Dissertação de Mestrado

Single and Multi-Objective Epistasis Scoring: A Matter of Frequency

Madalena Carvalho de Azevedo Moreira — 2020

Informações chave

Autores:

Madalena Carvalho de Azevedo Moreira (Madalena Carvalho de Azevedo Moreira)

Orientadores:

Sergio Santander-Jiménez; Aleksandar Ilic (Aleksandar Ilic)

Publicado em

30 de setembro de 2020

Resumo

Epistasis detection studies focus on finding interactions between Single Nucleotide Polymorphisms (SNPs) that may be linked with susceptibility to and development of complex disease states. Since existing search and score methods for detecting significant SNP combinations focus heavily on the search algorithm, the question of how to best evaluate the epistatic contribution of these interactions still lacks a satisfactory answer. This dissertation proposes a novel methodology for evaluating the performance of six widely used objective functions for epistasis detection based on genotype distribution in the dataset. This analysis reveals a correlation between high scoring power and extreme frequency table values, defined by two parameters. The first is based on genotypes with extreme differences between counts of cases and controls and the second is a simplified heritability formulation taking into account the total number of observations and cases for each genotype. A threshold is defined for these parameters above which, for the simulated datasets analysed, an objective function can correctly and single-handedly identify associated SNP combinations. Below this threshold, the combination of two and three complementary objective functions in a multi-objective approach demonstrates an increase in scoring power. This frequency table based approach is innovative in the sense that there is not currently a defined methodology for evaluating and comparing the performance of objective functions. The defined parameters can be applied to real datasets, representing a first step in validating the results of existing epistasis detection methods and promoting the choice of the least complex scoring method possible for specific datasets.

Detalhes da publicação

Autores da comunidade :

Orientadores desta instituição:

Domínio Científico (FOS)

electrical-engineering-electronic-engineering-information-engineering - Engenharia Eletrotécnica, Eletrónica e Informática

Idioma da publicação (código ISO)

eng - Inglês

Acesso à publicação:

Embargo levantado

Data do fim do embargo:

26 de julho de 2021

Nome da instituição

Instituto Superior Técnico