as inputs for CNN (Zhang et al, 2019), and iDeepS uses stacked CNN and bidirectional long short-term memory (biLSTM) and takes both RNA sequences and their estimated secondary structures as inputs (Pan et al, 2018). However, existing models often lack interpretability because of the complex nature of the neural network and require additional information to RNA sequences.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Prediction of RNA-protein interactions using a nucleotide language model

Yamada

Hamada

2021

Preprint

View full text Add to dashboard Cite

Motivation:The accumulation of sequencing data has enabled researchers to predict the interactions between RNA sequencesand RNA-binding proteins (RBPs) using novel machine learning techniques. However, existing models are often difficult tointerpret and require additional information to sequences. Bidirectional encoder representations from Transformer (BERT)is a language-based deep learning model that is highly interpretable. Therefore, a model based on BERT architecture canpotentially overcome such limitations.Results:Here, we propose BERT-RBP as a model to predict RNA-RBP interactions by adapting the BERT architecture pre-trained on a human reference genome. Our model outperformed state-of-the-art prediction models using the eCLIP-seq data of154 RBPs. The detailed analysis further revealed that BERT-RBP could recognize both the transcript region type and RNAsecondary structure only from sequential information. Overall, the results provide insights into the fine-tuning mechanism ofBERT in biological contexts and provide evidence of the applicability of the model to other RNA-related problems.Availability:Python source codes are freely available at https://github.com/kkyamada/bert-rbp.

show abstract

Section: Baseline Modelsmentioning

confidence: 99%

Section: Baseline Modelsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Prediction of RNA-protein interactions using a nucleotide language model

Yamada

Hamada

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Deep learning for sequencing data has been used [11][12][13]; it has also been used in some works to predict sgRNA on-target and off-target efficiency. For example, a deep learning approach to predict off-target effects is described by Lin in [14].…”

Section: Related Workmentioning

confidence: 99%

CRISPRLearner: A Deep Learning-Based System to Predict CRISPR/Cas9 sgRNA On-Target Cleavage Efficiency

et al. 2019

View full text Add to dashboard Cite

CRISPRLearner, the system presented in this paper, makes it possible to predict the on-target cleavage efficiency (also called on-target knockout efficiency) of a given sgRNA sequence, specifying the target genome that this sequence is designed for. After efficiency prediction, the researcher can evaluate its sequence and design a new one if the predicted efficiency is low. CRISPRLearner uses a deep convolutional neural network to automatically learn sequence determinants and predict the efficiency, using pre-trained models or using a model trained on a custom dataset. The convolutional neural network uses linear regression to predict efficiency based on efficiencies used to train the model. Ten different models were trained using ten different gene datasets. The efficiency prediction task attained an average Spearman correlation higher than 0.40. This result was obtained using a data augmentation technique that generates mutations of a sgRNA sequence, maintaining the efficiency value. CRISPRLearner supports researchers in sgRNA design task, predicting a sgRNA on-target knockout efficiency.

show abstract

Predicting RNA Secondary Structure Using In Vitro and In Vivo Data

Ponti

Tartaglia

2021

Methods in Molecular Biology

View full text Add to dashboard Cite

Prediction of the RBP binding sites on lncRNAs using the high-order nucleotide encoding convolutional neural network

Cited by 20 publications

References 29 publications

Prediction of RNA-protein interactions using a nucleotide language model

Prediction of RNA-protein interactions using a nucleotide language model

CRISPRLearner: A Deep Learning-Based System to Predict CRISPR/Cas9 sgRNA On-Target Cleavage Efficiency

Predicting RNA Secondary Structure Using In Vitro and In Vivo Data

Contact Info

Product

Resources

About