CFA: An explainable deep learning model for annotating the transcriptional roles of cis-regulatory modules based on epigenetic codes

Yang, Tzu-Hsien; Yu, Yu-Huai; Wu, Sheng-Hang; Zhang, Fang‐Yuan

doi:10.1016/j.compbiomed.2022.106375

Cited by 6 publications

(11 citation statements)

References 56 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Next, we investigated the possible reasons for the superiority of our designed interaction-aware model. In deep learning models, governing recognition patterns can be extracted from the data using the most discriminating input elements during model classification. ,, For this purpose, we first utilized the integrated gradient method from explainable artificial intelligence (XAI) to calculate the base-by-base importance enforced by the designed network. In addition, the miRNA sequences and target segments were approximately complement-aligned using RNAup .…”

Section: Results and Discussionmentioning

confidence: 99%

“…We also generated ROC (receiver operating characteristic) curves and calculated the area under the ROC curve (auROC) to estimate the intrinsic performance of different models. 31 In brief, an ROC curve plots the recall values to the corresponding FPR values while varying the prediction thresholds. A high auROC value indicates that the model is better at retaining high recalls when the corresponding false-positive rates are tightly controlled.…”

Section: ■ Data and Methodsmentioning

confidence: 99%

“…One-dimensional (1D) convolution operations have previously been shown to help extract sequence features from nucleotide one-hot encodings. , In this network, the miRNA and mRNA sequence inputs are designed to be padded to 30 and 40 nt, respectively, using the notation “L”. The deep learning model will then encode the “L” notation into zeros.…”

Section: Methodsmentioning

confidence: 99%

“…In the final constructed tool of this interaction-aware model, the probability p > 0.5 was chosen as the binding event prediction threshold. We also generated ROC (receiver operating characteristic) curves and calculated the area under the ROC curve (auROC) to estimate the intrinsic performance of different models . In brief, an ROC curve plots the recall values to the corresponding FPR values while varying the prediction thresholds.…”

Section: Methodsmentioning

confidence: 99%

See 3 more Smart Citations

Identifying Human miRNA Target Sites via Learning the Interaction Patterns between miRNA and mRNA Segments

Yang,

Chen,

Lee

et al. 2023

J. Chem. Inf. Model.

Self Cite

View full text Add to dashboard Cite

miRNAs (microRNAs) target specific mRNA (messenger RNA) sites to regulate their translation expression. Although miRNA targeting can rely on seed region base pairing, animal miRNAs, including human miRNAs, typically cooperate with several cofactors, leading to various noncanonical pairing rules. Therefore, identifying the binding sites of animal miRNAs remains challenging. Because experiments for mapping miRNA targets are costly, computational methods are preferred for extracting potential miRNA–mRNA fragment binding pairs first. However, existing prediction tools can have significant false positives due to the prevalent noncanonical miRNA binding behaviors and the information-biased training negative sets that were used while constructing these tools. To overcome these obstacles, we first prepared an information-balanced miRNA binding pair ground-truth data set. A miRNA–mRNA interaction-aware model was then designed to help identify miRNA binding events. On the test set, our model (auROC = 94.4%) outperformed existing models by at least 2.8% in auROC. Furthermore, we showed that this model can suggest potential binding patterns for miRNA–mRNA sequence interacting pairs. Finally, we made the prepared data sets and the designed model available at .

show abstract

Section: Results and Discussionmentioning

confidence: 99%

Section: ■ Data and Methodsmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

See 2 more Smart Citations

Identifying Human miRNA Target Sites via Learning the Interaction Patterns between miRNA and mRNA Segments

Yang,

Chen,

Lee

et al. 2023

J. Chem. Inf. Model.

Self Cite

View full text Add to dashboard Cite

show abstract

“…In the above equation, the convolution operation conv(n,T) with n 1D kernels of size 3 on the l × c tensor T (using stride 2) is defined as previously suggested 32,33 (eq 3)…”

Section: ■ Introductionmentioning

confidence: 99%

DEBFold: Computational Identification of RNA Secondary Structures for Sequences across Structural Families Using Deep Learning

Yang

2024

J. Chem. Inf. Model.

Self Cite

View full text Add to dashboard Cite

It is now known that RNAs play more active roles in cellular pathways beyond simply serving as transcription templates. These biological mechanisms might be mediated by higher RNA stereo conformations, triggering the need to understand RNA secondary structures first. However, experimental protocols for solving RNA structures are unavailable for large-scale investigation due to their high costs and time-consuming nature. Various computational tools were thus developed to predict the RNA secondary structures from sequences. Recently, deep networks have been investigated to help predict RNA structures directly from their sequences. However, existing deep-learning-based tools are more or less suffering from model overfitting due to their complicated problem formulation and defective model training processes, limiting their applications across sequences from different structural families. In this research, we designed a two-stage RNA structure prediction strategy called DEBFold (deep ensemble boosting and folding) based on convolution encoding/decoding and self-attention mechanisms to enhance the existing thermodynamic structure models. Moreover, the model training process followed rigorous steps to achieve an acceptable prediction generalization. On the family-wise reserved test sets and the PDB-derived test set, DEBFold achieves better structure prediction performance over traditional tools and existing deep-learning methods. In summary, we obtained a cutting-edge deep-learning-based structure prediction tool with supreme across-family generalization performance. The DEBFold tool can be accessed at https://cobis.bme.ncku.edu.tw/DEBFold/.

show abstract

RDDL: A systematic ensemble pipeline tool that streamlines balancing training schemes to reduce the effects of data imbalance in rare-disease-related deep-learning applications

Yang

Liao

et al. 2023

Computational Biology and Chemistry

View full text Add to dashboard Cite

CFA: An explainable deep learning model for annotating the transcriptional roles of cis-regulatory modules based on epigenetic codes

Cited by 6 publications

References 56 publications

Identifying Human miRNA Target Sites via Learning the Interaction Patterns between miRNA and mRNA Segments

Identifying Human miRNA Target Sites via Learning the Interaction Patterns between miRNA and mRNA Segments

DEBFold: Computational Identification of RNA Secondary Structures for Sequences across Structural Families Using Deep Learning

RDDL: A systematic ensemble pipeline tool that streamlines balancing training schemes to reduce the effects of data imbalance in rare-disease-related deep-learning applications

Contact Info

Product

Resources

About