MicroRNAs are small RNA molecules that play a major role in controlling gene expression. By regulating how genes are translated into proteins, microRNAs influence many biological processes, including development, immune responses, cancer progression, and numerous other diseases.

As microRNA research has expanded, microRNA sequencing has become an important tool for measuring changes in microRNA expression. However, many of the statistical methods used to analyze microRNA sequencing data were originally developed for messenger RNA (mRNA) sequencing experiments. While both technologies generate sequencing data, microRNAs have unique biological and statistical characteristics that can make traditional mRNA analysis methods less reliable.

Researchers at the University of Rochester Medical Center have now developed a new analytical approach called Negative Binomial Softmax Regression, or NBSR, specifically designed for microRNA sequencing data.

The researchers found that applying conventional mRNA-based statistical methods to microRNA sequencing datasets can lead to elevated false discovery rates. In practical terms, this means researchers may incorrectly identify microRNAs as significantly different between experimental groups when they are not.

To address this challenge, the team developed NBSR, a statistical framework that accounts for the unique properties of microRNA expression data. One key feature of the method is its use of a measurement called the log relative abundance ratio, which provides a more interpretable way to compare microRNA expression levels between experimental conditions.

The new model also improves statistical power, increasing the likelihood of detecting real biological differences. At the same time, it generates narrower confidence intervals, giving researchers greater confidence in their results.

Another advantage of NBSR is its ability to handle highly variable and sparsely expressed microRNAs. These low-abundance microRNAs are often difficult to analyze accurately using existing methods, yet they may play important biological roles in disease and cellular regulation.

The researchers also demonstrated that the model can improve estimates of absolute expression changes by correcting bias in relative abundance measurements. This capability may be particularly useful when only a small number of microRNAs differ between experimental groups.

To evaluate performance, the team tested NBSR using both real-world and simulated datasets. Across these analyses, the method consistently improved sensitivity while maintaining more accurate statistical inference than traditional approaches.

As microRNA sequencing continues to be used for biomarker discovery, disease research, and therapeutic development, specialized analytical approaches will become increasingly important. The NBSR framework represents a significant step toward more accurate interpretation of microRNA sequencing data and may help researchers uncover biologically meaningful signals that might otherwise be missed.

By recognizing that microRNA datasets require different statistical treatment than messenger RNA datasets, this work provides a valuable new tool for the growing field of microRNA research.\

Availability – The NBSR implementation is available on https://github.com/junseonghwan/nbsr/ and scripts for generating the figures on https://github.com/junseonghwan/nbsr-experiments/.