Analysis of Genotype-Phenotype Association using Genomic Informational Field Theory (GIFT)

TL;DR

This paper introduces a novel information-theoretic approach called GIFT to quantify genotype-phenotype relationships, especially effective for weak or unusual dependencies, generalizing traditional GWAS methods.

Contribution

It develops a field-theoretic framework for analyzing continuous phenotype data without assuming Gaussian distributions, enabling detection of subtle genotype effects.

Findings

Effective at extracting weak genotype-phenotype dependencies

Provides a method to calculate p-values for significance testing

Generalizes traditional GWAS analysis techniques

Abstract

We show how field- and information theory can be used to quantify the relationship between genotype and phenotype in cases where phenotype is a continuous variable. Given a sample population of phenotype measurements, from various known genotypes, we show how the ordering of phenotype data can lead to quantification of the effect of genotype. This method does not assume that the data has a Gaussian distribution, it is particularly effective at extracting weak and unusual dependencies of genotype on phenotype. However, in cases where data has a special form, (eg Gaussian), we observe that the effective phenotype field has a special form. We use asymptotic analysis to solve both the forward and reverse formulations of the problem. We show how $p$-values can be calculated so that the significance of correlation between phenotype and genotype can be quantified. This provides a significant generalisation of the traditional methods used in genome-wide association studies GWAS. We derive a field-strength which can be used to deduce how the correlations between genotype and phenotype, and their impact on the distribution of phenotypes.