Project of DeDaL was born in 2014 (or long before) in a context of cancer research in the lab of Computational Systems Biology of Cancer in Institut Curie in Paris in the mind of my research advisor Andrei Zinovyev.
The idea was to develop a tool that would combine multidimensional meta data associated with network and network itself in a form of an intelligent layout. We achieved it through performing PCA and Elastic Map (non-linear version of PCA) on the meta dataset and mapping network connexions into this MDS layout.
The difference between DeDaL and other MDS layouts available, it that DeDaL uses meta data and not the network structure in order to perform the layout. It also contains additional and more advanced features like network smoothing, alignment and morphing. DeDaL helps to interpret and visualize multidimensional data associated with network.
Today, DeDaL is a piece of software, I developed in Java together with A. Zinovyev, that can be used in Cytoscape 3.0 framework. It has been applied to a number of biological data which resulted in publication in BMC Systems Biology. Together with DeDaL website that contains full description and step-by-step tutorial and github repository, the software is well documented and open.
However, DeDaL finds its application not only in biological data. Here, I post a little demo of DeDaL applied to International Currencies 1890-1910 data from Flandreau, M. and C. Jobst to encourage you to try it out with different data.
‘For better or worse – and opinions differ on this – the choice of which language and which currency is made not on merit, or moral worth, but on size.’
(Kindleberger, 1967, p. 11)
The dataset gives measures for the international role of 45 currencies between 1890 and 1910 as well as a set of variables that can explain the relative position of these currencies.
Data have been collected for the years 1890, 1900 and 1910. The exchange structure matrix is complete from an empirical point of view, including virtually all currencies of the world in 1890-1910. It also has the nice conceptual feature to identify two different exchange markets for a given currency pair, corresponding to the home countries of the two currencies involved.
Here, using DeDaL I will try to see how much we can learn about currency market without previous knowledge in economy, based on the network and metadata associated with each country.
Details & code are omitted. Can be found in the full version of the post in my portfolio.
Organic layout of the imported network of currencies in Cytoscape 3.0. Edge width corresponds to distance, edge colors to bitrade, edge line type to colony variables.
I perform PCA with DeDaL in Cytoscape. Which results in the following layout:
From the PCA we can read that the factors best explaining variance in the data is thelevel of democracy, market rate (PC1) and coverage & debt (PC2). Therefore, we can see GBP is probably the most important factor. While DEU and FRA are somehow outliers. DEU being the highest in coverage and FRA in democracy index. On the opposite side, there is NZL, AUS and surprisingly CAN & ESP that have big debt. COL, JPN and ICH seem to have high market rate. We can also observe that the countries placed on the right have higher values in bitrade and are the most connected.
I also performed Elastic Map layout.
In this case Elastic Map layout is quite similar to PCA meaning the problem can be explained in with linear methods.
Another cool function of DeDaL is network morphing, in this example networks goes from PCA layout to organic layout. Organic layout network is rotated compared to original one to minimize bias due to rotation and mirroring.
It is possible to morph any two layouts (i.e. manually created topological layout into organic layout etc.)
From the data we can easily read the most important countries. Through PCA, it is possible to identify key variables for analysis of data variability. It cannot be compared, of course, to an expert analytical insight and models build with domain knowledge. However, DeDaL provides good and fast insight into the data and allows its visualization. It can be used as exploratory tool with different types of data associated with network. Don’t hesitate to download, like and share the tool!
Your imagination is the limit!
Full article: http://urszulaczerwinska.github.io/works/DeDaL.html
Data source: The international circulation of currencies 1890-1910
Flandreau, M. and C. Jobst (2005), ‘The ties that divide: a network analysis of the international monetary system 1890–1910’, Journal of Economic History, vol. 65(4).
Flandreau, M. and C. Jobst (2009), ‘The empirics of international currencies: Networks, history and persistence’, The Economic Journal, vol. 119(April).