This title appears in the Scientific Report :
2016
Complementing HTP phenotyping with improved and semiautomated experimental design and analysis pipelines
Complementing HTP phenotyping with improved and semiautomated experimental design and analysis pipelines
Plant phenotyping includes monitoring of rather complex plant processes which are linked to environmental factors. As complete homogeneity of the factors like temperature, light, water and airflow is usually impossible, it is of importance to optimize experiment design to exclude the impact of these...
Saved in:
Personal Name(s): | Li, Jinquan (Corresponding author) |
---|---|
Koerber, Niklas / Fiorani, Fabio (Corresponding author) | |
Contributing Institute: |
Pflanzenwissenschaften; IBG-2 |
Imprint: |
2016
|
Conference: | Plant 2030 Status Seminar, Potsdam (Germany), 2016-03-14 - 2016-03-16 |
Document Type: |
Poster |
Research Program: |
Deutsches Pflanzen Phänotypisierungsnetzwerk Plant Science |
Publikationsportal JuSER |
Plant phenotyping includes monitoring of rather complex plant processes which are linked to environmental factors. As complete homogeneity of the factors like temperature, light, water and airflow is usually impossible, it is of importance to optimize experiment design to exclude the impact of these compound factors. Furthermore, an optimization of cost, throughput and reliability of the data for high-throughput (HTP) phenotyping depends largely on the experimental design. To do this, an existing R package “agricolae” in statistical software R (de Mendiburu, 2009) was modified to generate 13 frequently used experimental designs (completely randomized design, randomized complete block design, Latin squared design, split-plot design, Graece-Latin design, augmented design, Youden design, Balance incomplete design, cyclic design, lattice design, alpha design, strip-plot design, and factorial design). The modified R package has been applied in generating different experimental designs for our common experiment with maize, barley, rapeseed, wheat, and sugar beet in the field and Arabidopsis 1001 genome phenotyping experiment in the screen chamber in 2015. A semi-automated pipeline for statistical analysis with these experimental designs and the data from high-throughput phenotyping experiments has been developed. The statistical functions include: descriptive statistics, frequency and correlation analysis, imputation of missing data with consideration of distribution of the sample data, detection of outlier data, determination of optimal replicates, verifying the assumptions of the linear model, data transformation, analysis of variance (ANOVA) with appropriate statistical models, interaction of different factors, multiple comparisons with different methods, time-serials or repeated measured data analysis, and so on. Graphic user interfaces (GUI) are under development for all statistical analysis procedures so that non-experienced R users can easily automate the selection of a suitable experimental design and the data analysis afterwards by using a step-by-step selection workflow as well as make it possible to monitor and analyze these experiments even during the run.Reference:1. Felipe de Mendiburu (2009) agricolae: Statistical procedures for agricultural research. R package version 1.2-3, https://cran.r-project.org/web/packages/agricolae/ |