A Statistical Toolbox for QTL Analysis

BigRiverQTL.jl is a package that makes it easier to perform quantitative trait locus (QTL) analysis in Julia. It consists of three components streamline the QTL analysis workflow: preprocessing, genome scans, and visualization.

• Preprocessing: The preprocessing functions are designed to seamlessly import and convert genomic data into an efficient and memory-conservative format. This component also offers function capabilities for quickly calculating kinship matrices, ensuring data readiness for subsequent analysis phases.

• Genome Scans: We provide the capabilities to perform a large number of univariate genome scans using BulkLMM.jl. For analyses involving multiple traits, we provide an interface to FlxQTL.jl to analyze longitudinal traits and complex trait correlations.

• Visualization: The third component of the package offers plotting tools designed to visualize genome scans including eQTL scans using the BigRiverQTLPlots.jl package.

To install BigRiverQTL.jl, you can use Julia's package manager. Here is the command:

using Pkg
Pkg.add("BigRiverQTL")

Contributions to BigRiverQTL.jl are welcome and appreciated. If you would like to contribute, please fork the repository, make your changes, and send us a pull request. If you have a question or think you have found a bug, please open an issue.

This example is also available as a notebook in the 'example' directory.

In this example, we will use a dataset available from the R/qtl2 package. Specifically, we will use the BXD dataset, which is obtained from the GeneNetwork website.

You can download the BXD genotype data from the following link: Download BXD Genotype Data

# Libraries
using BigRiverQTL
using Plots

We assume that the data is stored in ..\data\BXD directory.

########
# Data #
########
data_dir = joinpath(@__DIR__, "../data/BXD/");
file = joinpath(data_dir, "bxd.json");

Load bxd data using the function get_geneticstudydata():

# Load bxd data
data = get_geneticstudydata(file);

# The current version of `BigRiverQTL` does not have imputation functions.
# Remove the  missing data
data = get_data_completecases(data);

# Data types
# gmap contains 
# makers info 
gInfo = data.gmap;

# phenotype info 
pInfo = data.phenocov;
# phenotype values 
pheno = data.pheno.val;

# We can get the genotype matrix using the following command.
# For computing reasons, we need to convert the geno matrix in Float64.
# One way to do it is to multiply by 1.0
geno = reduce(hcat, data.geno.val).*1.0;

#################
# Preprocessing #
#################
traitID = 1112;
pheno_y = pheno[:, traitID];
pheno_y2 = ones(length(pheno_y));
idx_not_missing = findall(!ismissing, pheno_y)
pheno_y2[idx_not_missing] = pheno_y[idx_not_missing];

###########
# Kinship #
###########
kinship = kinship_gs(geno,.99);

########
# Scan #
########

single_results_perms = scan(
	pheno_y2,
	geno,
	kinship;
	permutation_test = true,
	nperms = 1000,
);

#########
# Plots #
#########

# QTL plots
plot_QTL(single_results_perms, gInfo, mbColname = "Pos")

# Manhattan plots
plot_manhattan(single_results_perms, gInfo, mbColname = "Pos")