Background
Wheat is a major global commodity. With more than a decade of stalled commercial yield in Australia, new approaches are vital to improve crop productivity. In this regard, the prediction of time-to-flowering is a valuable measurement for optimising farm resource allocation and yield. Pan-transcriptome analysis has shown promise for flowering time prediction across diverse wheat varieties. Herein we demonstrate the utility of pan-proteomics for predicting flowering time in mature plants using sample extracted from plants at the two-leaf stage.
Methodologies
A diverse panel of Australian wheat varieties were grown under short (8h) or long day length (16h) conditions to facilitate variation in flowering time. A total of 632 unique wheat samples were processed and measured by variable window SWATH acquisition along with control measurements. Peptide responses were extracted from raw SWATH data along with peak group false discovery rate estimates. These data were processed and analysed using a suite of scripts within the R environment. The resulting matrix of pan-wheat measurements was subject to multivariate and machine learning analyses to assess variance and quantify the ability of protein abundances to predict flowering time.
Findings
We show that the major proteome variation can be readily attributed to day length using t-Distributed Stochastic Neighbour Embedding machine learning. We also quantify the ability for pan-proteome measurements from plants at the two-leaf stage to predict the flowering time of the mature plant through application of random forest analysis.
Concluding
Wheat is a substantial source of global nutrition and economic benefit. With the growing population and coincidental requirement for nutrition from cereals projected to increase by 50% over the next two decades, efficiency gains in grain production are required. Herein we demonstrate the ability to predict wheat traits using pan-proteome measurements that can inform on-farm practices aimed at improving crop quality and yield.