Science Inventory

Normalization Methods and Statistical Inference to Identify Differentially Expressed MicroRNAs with an Application to a Residential Cohort Exposed to Environmental Toxins and Pollutants

Citation:

Pinkston, C., M. Cave, B. Chorley, M. Pavuk, L. Birnbaum, AND S. Rai. Normalization Methods and Statistical Inference to Identify Differentially Expressed MicroRNAs with an Application to a Residential Cohort Exposed to Environmental Toxins and Pollutants. Women in Statistics and Data Science Conference, Raleigh, NC, October 06 - 08, 2021. https://doi.org/10.23645/epacomptox.17430017

Impact/Purpose:

Presentation to the Women in Statistics and Data Science Conference October 2021. This presentation describes a method or microRNA normalization using a direct measurement method offered by Abcam. The importance of this method is put into the context of an environmental exposure study in a residential cohort.

Description:

Identification of disease through biomarkers in accessible biofluids, including microRNAs (miRs) – small single-stranded non-coding RNA – remains a promising field of research, especially in liver disease where biopsy is the gold standard. Multiple methods are available to measure miR signatures and there are many considerations for analysis, including normalization. Here, we use the Fireplex® platform technology by Abcam. Normalization methods used with other bioasssays – quantile normalization and rank normalization – significantly reduced technical variability seen in unnormalized data or normalized with manufacturer’s suggested methods GENorm, average normalization, or normalization based on investigator-selected miRs. We briefly describe and mathematically relate these approaches. Then, we demonstrate them using miR signatures of participants in the Anniston Community Health Survey, a residential cohort exposed to environmental toxins and pollutants. The effect of normalization methods with and without adjustment are thoroughly explained. Our results highlight the need to include relevant covariate data and to choose correct normalization methods.