QUANTITATIVE TRAIT LOCI MAPPING AND STATISTICAL MODELING OF COMPLEX TRAITS USING GENOMIC DATA

Abstract

Complex traits are affected by several genetic loci, and also complex interactions of the genotype and environment, and thus their accurate prediction is a major challenge of genomic studies research. This paper hypothesises a combined methodology that consists of quantitative trait loci (QTL) mapping and statistical modelling to describe and forecast complex traits with high-dimensional genomic data. After quality control, probable minor allele frequency filtering and missing data imputation were applied, a high-density single nucleotide polymorphism (SNP) dataset containing 3, 200 samples and 52, 400 filtered markers were used. A mixed linear model (MLM) was utilised to estimate the effects of a population structure and kinship in genome-wide QTL mapping to effectively identify important loci. The identified QTLs were then introduced as input features in statistical and machine learning analyses including linear regression, random forest, and extreme gradient boosting (XGBoost) in predicting phenotypic characteristics. Cross-validation indices like RMSE and coefficient of determination (R 2 ) were used to assess the model performance. The highest predictive accuracy (R 2 = 0.87, RMSE = 2.05) was obtained with the use of the XGBoost model which was better than the classification with the help of traditional linear techniques. The findings reveal that the combination of QTL mapping and machine learning can substantially increase the accuracy of predictions and predetermine the discovery of biologically important genomic areas. This model offers an efficient and scalable method of new breeding that utilises genomics in the study of complex traits.