Statistical Comparisons in Lands
From Array Suite Wiki
How OmicSoft calculates statistical comparisons in Lands
OmicSoft Lands contain tens of thousands of pre-computed statistical comparisons from our curated data, enabling quick discovery of datasets where your genes of interest were differentially expressed; you can also retrieve and compare full statistical results within the Lands or by exporting. These same comparisons are explorable within IPA Analysis Match to discover the potential biological mechanisms that explain the observed changes in gene expression.
Types of Comparisons in OmicSoft Lands
Despite the wide variety of comparison contrasts that are set up, they can be categorized into a small set of Comparison Categories, such as Disease vs. Normal, Disease1 vs. Disease2, CellType1 vs CellType2, Tissue1 vs. Tissue2, Treatment vs. Control, and Resistant vs. Sensitive.
How to find details of statistical models
Each study in the database is carefully reviewed by OmicSoft’s curators. After reviewing the source publication, sample meta data and study description, a proper statistical model will be proposed for the study to generate comparisons (log2ratio and p values). These details are displayed within the Comparison metadata, which can be viewed when a Comparison is selected:
In >95% of cases, the full statistical model is described in ComparisonContrast; in rare cases a more complex full model was used but not explicitly listed to aid clarity of the core comparison.
Batch Effect correction in Land comparisons
Within project batch effects are handled by statistical models on case-by-case basis.
OmicSoft does not correct for between-project batch effect to avoid the potential removal of true biological signals. If you are concerned about batch effect when performing meta-analyses across projects, OmicSoft recommends exporting similar projects using the data download functions to export expression and metadata.