This digital illustration demonstrates the application of Arabidopsis thaliana (L.) Heynh. (Brassicaceae) gene

expression data to train machine learning models for predicting tissue identity across diverse flowering plants. Arabidopsis thaliana, shown on the far left, serves as the foundational model species due to its rich genomic resources and central role in advancing plant biology research. The gene expression profi les (middle left) extracted from Arabidopsis are used to train computational models (center), which are then applied to predict tissue identity in other species, as illustrated by pumpkin (Cucurbita pepo L., Cucurbitaceae), sunfl ower (Helianthus annuus L., Asteraceae), and orange (Citrus sinensis (L.) Osbeck, Rutaceae) (far right). In the study “Expression-based machine learning models for predicting plant tissue identity,” Palande et al. compared multiple machine learning algorithms, fi nding that models trained within Arabidopsis data achieved near-perfect precision and recall, but performance dropped when predictions were extended across species. Notably, tissue prediction was most accurate for belowground tissues, and success was not correlated with phylogenetic distance from Arabidopsis. These findings emphasize the limitations of Arabidopsis-centric research and the need to integrate data from diverse plant species to enhance cross-species model generalizability. This study highlights the potential of gene expression signatures, rather than marker genes, for advancing tissue and cell type prediction in plants. It also underscores the value of computational tools in plant biology while challenging the reliance on a single model organism for  cross-species studies. Image credit: Drawing by Tamara Santos-Rougon; digital editing by Alejandra Rougon-Cardoso.