QTLNETWORK

QTLNetwork: Genetic Architecture and QTL Interaction Analysis

QTLNetwork implements full-quantitative trait locus (QTL) models to map and analyze the genetic architecture of complex traits in experimental populations derived from crosses between two inbred lines. It detects multiple QTLs, epistasis, QTL-by-environment interactions, and epistasis-by-environment interactions across multiple environments.

Key Features:

• Full-QTL Model: Integrates multiple QTL effects, epistasis (non-additive gene interactions), QTL-by-environment, and epistasis-by-environment interactions within a unified framework.
• Mixed Linear Model Mapping: Performs marker interval selection, genome scans, and detection of marker interval interactions using mixed linear models with environmental effects modeled as fixed or random.
• Statistical Testing Framework: Applies F-statistics based on Henderson's method III for hypothesis testing, permutation tests for genome-wide false positive control, and model selection to reduce ghost peaks in F-statistic profiles.
• Bayesian Parameter Estimation: Estimates full-QTL model parameters using Bayesian methods via Gibbs sampling and evaluates performance with Monte Carlo simulations.
• Experimental Design Support: Analyzes data from F2, backcross, recombinant inbred lines, double-haploid populations, immortalized F2, and BC(n)F(n) populations.
• Genotype Prediction Algorithms: Predicts superior multi-locus genotypes, lines, and hybrids based on additive, epistatic, and QTL-by-environment effects.

Scientific Applications:

• Complex Trait Dissection: Identifies QTLs and gene–environment interactions underlying phenotypes such as BXD mouse olfactory bulb weight and rice yield to support genotype optimization.

Methodology:

QTLNetwork applies a full-QTL mixed linear model to perform genome-wide scans and interaction detection. Hypothesis testing uses F-statistics derived from Henderson's method III with permutation-based control of genome-wide error rates. Model selection reduces spurious peaks. Parameter estimation is conducted using Bayesian inference via Gibbs sampling, and Monte Carlo simulations assess estimation reliability and efficiency.