About Flagship 2
Traditional crop breeding is based on the assessment of aboveground plant traits because these are the easiest to observe and measure. Aboveground traits are also much easier to assess by high-throughput phenotyping, using sensors mounted on drones or satellites.
However, the root network and its associated microbiome play the key role in nutrient and water acquisition. Root traits are therefore just as important in terms of selection targets as the more visible and accessible aboveground organs, but they are rarely considered in breeding programs.
Flagship Project 2 is developing methods for the comprehensive characterization of root traits, the root microbiome and soil properties, and is integrating the resulting datasets to improve breeding programs. To overcome the laborious and expensive process of sampling and measuring the root microbiome across many different genotypes in multiple environments, we are using a more streamlined approach to identify plant genes that control the properties of the microbiome, allowing those genes to be introduced into broader breeding programs. We are also developing an open-source computer program that will allow anyone to explore and select for microbes that will benefit crops on their land, thus optimizing belowground traits to maximize crop sustainability.
- Rapid screening of canola lines for bacteria that promote yield stability.
- Screening of field soils for microbiomes that provide consistent yields.
- A pencil core sampling method for the non-destructive collection of soil around plant roots.
- Spectroscopic methods for the rapid assessment of pencil cores for beneficial microbiomes.
- Rapid screening of canola lines for bacteria that promote nutrient use efficiency.
- Screening of field soils for microbiomes that enhance nutrient use efficiency.
- Analysis of key microbiome networks that promote nutrient use efficiency and yield stability.
- New unsupervised machine learning methods.
The following projects are currently underway within Flagship 2:
Identifying properties of the microbiome linked to value-added activities of crop breeding such as yield stability and nitrogen use efficiency, and determining whether these aspects of the microbiome are heritable.
Microbes associated with yield responses in crops can be isolated from soil using pencil cores, which are much simpler and less destructive than traditional sampling methods. We have used this approach to construct microbiome datasets allowing us to identify microbes linked to yield, nitrogen use efficiency and aboveground phenotypes in canola and wheat. We are also using spectroscopic methods to link spectral properties of rhizosphere soil samples with soil fertility, microbiome structure and canola/wheat performance. Finally, we are assessing the heritability of microbiome-associated traits to allow their use in canola and wheat breeding programs.
We have identified bacteria linked to canola yield stability and have collected samples for metagenome analysis to link differences in bacterial identity with key metabolic functions. We have completed a multi-location trial comparing the wheat microbiome collected from soil pencil cores with the more laborious method involving the excavation of plants, revealing that pencil core samples can identify differences between plant lines and the root microbiome. We have generated spectral libraries for 2711 microbiomes derived from canola and wheat genotypes. The FTIR spectra were particularly valuable, providing information about soil organic matter, plant fine roots, and the soil microbiome. We have conducted field trials on two sites at which samples were collected from canola plants using the traditional hand trowel method. The samples were collected before and during the flowering stage allowing the comparative analysis of the root microbiome at different times in the growing season.
Microbial Data Analytics and Machine Learning
Using data analytics and machine learning to evaluate the value of the microbiome in breeding programs, and identifying direct links between these microbiome constituents and specific plant genes.
Winnowing, an agricultural technique to separate grain from chaff, is the name of our analytical pipeline to derive useful information from microbiome data. Important data are selected on the basis of external factors such as yield and nitrogen use efficiency, allowing us to identify microbes associated with these factors. We are now developing an abundance-based network approach that will help to determine whether the microbiome has a causal effect on crop yields. Having defined important microbiome components, we are developing new data transformation methods linking those components to plant genes for use in breeding programs. We are converting our Python-based winnowing pipeline into a QIIME2 plug-in for distribution to the global plant–microbe and crop-breeding community.
We have explored multiple ways to define importance in microbiome datasets, including graph-based approaches and definitions based on the inertia or variable abundance of individual microbial species in response to changing external conditions. We have also linked changes in microbial abundance to other multivariate data, such as spectral image datasets. We have optimized the network graph heuristic to calculate the quality measure and sign pattern validity measure more efficiently using approximate solutions to differential equations, allowing process parallelization. We are currently processing the microbial metagenomic data to define the most important microbial components associated with yield and yield stability traits. We have evaluated several transformation methods for the integration of microbiome data with other large datasets and have created a beta version of a QIIME2 plugin for the original version of our winnowing pipeline. The more recent extended version is being ported to a QIIME2 plugin called VAVOOM (Variability of Added Value of Operational Microbiome).