Poly(A) tails are stretches of adenine nucleotides added to the end of messenger RNA (mRNA) molecules. These tails play a crucial role in determining how long RNA molecules survive in the cell, how efficiently they are translated into proteins, and which specific RNA isoforms are expressed. Because of their central role in RNA biology, accurately measuring poly(A) tail lengths is important for understanding gene expression, RNA stability, and overall transcriptome dynamics.

Researchers at the University of Melbourne used direct RNA sequencing from Oxford Nanopore Technologies to explore poly(A) tail lengths and compare different computational tools for estimating them. Direct RNA sequencing is a powerful approach that sequences RNA molecules in their full length, capturing the complete sequence of the RNA, including the poly(A) tail, without the need for conversion to DNA.

Poly(A) tail length estimates for each tool

(A, B) Predicted poly(A) tail length distributions for (A) R1 (30 nt) and (B) R2 (60 nt) Sequins. Outputs from BoostNano (yellow), Dorado (red), tailfindr (blue), and nanopolish (green). RNA002 (solid) and RNA004 (dashed) are shown as different line types. The x-axis shows the predicted poly(A) tail lengths of all reads, and the y-axis reveals the density of the poly(A) tail lengths. nt: nucleotides. Black vertical dashed lines indicate the known lengths. (C) RNA002 poly(A) tail estimates of eGFP synthetic RNA from the study by Krause et al. [42], ranging between 10 and 150 nt in poly(A) length. Red dashed lines indicate the corresponding known poly(A) lengths.

The team developed a new deep learning-based tool called BoostNano and benchmarked it against three existing poly(A) estimation tools: tailfindr, nanopolish, and Dorado. To test the accuracy of these tools, the researchers used synthetic RNA standards with known poly(A) tail lengths, including sequences with tails of 10 to 150 nucleotides. This approach allowed them to measure how well each tool could estimate the true tail length under controlled conditions.

The analysis revealed that tool performance depended on the RNA length and the dataset, with longer tails or complex sequences sometimes producing higher error rates. They also found that averaging measurements across multiple reads improved accuracy, highlighting the importance of robust data handling in RNA sequencing experiments. Among the tested tools, Dorado stood out as the most reliable option, offering fast runtimes, low mean error, and convenient integration with standard base-calling workflows.

These findings are significant because poly(A) tail length affects many aspects of RNA function, including transcript stability, translation efficiency, and gene regulation. By providing a thorough benchmarking of available tools, Chang and colleagues give the research community a practical reference for selecting the best approaches when analyzing RNA sequencing data. This work contributes to a more accurate understanding of RNA biology and provides guidance for studies linking poly(A) tail length to gene expression, transcript stability, and cellular regulation.

Availability – https://github.com/haotianteng/BoostNano