Example 3:

Metastatic vs. Primary Melanoma Data Set

We use Martini on a dataset that consists of 290 genes highly expressed in metastatic melanoma, and 899 genes highly expressed in primary melanomas, based on the microarray analysis from Riker, A.I. et al., 2008. Comparing with other tools we find that Martini can give more precise and detailed description of the biological processes occuring.

Keyword Enhancement of the Melanoma Data: :

The Keyword Enhancement Results of the Melanoma Data Set are strikingly interesting. Martini finds 264 keywords, nearly half of which are obviously related to mitosis. In addition, several specific gene names are being identified as significantly enhanced.
For this dataset, about 35% of the keywords returned by Martini are 'uninformative': some are too general (e.g. 'Topologies'), some relate to experimental techniques (e.g. 'Mass spectrometries') or model organisms (e.g. 'Saccharomyces'). Although these terms are overrepresented in the literature, they are clearly not helpful in understanding the biological process that distinguishes metastatic from primary melanoma.
Most interestingly, however, a third category of keywords occurs that are potentially informative, and are not obviously related to mitosis. In this category, Martini finds keywords related to skin cell types or diseases, and also keywords associated with polyploidy, which has long been recognized as associated with malignant melanoma (Whang-Peng, J., Chretien, P. & Knutsen, T., 1970).


We compare Martini's results using CoPub and ProfCom to analyze the melanoma datasets.
ProfCom does not find any significantly enhanced GO terms with these data.
CoPub finds 63 keywords, mostly related to mitosis, however CoPub returns almost only process names whereas Martini returned specific gene names as well.


Perhaps the most interesting keywords found by Martini are the names of several of the MAGE (melanoma-associated genes) family. These genes are normally expressed only in developing sperm, where they play a role in meiotic cell division. However, these genes are known to be expressed in melanoma (De Plaen, E. et al., 1994 and Simpson, A.J. et al., 2005). CoPub does not find the MAGE gene family, but unlike Martini finds the keywords 'meiosis' and 'spermatogenesis'.
In this case, as with the cell cycle dataset, Martini returns a precise and detailed description of the biological process occurring (the MAGE genes), whereas the keywords found by CoPub correctly point to a more general connection, in this case between melanoma and meiosis.