4. Pairwise Associations

Last modified by Erik S on 2019/05/29 16:29

In the All Pairs analysis, the statistical significance of each pairwise association is assessed using rank-ordered data and a statistical test appropriate to each data type pair, e.g. Fisher's test (categorical-categorical), F-statistic (continuous-continuous), and ANOVA (continuous-categorical).

Typical feature matrices represent data for hundreds of patients and tens of thousands of features.  A matrix with 30,000 features results in close to 450 million pairwise tests, out of which only the most significant ~2% are stored in a database for subsequent interactive exploration.  The results of each pairwise test includes the number of samples used in the test (excluding missing values), the Spearman correlation, and the -log10 (p-value).

The Genome-level View displays two types of pairwise associations:

  1. Significant associations between two molecular features which each map to genomic coordinates.
  2. Significant associations between one molecular feature which maps to a genomic coordinate and one which is unmapped.

1.  Associations with Two Features Mapped to Genomic Coordinates

Below is an example of associations with mapped coordinates:

Regulome Explorer All Pairs - Mapped Associations.png

The association arcs in the center indicate significant associations between two molecular features with the ends of the arcs positioned according to the genomic coordinates of each of the two features.  

Hovering over an association arc displays an Edge Card with information on the two features being associated.  Clicking on an arc pops up a Details window with a Data Plot tab that shows the scatter plot of the underlying data, and a MEDLINE tab that lists titles and abstracts of all articles in MEDLINE that jointly mention both the associated features.

   

Hovering over an endpoint of an association arc displays an Information Card.  Clicking on an endpoint brings up the Chromosome-Level view of the feature location.

2.  Associations with Features Lacking Genomic Coordinates

Below is an example of associations with one mapped coordinate and one unmapped coordinate:

Regulome Explorer All Pairs - Unmapped Associations.png

The circles in the innermost gray band indicate associations between one feature that has a genomic coordinate and another that does not, in this case, age at diagnosis.

Hovering over an association circle displays an Edge Card.  Clicking on a circle pops up a Details window with a Data Plot tab that shows the scatter plot of the underlying data, and a MEDLINE tab that lists titles and abstracts of all articles in MEDLINE that jointly mention both the associated features.

Tags:
Created by Erik S on 2019/05/29 16:29

Need help?

If you need help with XWiki you can contact: