Leveraging Machine Learning for Predicting Circadian Transcription in mRNAs and lncRNAs

Area of Science:

• Genetics
• Molecular Biology
• Computational Biology

Background:

• The circadian clock orchestrates daily rhythms in mammalian gene expression.
• Transcriptional regulation of rhythmic messenger RNAs (mRNAs) is well-understood.
• Regulatory mechanisms for rhythmic long non-coding RNAs (lncRNAs) are largely unknown.

Purpose of the Study:

• To investigate and compare the regulatory mechanisms of rhythmic lncRNA transcription with those of mRNAs.
• To identify DNA sequence features influencing rhythmic transcription in promoters.

Main Methods:

• Applied machine learning models to predict rhythmic transcription patterns.
• Utilized k-mer-based DNA sequence features from promoter regions.
• Trained models on mRNA data and tested on lncRNA data, and vice versa.
• Employed SHAP analysis to identify critical DNA features.

Main Results:

• Demonstrated significant differences in regulatory mechanisms between rhythmic mRNA and lncRNA transcription.
• Identified key DNA sequence features driving rhythmic transcription for both mRNA and lncRNA.
• Showcased the effectiveness of machine learning in predicting rhythmic gene expression from sequence data.

Conclusions:

• Regulatory mechanisms for rhythmic lncRNA transcription differ from those of mRNAs.
• Specific DNA sequence features are critical for rhythmic RNA transcription.
• Machine learning provides a powerful tool for understanding gene expression regulation.