Small RNA sequencing (sRNA-seq), long used for profiling microRNAs, may also hold valuable information about messenger RNA expression. A new preprint in bioRxiv by a team of researchers from the University of Essex demonstrates that fragments of protein-coding transcripts found in sRNA-seq data can accurately reflect gene expression levels, even without matched total RNA-seq experiments.
Using human, mouse, and chicken datasets with paired total RNA-seq and sRNA-seq data, the team found that expression levels inferred from sRNA-seq correlated strongly with those from conventional RNA-seq, with coefficients of determination (R²) ranging from 0.33 to 0.76. The correlations were particularly robust in neural tissues. Importantly, transcript half-life and the presence of highly abundant housekeeping genes did not significantly bias results.
The researchers processed reads with cutadapt, HISAT2, and featureCounts, followed by voom transformation and DESeq2 analysis. They showed that fragments 25 nucleotides or longer, sequenced at a depth of at least five million reads, were sufficient for accurate dual profiling of microRNAs and coding genes. Functional enrichment analysis confirmed that the inferred gene expression captured biologically relevant tissue-specific pathways.
Applying the method to breast cancer datasets containing only sRNA-seq data, the team successfully detected coding genes linked to tumor biology. Validation using qPCR-confirmed miR-10b-5p targets yielded 75% recall and 64% accuracy. These findings indicate that mRNA fragments within small RNA libraries can provide reliable estimates of coding gene activity, enabling reanalysis of archived sRNA-seq studies where total RNA data are unavailable.
Publication Details
Azadova A, Ekperuoh A, Brooke GN, Marco A. Analysis of coding gene expression from small RNA sequencing. bioRxiv 2024.06.21.600062; doi: https://doi.org/10.1101/2024.06.21.600062
Using human, mouse, and chicken datasets with paired total RNA-seq and sRNA-seq data, the team found that expression levels inferred from sRNA-seq correlated strongly with those from conventional RNA-seq, with coefficients of determination (R²) ranging from 0.33 to 0.76. The correlations were particularly robust in neural tissues. Importantly, transcript half-life and the presence of highly abundant housekeeping genes did not significantly bias results.
The researchers processed reads with cutadapt, HISAT2, and featureCounts, followed by voom transformation and DESeq2 analysis. They showed that fragments 25 nucleotides or longer, sequenced at a depth of at least five million reads, were sufficient for accurate dual profiling of microRNAs and coding genes. Functional enrichment analysis confirmed that the inferred gene expression captured biologically relevant tissue-specific pathways.
Applying the method to breast cancer datasets containing only sRNA-seq data, the team successfully detected coding genes linked to tumor biology. Validation using qPCR-confirmed miR-10b-5p targets yielded 75% recall and 64% accuracy. These findings indicate that mRNA fragments within small RNA libraries can provide reliable estimates of coding gene activity, enabling reanalysis of archived sRNA-seq studies where total RNA data are unavailable.
Publication Details
Azadova A, Ekperuoh A, Brooke GN, Marco A. Analysis of coding gene expression from small RNA sequencing. bioRxiv 2024.06.21.600062; doi: https://doi.org/10.1101/2024.06.21.600062