A high-throughput differential chemical genetic screen uncovers genotype-specific compounds altering plant growth.

Abstract

The identification of chemical compounds regulating plant growth in a genetic context can greatly enhance our understanding of biological mechanisms. Here, we have developed a high-throughput phenotype-directed chemical screening method in plants to compare two genotypes and identify small molecules inducing genotype-specific phenotypes. We used Arabidopsis thaliana wild type and mus81 , a DNA repair mutant, and screened off-patent drugs from the Prestwick library to selectively identify molecules affecting mus81 growth. We developed two complementary convolutional neural networks (CNN)-based image segmentation and classification programs to quantify Arabidopsis seedling growth. Using these approaches, we detected that about 10% of Prestwick molecules cause altered growth in both genotypes, suggesting their toxic effects on plant growth. We identified three Prestwick molecules specifically affecting mus81 . Overall, we developed a straightforward, accurate, and adaptable methodology for performing high-throughput screening of chemical libraries in a time-efficient manner, accelerating the discovery of genotype-specific chemical regulators of plant growth.

Plant phenotyping relevance

植物成長を定量するCNN画像セグメンテーション・分類法と、化学ライブラリのハイスループット表現型スクリーニング手法の開発が研究の中心である。

Code and data availability

The paper describes CNN-based classification/segmentation tools and seedling image datasets, but no public deposit of the authors' code, trained models, or image/phenotype data is stated. Fiji, PyTorch, OpenCV, and BIP are generic third-party tools, not paper-specific assets.

No evidence-backed public reproduction asset is currently recorded.