1. solGS:
a
Web-‐based
Solution
for
Genomic
Selection
Genomic selection, due to its reliance on dense genome-wide markers and statistical complexity,
presents significant challenges in data management, analysis and sharing results. solGS, a web-based
tool, meets these challenges;; it has a database to store phenotype and genotype data and an intuitive
web-interface for statistical analyses and data visualization. It uses RR-BLUP for the statistical
modeling and GBLUP method for genomic breeding values estimation (GEBV). It performs also
descriptive statistics, population structure, phenotypic and genetic correlations, and selection index
analyses. It visualizes data in interactive plots onthe browser. solGS is, currently, used by the NextGen
Cassava Breeding Project (http://nextgencassava.org) and implemented on
http://cassavabase.org/solgs. GS breeders canadapt the tool for any organism.
Funding sources http://cassavabase.org
Contact: sgn-feedback@solgenomics.net
Data andcode
ftp://ftp.solgenomics.net
http://github.com/solgenomics
Isaak
Y
Tecle,
Naama
Menda,
Guillaume
Bauchet
and
Lukas
Mueller
Boyce
Thomson
Institute,
Cornell
University,
Ithaca,
NY
14853.
solGS enables breeders to store raw data and estimate GEBVs of individuals online, in an
interactive workflow. The code is open-source and can be adapted to any breeding program.
Conclusion
Fig 4B. You can also evaluate
the phenotypic and genetic
correlation between traits.
Fig 3. A single trait prediction
model output page displays
description of the model
inputs including a graphica l
plot of the phenotype data
(Fig 4A), model output
including model accuracy
and variance components,
the GEBVs of individuals
used in the model and
marker effects (A). From the
same model page, you can
search for relevant selection
populations and predict their
GEBVs using the model (B).
Single trait prediction model output
Exploratory tools
Fig 4A. The phenotype data
used in the model is
displayed in scatter plot.
Fig 4C. Using the marker
data, you can evaluate the
population structure.
Fig 1. solGS enables you
to choose a training
population for your model
in three ways. One way is
using a trait name to
search the database for
individuals phenotyped for
that trait and select any
number of trials (A).
Second way is to search
for trials and use
individuals from any
number of trials (B). The
third way is to make your
own list of individuals
using a ‘List’ tool (C).
Options B and C lead you
to a training population
page (Fig 2).
A
B
C
Creating a training population
In simple terms, aGS process has 3 steps: (1) building a prediction model using a training population,
with phenotype and genotype data;; (2) validation of model accuracy, e.g. using cross-validation;; (3)
applying a prediction model to estimate GEBVs of selection candidates, with genotypedata only.
A
B
Multiple
traits
models
output
Reference
Tecle et al. solGS: a web-based tool for genomic selection. BMC Bioinformatics 2014, 15: 398.
C
A
B
Training population detail page
Fig 2. From the training
population, you can select
traits to build models and
also run phenotypic
correlation analysis. If you
run a model for a single
trait, you get an output
page as shown in Fig 3. If
you run models
simultaneously for multiple
traits, you get an output
page as shown in Fig 5.
Phenotype Data
Correlation
Population structure
GEBVs vs Phenotypes
Fig 5. Detail page when you
run models for multiple traits
from a training population. In
section A are, e.g., two models
for two traits with their
corresponding models
accuracy and heritability.
Following the trait links, you
can view the model output in
detail as in Fig 3. In section B,
you can simultaneously apply
models to predict GEBVs of a
selection population for the
traits. In section C, you can
calculate genetic correlation
between the traits. In section
D, you can calculate selection
indices by assigning relative
weights for any number of
traits with GEBVs from the
same population.
D Fig 4D. OLS regression of
GEBVs on the phenotype
value shows the relationship
between the two variables.
A
B
C
D