From obtaining allele substitution values to building predictive models, SNP & Variation Suite has all the tools for genomic prediction and visualization. Compare and contrast results using the available methods or pick your favorite method. Covariates can be included in every analysis and X-Chromosome correction is also available. SVS simplifies the entire genomic prediction process from data management to model building to visualization.
Genomic Prediction Methods
Methods available in SVS include Genomic Best Linear Unbiased Predictors (GBLUP), Bayes C and Bayes C-Pi. These tools create and find a solution to, or an approximate solution to, one or more sets of mixed linear model equations. The genomic information from the samples is included in every model to obtain a "genomic prediction".
Given the available dataset, genomic prediction methods can be used to build a prediction model that explains the association between the genotypes (genetic data) and the phenotype information best. This model can then be used in research to better understand the phenotype, and in commercial applications to improve decision making.
K-Fold Cross Validation
Automatically build training and validation sets within SVS using K-Fold Cross Validation. Account for stratification when picking the samples for each set to ensure balanced sets to obtain the best prediction models. Then run genomic prediction for one or more genomic prediction methods directly from K-Fold Cross Validation to save time and mouse clicks. SVS's K-Fold Cross Validation will also ensure major and minor alleles are consistently encoded through each data subset to ensure consistent direction of effect.
Applying a Prediction Model to New Data
After building a model, apply it to a new dataset to predict the phenotype. If the phenotype values are known this can be used to validate the model. If unknown, this can be used to make decisions based on the genetic data for the samples without phenotype information based on the samples used to build the prediction model. SVS automatically adjusts for strand information to ensure consistent direction of effect between the model used for prediction and the dataset the model is applied to.
Visualize the predicted versus actual phenotypes in a cluster plot to gauge the accuracy of the prediction model.
Getting to a scatter plot with a trend line is straightforward and you can color the data points by any covariates or by a stratifying variable.
The normalized log-transformed allele substitution values are genomic data and as with all genomic data in SNP & Variation Suite, plotting these values with GenomeBrowse provides you with the genomic context to interpret the markers with the largest influence in the prediction model to interpret key genes. Our live-streaming annotation repository as well as custom annotations for dozens of species can help decipher the significance of any results in the context of your research. Learn more about GenomeBrowse as a standalone tool.
Case Study - Raman Babu, CIMMYT
Being able to predict the phenotypic performance of a plant without testing it in the field ... is at the heart of the research within the plant breeding community."
Webcast - Genomic Analyses for Palatability of Beef
Genomic Prediction eBook
Genomic Prediction in Agriculture
by Dr. Andreas Scherer
View a sample project in SVS
Explore a Genomic Prediction Project in SVS - FREE!