GRANA: An AI-based tool for accelerating chloroplast grana nanomorphology analysis using hybrid intelligence.

Abstract

Grana are fundamental structural units of the intricate chloroplast membrane network. Investigating their nanomorphology is essential for understanding photosynthetic efficiency regulation. Here, we present GRANA (Graphical Recognition and Analysis of Nanostructural Assemblies), an artificial intelligence-enhanced, user-friendly software tool that recognizes grana on thylakoid network electron micrographs and generates a complex set of their structural parameters. GRANA employs 3 artificial neural networks of different architectures and binds them in a 1-click workflow. Its output is designed to facilitate hybrid intelligence analysis, securing fast and reliable results from large datasets. The GRANA tool is over 100 times faster compared with currently used manual approaches. As a proof of concept, we have successfully applied GRANA software to diverse grana structures across different land plant species grown under various conditions, demonstrating the wide range of potential applications for our software. GRANA tool supports large-scale analysis of grana nanomorphological features, facilitating advancements in photosynthesis-oriented studies.

Code and data availability

The paper's raw TEM images used for grana nanomorphology analysis are publicly deposited under DOI 10.58132/HTWCC1. The authors' analysis code (github.com/center4ml/GRANA) is mentioned but that URL is not among the allowed URLs, so it cannot be listed.