DNA expression microarrays may be the wrong tool to identify biological pathways
Correspondence: (Login to view email address)
- Oxford Radcliffe Hospitals NHS Trust, Oxford, United Kingdom
- Bioinformatics Institute, Agency for Science, Technology and Research, Singapore
- Environment and Health Department, Istituto Superiore di Sanità, Roma, Italy
- Document Type:
- Manuscript
- Date:
- Received 18 September 2007 13:09 UTC; Posted 19 September 2007
- Subjects:
- Biotechnology, Genetics & Genomics, Molecular Cell Biology, Pharmacology, Bioinformatics
- Abstract:
DNA microarray expression signatures are expected to provide new insights into patho- physiological pathways. Numerous variant statistical methods have been described for each step of the signal analysis. We employed five similar statistical tests on the same data set at the level of gene selection. Inter-test agreement for the identification of biological pathways in BioCarta, KEGG and Reactome was calculated using Cohen’s k- score. The identification of specific biological pathways showed only moderate agreement (0.30 < k < 0.79) between the analysis methods used. Pathways identified by microarrays must be treated cautiously as they vary according to the statistical method used.
Discussion
- Votes:
-
7 votes
- Comments:
-
6 comments
Information gain from microarrays is possible, if one wields the tool correctly. Two class prediction, as used in diagnosis, is as good as what experienced physicians achieve (see our other manuscript on this site). In other applications, the information gain must be seen more critically. One might consider using several tests in parallel, and then further investigate the top ranking genes of interest from all tests, rather than limit to only one test.
Totally agree with the comment that microarray data needs to be viewed with caution. Thats about choosing which genes to follow up after the microarray data. What about losing out on data in the name of “house-keeping genes”. Isogenes of housekeeping genes such as GAPDH are differentially regulated depending on cis-regulatory motifs, but are often not considered important due to their house-keeping nature in an attempt to look for genes specific to a certain stimulus or developmental stage or tissue. Whatever the statistical treatment of the signal for such genes they are often missed out for further inquiry into their role.
The main trouble is that microarray technology is pushed by commercial interests. I am very much in favour of commercial research, to the extent that I try to prescribe original drugs rather than generics to feed money back to companies actively involved in R&D.
Having said that, I am saddened to see how often good tools (like microarrays) become fashionable for a while, are used irresponsively (i.e. for the wrong tasks) and then fall into oblivion once the next big thing comes around.More specifically regarding housekeeping genes: I think their contribution to the state of biological systems is largely lost in high throughput analysis unless the conditions under which the samples were taken are very precisely documented. Usually, that is not the case, and their analysis on microarrays thus becomes post- hoc, which is usually regarded as dubious information.
Has anyone published the changes in a selection of say 5 housekeeping genes on say an hour-to-hour basis for a population of healthy humans monitoring food and water intake, and exercise?
It might be buried in a really early paper on microarrays… I suppose…?
@Bronwen Dekker: Not to my knowledge. I would think that early technology users would have focused their activities on some of the “buzzwords”- cancer, heart disease etc.
- (Login to share with a colleague)
Additional information
- License:
- This document is licensed to the public under the Creative Commons Attribution 2.5 License
- How to cite this document:
-
Mondry, Adrian, Loh, Marie, and Giuliani, Alessandro. DNA expression microarrays may be the wrong tool to identify biological pathways . Available from Nature Precedings <http://dx.doi.org/10.1038/npre.2007.1036.1> (2007)
- Version info:
-
Other versions of this document in Nature Precedings
None.
Other versions of this document elsewhere on the web
None known.
Tania Oh on 14 October 2007 19:01 UTC
Very interesting, and disturbing. Do you think one can rely on information gained from microarrays?