Technological developments have revolutionized measurements on plant genotypes and phenotypes, leading to the routine production of large, complex data sets. Phenotypes, broadly defined, include molecular or physiological activities (e.g., transcription, water use efficiency), and macroscopic traits (e.g., yield). A resulting challenge in both fundamental and applied plant sciences (e.g., breeding) is to explain or predict phenotypes from the underlying genotypes under different environments. Genotypic variation in combination with environmental variation leads to differences in the biochemical makeup of cells, measurable as molecular phenotypes. These in turn influence organ formation, plant growth, and eventually traits relevant in agriculture, such as yield and stress tolerance. Thus, relating genotypes to phenotypes yields fundamental insights into the regulation of important processes in plant development and physiology as well as the ability to predict yield and quality traits in specific environments, which is essential in basic plant science and molecular breeding of resilient plants in changing environments.
Analyzing phenotypes measured at different scales or linking these phenotypes to genotypes increasingly calls for processing and integration of large, noisy, and heterogeneous data sets. To exploit the full potential of these data, artificial intelligence and machine learning methods are essential tools. As a result, AI/ML algorithms are now starting to be widely applied in plant science and plant breeding. Next to applications of existing AI/ML methods, a novel methodology is being developed for challenges specific to this area (e.g., comparative and evolutionary analyses of wide varieties of complex genomes, and reconstruction of molecular networks) and specific applications in plant breeding, such as genomic prediction and selection. In this research topic, we intend to collect contributions at the interface of AI/ML and plant sciences.
We invite submissions discussing applications of AI/ML in the context of plant science and plant breeding, particularly focusing on analyses to connect genotypes to phenotypes at different levels, from molecular (transcripts, proteins, metabolites, etc.) to macroscopic (shape, growth, yield, etc.). Contributions can cover, but are not limited to methods for and applications of:
? Integrating and interpreting -omics data, also across space and time;
? Decision support in experimentation, breeding programs, etc. (explainable AI/ML, causal inference, active learning);
? Collecting and integrating prior (biological) knowledge (NLP, knowledge engineering);
? Exploiting unlabeled data for prediction (embeddings, self-training, semi-supervised);
? Enhancing understanding of underlying mechanisms (interpretable/explainable AI/ML);
? Break down and/or aggregate complex traits into more easily interpretable/measurable components;
? Translating models between model organisms and relevant other (crop) species (transfer learning);
? Bridging the gap between traditional statistical approaches and advances in AI/ML;
? Connecting bottom-up (“systems biology”) / mechanistic (“crop modeling”) models and AI/ML;
? ML methods for incorporating and predicting genotype-by-environment interactions.
Overall, the focus is on already measured phenotypes rather than on analysis of raw phenotype data (e.g. through computer vision).
Technological developments have revolutionized measurements on plant genotypes and phenotypes, leading to the routine production of large, complex data sets. Phenotypes, broadly defined, include molecular or physiological activities (e.g., transcription, water use efficiency), and macroscopic traits (e.g., yield). A resulting challenge in both fundamental and applied plant sciences (e.g., breeding) is to explain or predict phenotypes from the underlying genotypes under different environments. Genotypic variation in combination with environmental variation leads to differences in the biochemical makeup of cells, measurable as molecular phenotypes. These in turn influence organ formation, plant growth, and eventually traits relevant in agriculture, such as yield and stress tolerance. Thus, relating genotypes to phenotypes yields fundamental insights into the regulation of important processes in plant development and physiology as well as the ability to predict yield and quality traits in specific environments, which is essential in basic plant science and molecular breeding of resilient plants in changing environments.
Analyzing phenotypes measured at different scales or linking these phenotypes to genotypes increasingly calls for processing and integration of large, noisy, and heterogeneous data sets. To exploit the full potential of these data, artificial intelligence and machine learning methods are essential tools. As a result, AI/ML algorithms are now starting to be widely applied in plant science and plant breeding. Next to applications of existing AI/ML methods, a novel methodology is being developed for challenges specific to this area (e.g., comparative and evolutionary analyses of wide varieties of complex genomes, and reconstruction of molecular networks) and specific applications in plant breeding, such as genomic prediction and selection. In this research topic, we intend to collect contributions at the interface of AI/ML and plant sciences.
We invite submissions discussing applications of AI/ML in the context of plant science and plant breeding, particularly focusing on analyses to connect genotypes to phenotypes at different levels, from molecular (transcripts, proteins, metabolites, etc.) to macroscopic (shape, growth, yield, etc.). Contributions can cover, but are not limited to methods for and applications of:
? Integrating and interpreting -omics data, also across space and time;
? Decision support in experimentation, breeding programs, etc. (explainable AI/ML, causal inference, active learning);
? Collecting and integrating prior (biological) knowledge (NLP, knowledge engineering);
? Exploiting unlabeled data for prediction (embeddings, self-training, semi-supervised);
? Enhancing understanding of underlying mechanisms (interpretable/explainable AI/ML);
? Break down and/or aggregate complex traits into more easily interpretable/measurable components;
? Translating models between model organisms and relevant other (crop) species (transfer learning);
? Bridging the gap between traditional statistical approaches and advances in AI/ML;
? Connecting bottom-up (“systems biology”) / mechanistic (“crop modeling”) models and AI/ML;
? ML methods for incorporating and predicting genotype-by-environment interactions.
Overall, the focus is on already measured phenotypes rather than on analysis of raw phenotype data (e.g. through computer vision).