Four general issues of concern have emerged: normalization, descriptive statistics, demographic categorization, and data censorship.
The proposed normalization method incorporates demographic and lifestyle variables by considering the distances (weights) among subjects based on their characteristics, and considers these weights in equating the quantiles of distributions of miRNA expressions.
The other big piece of this is that, like Heroku, Precog offers an add-on marketplace, in which developers can select from a variety of data enrichments, like sentiment analysis, demographics by email, or contact normalization.
In this paper, we incorporate subject-level covariates (specifically the demographic and lifestyle variables) in miRNA normalization, which has not been done before.
In this respect, we propose a novel approach that removes the non-biological differences across samples while incorporating the demographic and lifestyle characteristics of sample-subjects in normalization.
The application of conventional quantile normalization can mask important expression differences by ignoring demographic and environmental factors.
The application of the weighted quantile normalization method only increases power (compared to conventional quantile normalization) when demographic and lifestyle variables (on which between-subject distance is based) are relevant to the "treatment" group comparison of interest.
Thus, the weighted quantile normalization performs at least as well as the conventional quantile normalization when demographic and lifestyle variables are not associated with the treatment group.
These data may be helpful to consider not only the artificial intra- and inter-array differences, but also the differences caused by the demographic and lifestyle characteristics of patients, and to maintain only biological differences during the normalization procedure.
This suggests that their similarity is merely an artifact of the normalization, and that important differences in gene expression could be masked by ignoring demographic and environmental factors.
The algorithm of our proposed method is based on the quantile normalization algorithm from Bolstad et al. [ 9], but accounts for the weighted distance metrics of demographic and lifestyle variables.
