Physics-informed deep learning characterizes morphodynamics of Asian soybean rust disease

Henry Cavanagh,Rob Lind,Robert G Endres,Andreas Mosbach,Gabriel Scalliet

doi:10.1038/s41467-021-26577-1

Henry Cavanagh, Rob Lind + Show 3 more

Open Access

https://doi.org/10.1038/s41467-021-26577-1

Copy DOI

Abstract

Medicines and agricultural biocides are often discovered using large phenotypic screens across hundreds of compounds, where visible effects of whole organisms are compared to gauge efficacy and possible modes of action. However, such analysis is often limited to human-defined and static features. Here, we introduce a novel framework that can characterize shape changes (morphodynamics) for cell-drug interactions directly from images, and use it to interpret perturbed development of Phakopsora pachyrhizi, the Asian soybean rust crop pathogen. We describe population development over a 2D space of shapes (morphospace) using two models with condition-dependent parameters: a top-down Fokker-Planck model of diffusive development over Waddington-type landscapes, and a bottom-up model of tip growth. We discover a variety of landscapes, describing phenotype transitions during growth, and identify possible perturbations in the tip growth machinery that cause this variation. This demonstrates a widely-applicable integration of unsupervised learning and biophysical modeling.

Highlights

Medicines and agricultural biocides are often discovered using large phenotypic screens across hundreds of compounds, where visible effects of whole organisms are compared to gauge efficacy and possible modes of action
Recent progress in automated image analysis has popularized static descriptors beyond mean growth rates and metabolic fluxes[3], the incorporation of dynamics can provide more complete system descriptions[4] and may aid the development and validation of mechanistic models[5]. We developed such a framework and used it to interpret how fungicides affect the morphodynamics of Phakopsora pachyrhizi, the pathogen that causes rust disease in the soybean crop worldwide
Manifold-based dimensionality reduction can provide a continuous low-dimensional space where dynamics are as simple as possible[26]. This is because an imaged dynamic system with n degrees of freedom traces out an n-dimensional manifold within the higher-dimensional pixel space, irrespective of the image dimensionality

Summary

Introduction

Medicines and agricultural biocides are often discovered using large phenotypic screens across hundreds of compounds, where visible effects of whole organisms are compared to gauge efficacy and possible modes of action. Recent progress in automated image analysis has popularized static descriptors beyond mean growth rates and metabolic fluxes[3], the incorporation of dynamics can provide more complete system descriptions[4] and may aid the development and validation of mechanistic models[5] We developed such a framework and used it to interpret how fungicides affect the morphodynamics of Phakopsora pachyrhizi, the pathogen that causes rust disease in the soybean crop worldwide. Existing methods have uncovered remarkable behavioral patterns, revealing chemotactic strategies, temporal processing, and social cooperation in a range of organisms[10,11,12,13], typical shortcomings are as follows: first, they are often based on particular shape descriptors (e.g., one-dimensional centerlines), restricting analysis to a narrow range of morphologies, often requiring sophisticated feature extraction algorithms They focus on stereotyped behaviors, which may not be characteristic of early development. Our work shows how intuitive system characterizations can be acquired directly from images, by integrating unsupervised learning and biophysical modeling

Objectives

Methods

Results

Conclusion