Briefings in Bioinformatics Advance Access originally published online on February 6, 2006
Briefings in Bioinformatics 2006 7(1):121-122; doi:10.1093/bib/bbk010
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
List mania: interpreting microarray results with the L2L server
ABSTRACTA new server for interpreting microarray results, list to list (L2L), is described. This tool offers a unique approach to understand the meaning of microarray results, based on comparing them to previously identified lists of differentially expressed genes. The usefulness of the server is demonstrated by studying differential expression in primary tumours versus metastases in colon cancer.
The need to make sense of transcription profiling results—often lists of hundreds of differentially regulated genes—has led to the development of several tools for gene list analysis. Today, a multitude of servers offer gene-ontology enrichment analysis, seeking biological features that are common to a high proportion of the differentially expressed genes. This type of analysis provides insight into the molecular functions or processes that are affected in response to a particular treatment.
Standing out in this crowd of tools is a simple server known as list to list (L2L). Developed by John Newman and Alan Weiner at the University of Washington (http://depts.washington.edu/l2l/), L2L takes an empirical approach toward understanding microarray results. It is based on the same concept as that behind sequence similarity searching with engines such as BLAST: if two experiments yield similar results, then this similarity has meaning. In sequence analysis, a common ancestor/function is deduced, providing insight into a gene's function. In microarray analysis, L2L offers a way to determine which experiments have produced a gene list similar to yours. Such a similarity implies similarity in the biological meaning and response of different treatments.
There are two components to the L2L server. First, there is a database of differentially expressed genes. Again, an ingeniously simple approach is used: rather than comparing expression levels across platforms and laboratories, L2L has taken authors at face value and has collected the conclusions of microarray experiments from multiple articles. Sounds easy, but considering that many publications offer these lists only in the form of images, the developers hint that a great deal of work went into manually entering lists from articles and web sites. The lists are supported by conversion files, which map probes from different microarray types.
The second component of the server is the enrichment analysis. Your gene list is compared to all lists, and genes with a statistically surprising number of common genes are reported through a simple and powerful table. (Here, I feel the authors may have taken simplicity a step too far, choosing approximated statistics to speed up the comparison.) Rich links are offered, connecting the resulting lists to external sources of information such as Genecards. As a bonus, a colour-coded table indicates the gene ontology categories to which the common genes belong. L2L also offers a gene ontology enrichment analysis, but it is not that unique.
I tried L2L for a mock question (Figure 1), picking at random a GEO set that compared the expression profile of primary colon tumours and metastases. One of the lists found to be similar to this set represents genes known to be upregulated by hypoxia. This could have been an interesting observation about metastasis except that hypoxia is already known to be associated with rapidly growing tumours. Still, considering the ease with which L2L uncovered this association, I am already trying the server for my favourite human and mouse gene lists.
|
National Institute of Biotechnology in the Negev Ben Gurion Univerity Israel
Submitted: December 7, 2005. Received (in revised form): December 13, 2005.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||