# Limitations of Single Prediction Tools in miRNA Profiling of Grapevine Viral Coinfection

## TL;DR

This study compares different tools for predicting miRNA expression in grapevines infected with multiple viruses and finds that combining tools improves results.

## Contribution

The study reveals how viral coinfections affect miRNA expression in grapevines and highlights the limitations of single prediction tools.

## Key findings

- Three miRNA prediction tools identified largely different sets of differentially expressed miRNAs.
- An integrated approach combining all three tools yielded more consistent and biologically meaningful results.
- Experimental validation confirmed the differential expression of several miRNAs.

## Abstract

Background/objectives: Grapevine (Vitis vinifera L.) is one of the most economically and culturally important fruit crops worldwide and hosts more than 100 viruses. Viral infections can cause severe yield losses, but plants can adapt to infection through changes in miRNA-mediated regulatory pathways. MicroRNAs are key regulators of plant development and stress responses. Several prediction tools are available for miRNA detection from small RNA sequencing data, each relying on different algorithms. The aim of this study was to compare miRNA predictions generated by three widely used tools (miRador, ShortStack, and miRDeep2) and to evaluate how viral coinfections influence miRNA expression in grapevine. Methods: Two grapevine cultivars, Refošk (“Terrano”) and Zeleni Sauvignon (“Sauvignon Vert”), were analyzed. Small RNA sequencing was performed on virus-free plants and plants coinfected with grapevine Pinot gris virus (GPGV), grapevine rupestris stem pitting-associated virus (GRSPaV), and grapevine rupestris vein feathering virus (GRVFV). Three miRNA prediction tools were used to identify miRNAs annotated in public databases. Differential expression analysis was performed separately for each tool and by using an integrated approach that combined all three datasets. The expression of selected miRNAs was further evaluated using stem-loop RT-qPCR. Results: The three prediction tools detected markedly different numbers of miRNAs, resulting in largely distinct sets of differentially expressed miRNAs and limited overlap between individual analyses. The integrated approach yielded a separate set of differentially expressed miRNAs, most of which overlapped with at least one individual dataset. Stem-loop RT-qPCR analysis supported the differential expression of several selected miRNAs. Conclusions: This study provides new insight into miRNA expression in grapevine under mixed-virus infection and demonstrates that miRNA profiling outcomes are strongly influenced by the choice of bioinformatic prediction tool. Our results highlight the importance of integrated analytical strategies combined with experimental validation to obtain robust and biologically meaningful interpretations of miRNA expression in plants.