Date of Award

5-12-2017

Degree Type

Thesis

Degree Name

Master of Public Health (MPH)

Department

Public Health

First Advisor

Ruiyan Luo

Second Advisor

Mehul Suthar

Abstract

INTRODUCTION:

Host genetic variants are known to impact infectious disease susceptibility and outcomes. However, the genes underlying these impacts are not well characterized. Multiplex bead assays (MBA) provide an affordable and rapid means of large scale screening for multiple phenotypic measures of immune response. Transformation and normalization approaches for MBA data have not been agreed upon, especially concerning screening applications.

AIM:

To compare preprocessing techniques in improving validity of quantitative loci trait analyses which utilize MBA phenotypic data with high levels of predictor technical variability using experimental data.

METHODS:

This research uses primary dendritic cells derived from a set of sixty-one genetically diverse mouse strains to study activation response of an antiviral pathway (RIG-I). Primary outcomes were IFNα and IFNβ secretion following RIG-I agonist treatment. Multiple transformation and normalization approaches were used to estimate true IFNα and IFNβ responses. Evaluation criteria included three quantitative measures (tail length, kurtosis, skewness) and three qualitative measures (QQ-plot, Bland-Altman plot, Mean-SD plot).

RESULTS:

Most qualitative measures and quantitative measures found log transformation with quantile normalization was most appropriate for normalizing data and reducing technical variability between batches and replicates. Unfortunately, no statistically significant (α = 0.90) loci of interest were identified with this normalized data.

DISCUSSION:

The data used to test these methods had notable limitations, mainly only two phenotypic markers and dramatic variability in both technical and biological replicates. While normalization and transformation techniques did ameliorate these issues, additional approaches such as mixed effects modeling may be able to further improve these types of analysis.

Share

COinS