A network science project about a drugs interaction network.
To study the network I extract the following analytics, dividing the work in two parts:
- Part I:
- average degree, higher moments;
- average distance, diameter;
- CCDF and ML fitting;
- clustering coefficient and its distribution;
- assortativity;
- robustness to random failures and attacks;
- Part II:
- PageRank;
- HITS;
- ranking comparison: degree, betweenness, PageRank and HITS;
- communities detection;
- link-prediction.
The project has been developed and tested with the following tools:
- Python: version 3.7.5, using Pandas external library; IDE: Spyder 4.0.1.
- MATLAB: version R2019b, service update 5.
- Gephi: version 0.9.2.
The network, represented with Gephi.
The clustering coefficients in blue, ⟨C⟩ in red.
Assortativity of the network.
Random and adversary attack on the network.
PageRank and HITS comparison plot.
Conductance of the network.
Comparison rank table based on nodes degree, betweenness, PageRank and HITS score. I highlight with the same colours the same drugs among different indexes. Conductance of the network.*
Visual representation of the a) degree and b) betweenness ranking score c) PageRank and d) HITS ranking score (Gephi).
AUC and Precision (with L = 151) for CM, AA, RA, LP, Kats and PWR similarity.
For the papers, the license used is: Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
