Abstract:MicroRNAs (miRNAs) mediates gene expression regulation by targeting specific messenger RNAs (mRNAs) in the cytoplasm. They can function as both tumor suppressors and oncogenes depending on the specific miRNA and its target genes. Detecting miRNA-target interactions (MTIs) is critical for unraveling the complex mechanisms of gene regulation and promising towards RNA therapy for cancer. There is currently a lack of MTIs prediction methods that simultaneously perform feature learning from heterogeneous gene regulatory network (GRN) and genomic sequences. To improve the prediction performance of MTIs, we present a novel transformer-based multi-view feature learning method -- MiRGraph, which consists of two main modules for learning the sequence-based and GRN-based feature embedding. For the former, we utilize the mature miRNA sequences and the complete 3'UTR sequence of the target mRNAs to encode sequence features using a hybrid transformer and convolutional neural network (CNN) (TransCNN) architecture. For the latter, we utilize a heterogeneous graph transformer (HGT) module to extract the relational and structural information from the GRN consisting of miRNA-miRNA, gene-gene and miRNA-target interactions. The TransCNN and HGT modules can be learned end-to-end to predict experimentally validated MTIs from MiRTarBase. MiRGraph outperforms existing methods in not only recapitulating the true MTIs but also in predicting strength of the MTIs based on the in-vitro measurements of miRNA transfections. In a case study on breast cancer, we identified plausible target genes of an oncomir.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to accurately predict the interactions between microRNAs (miRNAs) and target genes (MTIs), namely microRNA - target interactions. This problem is of great significance in revealing the regulatory mechanisms of complex diseases and designing miRNA - based treatment regimens. ### Problem Background - **The Role of miRNA**: miRNAs regulate gene expression by binding to the 3' untranslated region (3'UTR) of mRNA, inhibiting the translation of mRNA or causing its degradation. They can act as either tumor suppressors or oncogenes. - **The Importance of MTI Prediction**: Detecting the interactions between miRNAs and target genes is crucial for understanding gene regulatory mechanisms and developing RNA therapies for cancer. - **Limitations of Existing Methods**: - **Experimental Methods**: Traditional experimental methods (such as HITS - CLIP, CLASH, etc.) are costly and are biased towards strong - binding MTIs. - **Computational Methods**: Existing computational methods are divided into rule - based methods and machine - learning - based methods. The former rely on manually - designed features, resulting in inconsistent predictions and a high false - positive rate; the latter, although using deep - learning models, mostly rely only on sequence information and fail to fully utilize the information of miRNA - mediated gene regulatory networks (GRNs). ### The Solution Proposed in the Paper To overcome the above limitations, the paper proposes a new hybrid deep - learning framework - MiRGraph. This framework aims to simultaneously learn features from heterogeneous gene regulatory networks (GRNs) and genomic sequences to improve the accuracy of MTI prediction. ### Main Contributions of MiRGraph 1. **Sequence Feature Learning**: - Using the complete mature miRNA sequence and the complete 3'UTR sequence of the target mRNA, a hybrid architecture (TransCNN) that combines a convolutional neural network (CNN) and a Transformer is designed to encode sequence features. 2. **Network Feature Learning**: - A comprehensive heterogeneous graph network containing miRNA - miRNA, gene - gene, and miRNA - target - gene interactions is constructed, and the heterogeneous graph Transformer (HGT) module is used to extract relationship and structural information. ### Experimental Results - **Performance Improvement**: MiRGraph significantly outperforms existing methods on multiple evaluation metrics, especially performing best in terms of AUROC and AUPR. - **Generalization Ability**: Even in more challenging scenarios (such as predicting MTIs of unseen miRNAs), MiRGraph still performs excellently. - **Specific Application**: In a breast cancer case study, MiRGraph successfully identified potential target genes, verifying its effectiveness in practical applications. Through these improvements, MiRGraph provides a powerful tool for more accurately predicting the interactions between miRNAs and target genes, which helps to deepen the understanding of gene regulatory mechanisms and promotes the development of miRNA - based therapies.

MiRGraph: A hybrid deep learning approach to identify microRNA-target interactions by integrating heterogeneous regulatory network and genomic sequences

Gra-CRC-miRTar: The pre-trained nucleotide-to-graph neural networks to identify potential miRNA targets in colorectal cancer

DCTGM: A Novel Dual-channel Transformer Graph Model for Mirna-Disease Association Prediction

MiRTDL: A Deep Learning Approach for Mirna Target Prediction

LncMirNet: Predicting LncRNA–miRNA Interaction Based on Deep Learning of Ribonucleic Acid Sequences

Multi-view graph neural network with cascaded attention for LncRNA-miRNA interaction prediction

HGTMDA: A Hypergraph Learning Approach with Improved GCN-Transformer for miRNA-Disease Association Prediction

ncRNAInter: a novel strategy based on graph neural network to discover interactions between lncRNA and miRNA

Integrative Approaches for Microrna Target Prediction: Combining Sequence Information and the Paired Mrna and Mirna Expression Profiles

Predicting the potential human lncRNA–miRNA interactions based on graph convolution network with conditional random field

MHGTMDA: Molecular heterogeneous graph transformer based on biological entity graph for miRNA-disease associations prediction

Sequence pre-training-based graph neural network for predicting lncRNA-miRNA associations

A method for miRNA diffusion association prediction using machine learning decoding of multi-level heterogeneous graph Transformer encoded representations

Predicting miRNA-Disease Associations Based on Heterogeneous Graph Attention Networks

Predicting miRNA–Disease Associations by Combining Graph and Hypergraph Convolutional Network

Predicting MicroRNA Targets by Integrating Sequence and Expression Data in Cancer

Predicting miRNA-disease associations based on graph attention network with multi-source information

Predicting Mirna Targets By Integrating Gene Regulatory Knowledge With Expression Profiles

Graph convolution for predicting associations between miRNA and drug resistance

Predicting miRNA-disease association via graph attention learning and multiplex adaptive modality fusion