About Nephroseq Processing and Pre-Computing

Data Curation

Nephroseq collects experimental details and sample metadata from supplemental manuscript data and by directly corresponding with authors of published work. This data is added to the database and analyzed. To facilitate consistent analysis, the Nephroseq team cleans and curates the data, mapping all sample metadata data to the Nephroseq Ontology. This ontology is a multi-threaded hierarchy of terms to describe the data, including biological traits, demographics, and clinical measures (e.g. GFR, blood pressure, weight).

Pre-computing

With the data, the Nephroseq team pre-computes analyses that provide values as follows, presented in alphabetical order: (Taken from Rhodes et al. 2007. Oncomine 3.0, Neoplasia 9, 166-180)

Differential expression analysis: Nephroseq pre-computes differential expression profiles using Student’s t-test for two-class differential expression analyses (e.g., disease versus normal tissues) and standard correlation methods for multi-class ordinal analyses (e.g. age, GFR). In a two-class analysis, genes are over-expressed if they are more highly expressed in one of the groups compared to the control. In multi-class analyses, genes are over-expressed if they display progressively increasing expression with increasing attribute values. p-values are corrected for multiple hypothesis testing using the false discovery rate method.

t-tests and p-values: The two-sample t-test compares the means (averages) of two independent samples. The null hypothesis is presumed to be true, which is that the two groups have the same average value. A t-test generates a p-value, which indicates the likelihood of the null hypothesis. If the chance is less than five percent (p= <.05), then by convention the null hypothesis is rejected. Instead there is a real, statistically significant difference between the means of the two groups.

p-values and effect size: p-values measure whether the difference in means between two groups is likely to occur solely by chance. Effect size measures the amount of difference between the groups and is reflected in the t-test statistics.

Fold change and p-values: Fold change is a valuable complement to p- values to assess large absolute differences between groups of samples of classes measured in an analysis. When relying on p-value to assess differences, there can be analyses where p-values are very significant due to a large number of samples and low sample variability within groups; but the actual difference in the magnitude, or fold change, between groups is low. Fold changes are reported as log base 2 values. A change of 1.5 or more is considered significant.

Standardized analyses: Standardized analyses are median-centered. Based on the Nephroseq Ontology, standardized analyses are performed on every dataset.