CiteAllInOne.ris

TY  - MISC
AU  - Yoosefzadeh Najafabadi, Mohsen
AU  - Heidari, Ali
AU  - Rajcan, Istvan
TI  - AllInOne
PY  - 2022
DA  - 2022
AB  - Plant breeding is a great mixture of art and science that has been around
      for centuries and plays an important role in solving many of the world’s
      agricultural problems. The ultimate goal for most plant breeding programs
      is to develop new varieties that are better adapted for certain
      environments, more resistant to biotic/abiotic stresses, and/or have
      increased yields. The process usually starts with selecting and crossing
      two varieties as parental lines with specific features to create a
      population with a high recombination rate. Afterward, the created
      population is evaluated for several generations to select the superior
      varieties based on the plant breeding goal. The first evaluation of the
      breeding population mainly relies on the phenotypic data derived from
      different plant varieties growing in multi-locations for several years.
      The collected phenotypic data can be further used in combination with the
      environment and omics data to increase the accuracy of the breeding
      decisions. Therefore, the accuracy, quality, and nature of phenotypic data
      are the three most important factors in capturing true signals and
      avoiding false interpretations. In addition, the proper selection of
      statistical analysis methods is of high paramount in making the right
      decision to select superior genotypes based on a trait of interest. The
      accuracy of phenotypic data highly depends on how plant breeders collect
      data. Conventional data collection methods can bring more noise to the
      phenotypic data than high throughput methods. In addition, the
      contribution of more data collectors in measuring phenotypic data may
      decrease the level of accuracy. Moreover, the existence and abundance of
      genotypes that exhibit a particular phenotypic characteristic is an
      important determinant of the extent to which analysis methods can be used
      to select superior genotypes. Therefore, as plant breeders have been
      dealing with a large number of genotypes in different locations over
      several years, it is necessary to pre-process the phenotypic datasets
      before selecting any methods or making any decision. Pre-processing is one
      of the important procedures for increasing the quality and accuracy of the
      field phenotypic data, which is usually done in several steps, such as 1)
      detecting missing patterns in the dataset, 2) imputing missing data using
      different statistical methods, 3) data visualization in order to check
      data patterns and distribution, 4) detecting and refining outliers, 5)
      estimating correlations between dependant variables and also with and
      within independent variables, 6) normalizing data based on the optimum
      normalization methods for a given dataset, 6) estimating heritability and
      conducting spatial analysis, and finally, 7) calculating best linear
      unbiased prediction (BLUP) or/and best linear unbiased estimator (BLUE)
      based on the goal of the plant breeding program. Several packages in
      different languages (mainly in R) have been created to handle different
      pre-processing steps, such as 1) MICE to deal with missing data, 2)
      Bestormalizer to normalize the datasets using different methods, 3) lme4
      and ASReml R to handle variance analysis using different experimental
      designs, 4) ggplot for data visualization, etc. However, none of them are
      able to handle at least most of the pre-processing steps in plant breeding
      phenotypic datasets. In addition, they all require medium to advanced
      coding knowledge to run and adjust the functions based on the breeding
      preference. Furthermore, more sophisticated packages such as ASReml R are
      not free and require an annual renewal fee. Moreover, none of them
      provided a dynamical graphical interference for creating plots, detecting
      and refining outliers in a live mode, and using different datasets without
      directly importing from a file. Here, we introduce the AllInOne package as
      an open-Source, breeder-friendly, analytical R package for pre-processing
      phenotypic data. The basis of AllInOne is to utilize different R packages
      and develop the pipeline for pre-processing the phenotypic datasets in an
      accurate, easy, and timely manner without any coding skills required. In
      addition, several new features and abilities were added in AllInOne that
      can complement previously developed packages in this area.
UR  - https://allinone.shinyapps.io/allinone/
KW  - Pre-processing, plant breeding, data collection, curation, R shiny package, agriculture
ER  -