Design and analysis of high throughput, phenotyping experiments
H. Rabie, B. Berger, M. Tester, N. Al-Tamimi, H. Oakey, S. Saade, Y. S. Ho, S. Schmöckel, S. and N. Negrão
The Plant Accelerator facility incorporates normal glasshouse rooms and Smarthouses, the latter having the capacity to automatically measure many phenomic characters. It is recognized that there are trends in these rooms and taking these into account in the design used for an experiment might be approached in several different ways. These include: a) no design but move the plants about arbitrarily during the experiment; b) use a design that incorporates blocking, usually of rows and columns; c) use a design that allows for spatial correlation in rows and columns.
Further, some experiments are being conducted in the Plant Accelerator that are three-phase in that plants spend an initial phase in an ordinary glasshouse room during the intitial growth period. Then, in an imaging phase, they are transferred to a Smartroom for imaging during the main growth period. Finally, in the final phase, they are returned to an ordinary glasshouse room until they are harvested. Designing such experiments requires a separate design for each phase, but the design for one phase must take into account those for previous phases. Methods for producing such designs have been investigated.
We investigated the design of such experiments by running an experiment in the Plant Accelerator and the results are reported in Brien, Berger, Rabie and Tester (2013).
As far as analysis is concerned, the data set from an experiment of this type is typically large (15,000 values for a single trait) and complicated in that it involves longitudinal data which are to be used to measure plant growth. We have developed tecniques involving the fitting of splines, the derivation of continuous and interval growth rates, and mixed model analysis of growth traits for analyzing the data. The methods have been published in Al-Tamimi, Brien, Oakey, Berger, Saade, Ho, Schmöckel, Tester and Negrão (2016).
Brien, C. J., Berger, B., Rabie, H. and Tester, M. (2013) Accounting for variation in designing greenhouse experiments with special reference to greenhouses containing plants on conveyor systems. Plant Methods 9:5. Published paper.
Al-Tamimi, N., Brien, C., Oakey, H., Berger, B., Saade, S., Ho, Y. S., Schmöckel, S., Tester, M., and Negrão, N. (2016) Salinity tolerance loci revealed in rice using high-throughput non-invasive phenotyping. Nature Communications 7:13342. Published paper.
and analysis of incubator and greenhouse experiments to evaluate native
Brien, T. Tran,
J. Boland, R. A. Bailey, H. Mancini and J. Gibb
Revegetation along rail corridor has been considered as a means of solving rail track stability issues and weed and fire risk problems. One of the challenges of restoring vegetation in the rail corridor has been to identify an optimal planting regime in structure, density and species composition in order to address specific site problems, create stable plant communities and conform to industry standards for safety and maintenance requirements. The design and analysis of experiments to study different species, soils and soil-preparation treatments in incubator and greenhouse experiments has been investigated. This has resulted in the development of split plot designs in which both main plots and sub-plots employ two-dimensional designs and in which sub-plot treatments are latinized. A new class of row-column designs, the quasi-Latin rectangle designs, were developed for assigning several factors to the main plots. Also, a comparison of analysis methods for count data from multistratum experiments has been conducted. The methods compared were analysis of variance, both untransformed and transformed, generalized linear models and generalized linear mixed models.
Mancini, H., Tran, T. T., Gibbs, J., Brien, C. J. & Boland, J. (2006) Improving rail corridors by restoring native vegetation. In Ravitharan, R. (Ed.) Rail Achieving Growth: Proceedings of the Conference on Railway Engineering (CORE) 2006, Melbourne, 30 April - 3 May 2006. Melbourne, Railway Technical Society of Australasia (RTSA).
Tran, T. T. (2009) Design and Analysis of Experiments for Assessing Indigenous Plant Species. PhD thesis, School of Mathematics and Statistics, University of South Australia.
Brien, C.J., Bailey, R.A., Thao, T.T. and Boland, J. (2012) Quasi-Latin designs. Electronic Journal of Statistics, 6, 1900-1925. Published paper
and analysis of sugarcane breeding experiments
Brien, C. Demétrio, R. Sermarini and A. dos Santos
Both Australia and Brazil conduct sugarcane breeding experiments and, as for cereal breeding experiments, the early generation experiments involve a large number of lines. However, because of the restricted availabilty of seeds and experimental area, each line is either unreplicated or duplicated. Work in Australia by Jo Stringer and colleagues has shown that models that allow for genetic and residual competition are needed. We have been investigating whether the same models are required for Brazilian experiments. Further, the designs traditionally used for such experiments are systematic check-plot and augmented designs. We are investigating the use of the more recently developed partially-replicated (p-rep) designs and combinations of check plots and partial replication.