Share
1. Oct. 2026
Hundreds of pollen grains in a single microscope image, thousands across an experiment, and each one needs to be identified and counted. For researchers studying plant reproduction, counting viable pollen provides clues about a plant’s ability to produce seeds. But obtaining those numbers can mean hours of repetitive work. Researchers at CEITEC Masaryk University have developed the Pollen Analysis Tool (PAT), free desktop software that uses artificial intelligence to help with the counting. It has already helped the team uncover a genetic change that improves fertility in plants with reproductive problems.
Thousands of pollen grains, one pair of human eyes
For flowering plants, successful reproduction depends on pollen reaching a flower and fertilising its ovules, allowing seeds to develop. In many crops, this process also underpins the harvest we depend on for food. But pollen development is sensitive to heat and drought. When these stresses disrupt its formation, fewer viable pollen grains may be available, reducing the plant’s ability to produce seeds. Genes also influence this process. Some genetic changes interfere with pollen development, while others can help compensate for reproductive defects. Understanding these effects can reveal why fertility declines and how it might be restored.
Counting viable pollen gives researchers a practical way to investigate these questions. It helps them compare how plants respond to different growing conditions or identify genetic changes associated with improved fertility. But across a large experiment, obtaining those counts can take hours of repetitive work.
“When working on larger experiments, we need to analyse a large number of samples. Assessing pollen itself is not complicated, but it takes a lot of time. Tasks like this are ideal for automation,” explains Vivek Kumar Raxwal, a postdoctoral researcher at CEITEC Masaryk University who developed PAT.
Software counts, scientists look for connections
The team developed the Pollen Analysis Tool (PAT) to count viable pollen directly in microscope images of stained, intact anthers, the parts of a flower where pollen develops. They trained an AI model to distinguish viable pollen grains from non-viable pollen and the surrounding plant cells.
Researchers load their images, and PAT identifies and counts the viable grains. The software marks its detections on each image, allowing users to review the results and interactively correct mistakes. They can then compare pollen counts between plants grown under different conditions, such as heat or drought, or carrying different genetic changes. PAT also creates graphs and allows users to further train the model on their own images, without writing code.
“For us, the value of AI is in solving a practical research problem and making the solution accessible. We built PAT so researchers can work with larger numbers of samples and spend more time exploring what the results actually mean,” says Vivek Raxwal.
From counting pollen to a discovery
Producing viable pollen depends on meiosis, a specialised cell division that gives reproductive cells the correct number of chromosomes. Environmental stresses, including heat, can disrupt this process and reduce fertility. Understanding the genes that keep meiosis working is therefore an important part of understanding how plants reproduce, and why reproduction sometimes fails.
To discover previously unrecognised players in this process, the team studied Arabidopsis plants carrying a mutation that prevents meiosis from finishing properly. They then searched for additional genetic changes that could ease this defect. If another genetic change helps the plant overcome the original defect, it can point scientists towards another gene involved in the process.
“We wanted to find genes involved in the successful completion of meiosis. Plants producing more viable pollen gave us a clue that something had improved. PAT helped us compare promising candidates and decide which to investigate further,” explains Darya Volkava, the study’s first author.
Further experiments identified a mutation in a gene involved in organising chromosomes. This genetic change helped the affected cells complete meiosis and increased viable pollen production, revealing a new link to how this process is brought to an end.
Less routine work, more room for biology
For Karel Říha, the research group leader who supervised the study, the value of PAT lies in making it easier to connect observations with new biological questions. Faster pollen counting allows researchers to compare more plants and identify differences worth investigating, whether they arise from genetic changes or growing conditions.
“Many experiments are limited by how much material we can realistically analyse. Tools like PAT help us extend that limit. We can examine more plants, make comparisons and use the results to ask new questions about how reproduction is controlled,” says Karel Říha.
PAT is free and open source, so other laboratories can use it and adapt the model to their own images. By making a routine measurement less time-consuming, the team hopes to support wider research into plant fertility and its responses to a changing environment
The findings are published in the Journal of Experimental Botany, and PAT is available online.
Author
Expert
