A new, improved and generalizable approach for the analysis of biological data generated by -omic platforms

• Published in: Journal of developmental origins of health and disease 2015-02, Vol.6 (1), p.17-26
• Main Authors: Pleasants, A. B., Wake, G. C., Shorten, P. R., Hassell-Sweatman, C. Z. W., McLean, C. A., Holbrook, J. D., Gluckman, P. D., Sheppard, A. M.
• Format: Article
• Language: English
• Subjects: Animals; Bioinformatics; Data Interpretation, Statistical; Deoxyribonucleic acid; DNA; DNA methylation; DNA Methylation - genetics; Epigenetics; Epigenomics - methods; Gene expression; Gene Expression Profiling - methods; Hypothesis testing; Mass spectrometry; Original Article; Probability Theory; Questioning; Regression Analysis; Scientific imaging; Sheep - genetics; Spectrometry, Mass, Matrix-Assisted Laser Desorption-Ionization - methods; Systems Biology - methods; Variables
• ISSN: 2040-1744; 2040-1752
• DOI: 10.1017/S2040174414000476

The principles embodied by the Developmental Origins of Health and Disease (DOHaD) view of ‘life history’ trajectory are increasingly underpinned by biological data arising from molecular-based epigenomic and transcriptomic studies. Although a number of ‘omic’ platforms are now routinely and widely used in biology and medicine, data generation is frequently confounded by a frequency distribution in the measurement error (an inherent feature of the chemistry and physics of the measurement process), which adversely affect the accuracy of estimation and thus, the inference of relationships to other biological measures such as phenotype. Based on empirical derived data, we have previously derived a probability density function to capture such errors and thus improve the confidence of estimation and inference based on such data. Here we use published open source data sets to calculate parameter values relevant to the most widely used epigenomic and transcriptomic technologies Then by using our own data sets, we illustrate the benefits of this approach by specific application, to measurement of DNA methylation in this instance, in cases where levels of methylation at specific genomic sites represents either (1) a response variable or (2) an independent variable. Further, we extend this formulation to consideration of the ‘bivariate’ case, in which the co-dependency of methylation levels at two distinct genomic sites is tested for biological significance. These tools not only allow greater accuracy of measurement and improved confidence of functional inference, but in the case of epigenomic data at least, also reveal otherwise cryptic information.