The Quantitative Trait Loci (QTL) mapping problem aims to identify regions\r\nin the genome that are linked to phenotypic features of the developed organism that vary\r\nin degree. It is a principle step in determining targets for further genetic analysis and is\r\nkey in decoding the role of specific genes that control quantitative traits within species.\r\nApplications include identifying genetic causes of disease, optimization of cross-breeding\r\nfor desired traits and understanding trait diversity in populations. In this paper a new\r\nmulti-objective evolutionary algorithm (MOEA) method is introduced and is shown to\r\nincrease the accuracy of QTL mapping identification for both independent and epistatic\r\nloci interactions. The MOEA method optimizes over the space of possible partial least\r\nsquares (PLS) regression QTL models and considers the conflicting objectives of model\r\nsimplicity versus model accuracy. By optimizing for minimal model complexity, MOEA\r\nhas the advantage of solving the over-fitting problem of conventional PLS models. The\r\neffectiveness of the method is confirmed by comparing the new method with Bayesian\r\nInterval Mapping approaches over a series of test cases where the optimal solutions are\r\nknown. This approach can be applied to many problems that arise in analysis of genomic data sets where the number of features far exceeds the number of observations and where\r\nfeatures can be highly correlated.
Loading....