Expression-based machine learning models for predicting plant tissue identity

Abstract

The selection of Arabidopsis as a model organism played a pivotal role in advancing genomic science, firmly establishing the cornerstone of modern plant molecular biology. Competing frameworks to select an agricultural- or ecological-based model species, or to decentralize plant science and study a multitude of diverse species, were selected against in favor of building core knowledge in a species that would facilitate genome-enabled research that could assumedly be transferred to other plants. Here, over twenty years after sequencing the Arabidopsis genome and during which time sequencing data from other plant species has accumulated and computation enabling machine learning has evolved, we critically examine the ability of models based on Arabidopsis gene expression data to predict tissue identity in other flowering plant species. Comparing different machine learning algorithms, models trained and tested on Arabidopsis data achieved precision and recall values of up to 0.99 using the K-Nearest Neighbor method, whereas when tissue identity is predicted across the flowering plants using models trained on Arabidopsis data, precision values range from 0.70 to 0.75 and recall from 0.55 to 0.64, depending on the algorithm used. Below-ground tissue is more predictable than other tissue types, and the ability to predict tissue identity is not correlated with phylogenetic distance from Arabidopsis. Our data-driven results suggest that, in hindsight, the assertion that knowledge from Arabidopsis is translatable to other plants is not as strong as originally assumed, and that in the current era where sequencing data and computation abound, we should decentralize the scientific focus on Arabidopsis and embrace plant diversity.