MiREx: mRNA levels prediction from gene sequence and miRNA target knowledge

Area of Science:

• Genomics
• Computational Biology
• Molecular Biology

Background:

• Messenger RNA (mRNA) expression is vital for protein production and gene regulation.
• Existing models predict mRNA levels using DNA sequences, gene features, or transcriptional regulators like transcription factors (TFs) and microRNAs (miRNAs).
• Convolutional neural network (CNN) models like Xpresso have advanced prediction using promoter and mRNA half-life features.

Purpose of the Study:

• To develop an improved mRNA expression level prediction tool by incorporating miRNA target and expression data.
• To enhance the accuracy of mRNA expression prediction by leveraging the regulatory influence of specific miRNAs.

Main Methods:

• A CNN-based tool, miREx, was developed.
• The model integrates information on selected miRNA targets and their expression levels.
• Evaluation was performed on primary cancer site data from the Genomics Data Commons (GDC) database (lung, kidney, breast, corpus uteri).

Main Results:

• The miREx model demonstrated that incorporating specific miRNA targets and expression data significantly benefits mRNA level prediction.
• The inclusion of miRNA information improved the predictive accuracy compared to models without this data.
• The model's performance was validated across diverse cancer types.

Conclusions:

• miREx advances mRNA expression prediction by effectively utilizing miRNA regulatory information.
• The findings highlight the importance of considering miRNA-gene interactions for accurate gene expression modeling.
• Future work may involve integrating other regulatory factors or proteomics data.