Opened 15 years ago
Last modified 15 years ago
#1396 new enhancement
Using item properties as experimental factors - how could this be implemented? — at Version 1
| Reported by: | Johan Vallon-Christersson | Owned by: | everyone |
|---|---|---|---|
| Priority: | major | Milestone: | BASE Future Release |
| Component: | web | Version: | |
| Keywords: | Cc: |
Description (last modified by )
Would it be possible to somehow make item properties (or some selected item properties) available for use as experimental factors?
--
Use case: An experiment include 20 rawbioassays, e.g., gene expression analysis of lung cancer biopsies. The raw bioassays are derived from 20 samples (the biopsies) annotated with a number of biologically and clinically relevant annotations, e.g., lung cancer type, tumor size, and so on. These annotations are used as experimental factors and the user can analyze differences in gene expression between the annotation type values.
However, the user is concerned that a number of technical parameters, not related to the biology of lung cancer, might be confounding factors and affect the analysis. For example, if insufficient care has been devoted to the experimental design, technical factors, such as scanner (the hardware used to scan array slides), might correlate with experimental factors derived from the samples, e.g., if two different scanners have been used. That is, a user might like to (or rather should always carefully) examine whether there is a correlation between gene expression and technical parameters such as scanner, array batch, hybridization date, protocols, and so on.
--
Naturally, a user could export item properties, e.g., scanner, and then annotate the scans with scanner name using an annotation type. This is, however, not a particularly pretty solution as: the information is redundant (inputted twice), and there is no association between the property and the annotation type, i.e., if scan properties are changes the annotation values remain unaltered.
Some thoughts:
- Could a special form of annotation types be connected to item properties so that annotation values are derived from the item property value?
- Alternatively, could item properties be used directly as experimental factors?
- Is there already support for something similar?
It might be worthwhile to discuss this further and come up with a solution that enable users to analyze data in relation to item properties that may constitute confounding factors in experiments.
