6. Complex Networks (CoNe) Group
• Develop new algorithms for network “mining”
• Use the algorithms to study real-world networks
– Focus on biological (molecular) networks
tmilenko@nd.edu
8. • Map “similar” nodes between different networks
in a way that conserves edges
Network alignment
tmilenko@nd.edu
9. • IsoRank family (B. Berger, MIT, 2007-2009)
• Our methods (2010):
– GRAAL
O. Kuchaiev, T. Milenkovic, V. Memisevic, W. Hayes, N. Przulj, "Topological network alignment
uncovers biological function and phylogeny", Journal of the Royal Society Interface, 2010.
– H-GRAAL
T. Milenkovic, W.L. Ng, W. Hayes, N. Przulj, “Optimal Network Alignment with Graphlet Degree
Vectors”, Cancer Informatics, 2010.
• MI-GRAAL (N. Przulj, ICL, 2011)
• GHOST (C. Kingsford, CMU, 2012)
• …
• Mix-and-match existing methods to improve them
– F.E. Faisal, H. Zhao, and T. Milenković, “Global Network Alignment In The Context
Of Aging”, IEEE/ACM TCBB, 2014. Also, in ACM-BCB 2013.
• MAGNA
– V. Saraph and T. Milenković, “MAGNA: Maximizing Accuracy of Global Network
Alignment”, Bioinformatics, 2014.
Network alignment
tmilenko@nd.edu
10. Mix-and-match existing methods to improve them
• Network alignment – algorithmic components:
1. Node cost function (NCF)
2. Alignment strategy (AS)
tmilenko@nd.edu
11. Mix-and-match existing methods to improve them
• Network alignment – algorithmic components:
1. Node cost function (NCF)
2. Alignment strategy (AS)
tmilenko@nd.edu
12. Mix-and-match existing methods to improve them
• Network alignment – algorithmic components:
1. Node cost function (NCF)
2. Alignment strategy (AS)
tmilenko@nd.edu
13. Mix-and-match existing methods to improve them
• Our goal: mix and match node cost functions and
alignment strategies of state-of-the-art methods
– MI-GRAAL and IsoRankN
• Fair evaluation framework
• New superior method? YES!
• Follow-up study on MI-GRAAL and GHOST
– Same conclusions
J. Crawford, Y. Sun, and T. Milenković, “Fair evaluation of global network aligners”, submitted, 2014.
tmilenko@nd.edu
14. MAGNA: Maximizing Accuracy
in Global Network Alignment
• Existing methods:
– Rapidly identify from all possible alignments the “high-
scoring” alignments with respect to total NCF
– Evaluate alignments with respect to edge conservation
– So, align similar nodes between networks hoping to
conserve many edges (after the alignment is constructed!)
tmilenko@nd.edu
15. • MAGNA:
– Directly optimizes edge conservation while the
alignment is constructed
– Can optimize any alignment quality measure
• E.g., a measure of both node and edge conservation
– Outperforms existing state-of-the-art methods
• In terms both node and edge conservation
• In terms of both topological and biological quality
MAGNA: Maximizing Accuracy
in Global Network Alignment
tmilenko@nd.edu
16. • Key idea behind MAGNA:
– Cross parent alignments into a superior child alignment
• Parent alignments:
– Alignments of existing methods
– Or completely random alignments
– Evolve as long as allowed by computational resources
Software: http://nd.edu/~cone/MAGNA
MAGNA: Maximizing Accuracy
in Global Network Alignment
tmilenko@nd.edu
17. • MAGNA on synthetic networks
MAGNA: Maximizing Accuracy
in Global Network Alignment
tmilenko@nd.edu
19. MAGNA: Maximizing Accuracy
in Global Network Alignment
tmilenko@nd.edu
• Running time comparison
– MAGNA is run on random alignments
20. Network alignment in aging
Current knowledge about human aging
• Human aging - hard to study experimentally
– Long lifespan
– Ethical constraints
• Hence, sequence-based knowledge transfer from model
species
• I.e., current “ground truth” - computational predictions
• But
– Not all genes in model species have human orthologs (vice versa)
– Importantly, genes’ “connectivities” typically ignored
tmilenko@nd.edu
21. • But, genes, i.e., their protein products, carry out
biological processes by interacting with each other
• And this is exactly what biological networks model!
– E.g., protein-protein interaction (PPI) networks
Network alignment in aging
tmilenko@nd.edu
22. Network alignment in aging
So, predict novel “ground truth” knowledge
about human aging via network alignment
tmilenko@nd.edu
23. • GenAge: ~250 genes (3!)
• We predict novel aging-related candidates:
– 792 genes in human
– 311, 522, and 544 genes in yeast, fruitfly, and worm
• Examples of validation
– Significant overlap with independent “ground truth” data
– Significantly enriched diseases:
• Brain tumor
• Prostate cancer
• Cancer
– Literature validation: 91% of our top scoring predictions
Network alignment in aging
tmilenko@nd.edu
24. Other projects in my group
• E.g., dynamic network analysis
F.E. Faisal and T. Milenković, “Dynamic networks reveal key players in aging”, Bioinformatics, 2014.
25. Other projects in my group
• E.g., network clustering
R.W. Solava, R.P. Michaels, and T. Milenkovic, “Graphlet-based edge clustering reveals pathogen-
interacting proteins”, Bioinformatics, ECCB 2012 (acceptance rate: 14%).
26. Other projects in my group
• E.g., network de-noising via link prediction
Y. Hulovatyy, R.W. Solava, and T. Milenkovic, “Revealing missing parts of the interactome via link
prediction”, PLOS ONE, 2014.
B. Yoo, H. Chen, F.E. Faisal, and T. Milenkovic, “Improving identification of key players in aging via
network de-noising”, ACM-BCB 2014.
28. Protein degradation (with Lan Huang)
R. Kaake, T. Milenkovic, N. Przulj, P. Kaiser, and L. Huang, Journal of Proteome Research, 2010.
C. Guerrero, T. Milenkovic, N. Przulj, J. J. Jones, P. Kaiser, L. Huang, PNAS, 2008.
29. Netsense (with Aaron Striegel)
How do individuals interact in the “always-on” environment?
L. Meng, T. Milenković, and A. Striegel, “Systematic Dynamic and Heterogeneous Analysis of Rich Social
Network Data,” Complex Networks V, 2014.
L. Meng, Y. Hulovatyy, A. Striegel, and T. Milenković, “On the Interplay Between Individuals' Evolving
Interaction Patterns and Traits in Dynamic Multiplex Social Networks”, submitted, 2014.
30. Physiological networks (with Sidney D’Mello)
Y. Hulovatyy, S. D’Mello, R. Calvo, T. Milenković, “Network Analysis Improves Interpretation of Affective
Physiological Data,” Journal of Complex Networks, 2014. Also, in IEEE Proceedings of Complex Networks,
2013.
31. Acknowledgements
• NSF CCF-1319469 ($453K)
• NSF EAGER CCF-1243295 ($208K)
• NIH R01 Supplement 3R01GM074807-07S1 ($249K)
• Google Faculty Research Award ($33K)
tmilenko@nd.edu
32. 25. B. Yoo, H. Chen, F.E. Faisal, T. Milenković, "Improving identification of key players in aging via network de-noising", ACM-BCB 2014.
24. L. Meng, Y. Hulovatyy, A. Striegel, T. Milenković, "On the Interplay Between Individuals' Evolving Interaction Patterns and Traits in
Dynamic Multiplex Social Networks", submitted, 2014.
23. V. Saraph, T. Milenković, "MAGNA: Maximizing Accuracy in Global Network Alignment", Bioinformatics,
DOI: 10.1093/bioinformatics/btu409, 2014.
22. Y. Hulovatyy, S. D'Mello, R.A. Calvo, T. Milenković, "Network Analysis Improves Interpretation of Affective Physiological Data", Journal of
Complex Networks, DOI: 10.1093/comnet/cnu032, 2014.
21. F.E. Faisal, H. Zhao, T. Milenković, "Global Network Alignment In The Context Of Aging", IEEE/ACM Transactions on Computational
Biology and Bioinformatics, DOI: 10.1109/TCBB.2014.2326862, 2014.
20. F.E. Faisal, T. Milenković, "Dynamic networks reveal key players in aging", Bioinformatics, DOI: 10.1093/bioinformatics/btu089, 2014.
19. L. Meng, T. Milenković, A. Striegel, "Systematic Dynamic and Heterogeneous Analysis of Rich Social Network Data", In Proceedings of
Complex Networks V, 2014 (acceptance rate: 25%).
18. A.K. Rider, T. Milenković, G.H. Siwo, R.S. Pinapati, S.J. Emrich, M.T. Ferdig, N.V. Chawla, "Networks’ Characteristics Matter for Systems
Biology," Network Science, accepted, to appear, 2014.
17. Y. Hulovatyy, R.W. Solava, T. Milenković, “Revealing missing parts of the interactome via link prediction”, PLOS ONE, 9(3), 2014.
16. Y. Hulovatyy, S. D'Mello, R.A. Calvo, T. Milenković, “Network Analysis Improves Interpretation of Affective Physiological Data”, In
Proceedings of Workshop on Complex Networks and their Applications at SITIS 2013, DOI: 10.1109/SITIS.2013.82.
15. T. Milenković, H. Zhao, and F.E. Faisal (2013), “Global Network Alignment In The Context Of Aging”, In Proceedings of ACM-BCB 2013
(acceptance rate: 28%).
14. R. Solava, R. Michaels, T. Milenković, “Graphlet-based edge clustering reveals pathogen-interacting genes,” In Proceedings of ECCB 2012,
Bioinformatics, 28 (18): i480-i486, 2012.
13. T. Milenković, V. Memišević, A. Bonato, N. Pržulj, “Dominating biological networks,” PLOS ONE, 6(8), 2011.
12. Arabidopsis Interactome Mapping Consortium, "Evidence for Network Evolution in an Arabidopsis Interactome Map," Science,
333(6042):601-607, 2011.
11. T. Milenković, W.L. Ng, W. Hayes, N. Pržulj, “Optimal network alignment with graphlet degree vectors,” Cancer Informatics, 9, 2010.
10. R. Kaake, T. Milenković, N. Pržulj, P. Kaiser, L. Huang, “Characterization of cell cycle specific protein interaction networks of the yeast
26S proteasome complex by the QTAX strategy,” Journal of Proteome Research, 9(4): 2016-2029, 2010.
9. H. Ho, T. Milenković, V. Memišević, J. Aruri, N. Pržulj, A.K. Ganesan, “Protein Interaction Network Topology Uncovers Melanogenesis
Regulatory Network Components Within Functional Genomics Datasets,” BMC Systems Biology, 4:84, 2010 (Highly Accessed).
8. V. Memišević, T. Milenković, N. Pržulj,“Complementarity of network and sequence structure in homologous proteins,” Journal of
Integrative Bioinformatics, 7(3):135, 2010.
7. Memišević, T. Milenković, N. Pržulj, “An integrative approach to modeling biological networks,” Journal of Integrative Bioinformatics,
7(3):135, 2010.
6. O. Kuchaiev, T. Milenković, V. Memišević, W. Hayes, N. Pržulj, “Topological network alignment uncovers biological function and
phylogeny,” Journal of the Royal Society Interface, 7:1341-1354, 2010.
5. T. Milenković, V. Memišević, A.K. Ganesan, N. Pržulj, “Systems-level cancer gene identification from protein interaction network topology
applied to melanogenesis-related functional genomics data,” Journal of the Royal Society Interface, 7(44), 423-437, 2010.
4. T. Milenković, I. Filippis, M. Lappe, N. Pržulj, “Optimized Null Model of Protein Structure Networks,” PLOS ONE, 4(6): e5967, 2009.
3. C. Guerrero, T. Milenković , N. Pržulj, P. Kaiser, L. Huang, “Characterization of the proteasome interaction network using a QTAX-based
tag-team strategy and protein interaction network analysis,” PNAS, 105(36), 13333-13338, 2008.
2. T. Milenković & N. Pržulj, “Uncovering Biological Network Function via Graphlet Degree Signatures,” Cancer Informatics, 2008:6 257-273,
2008 (Highly Visible).
1. T. Milenković, J. Lai,N. Pržulj, “GraphCrunch: A Tool for Large Network Analyses,” BMC Bioinformatics, 9:70, 2008 (Highly Accessed).
Notes de l'éditeur
Analogous to genomic sequence research, biological network research is expected to impact our biological understanding, since genes, that is their protein products, carry out most biological processes by interacting with other proteins, and this is exactly what biological networks model. Thus, computational prediction of protein function and the role of proteins in disease from PPI networks have received attention in the post-genomic era.