Characterization of an Integrative Model for Candidate Gene Selection from Genomic Data with Simulated Data Analysis

Although multiple large data sets have been generated for identification of candidate genes for disease or economically important phenotypic traits, many of the key genes involved remain unknown.  Previously, we have developed an algorithm that incorporates weights for a gene, as quantified from multiple types of data, into an aggregated gene weight to assess their relative importance.  The objective of this project was to develop and analyze a simulated data set for improvement of the current model.  A simulated data set of QTL and gene expression (GE) data were created from data available from the Animal QTL Database and NCBI GEO, which led to the identification of myostatin in muscle hypertrophy phenotypes.  Genes from QTL and GE projects were assigned weights, normalized and aggregated to determine a specific gene’s relative importance.  Gene weights for QTL data were defined as the distance to the closest QTL peak, and weights for GE data were based on the log of the fold change.  From the available QTL data, simulated data sets were created with either high or low presence of QTL for myostatin present in the data (n=100 ‘projects’ each) and repeated for 10 simulations.  The ranks for myostatin were compared.  Myostatin was consistently ranked higher in the aggregated gene weight from muscle hypertrophy data sets than randomized data sets (P<0.05).   Utilization of the simulated data sets based on multiple traits associated with myostatin QTL and GE data led to an improved model algorithm and increased confidence in candidate gene outputs.