HotNet is an algorithm for finding significantly altered subnetworks in a large gene interaction network. While originally developed for use with cancer mutation data, the current release of HotNet also supports any application in which scores can be assigned to genes in the network (this version of the algorithm is called generalizedHotNet - see Vandin et al., PSB 2012).

People (strict random order): Fabio Vandin, Hsin-Ta Wu, Edward Rice, Layla Oesper, Adrien Deschamps, Max Leiserson, Jonathan Eldridge, Jason Schum, Eli Upfal, Ben Raphael.


The HotNet algorithm is described in the following publications:

  • F. Vandin, E. Upfal, and B.J. Raphael. (2010) Algorithms for Detecting Significantly Mutated Pathways in Cancer. Proceedings of the 14th Annual International Conference on Research in Computational Molecular Biology (RECOMB 2010).
  • F. Vandin, E. Upfal, and B.J. Raphael. (2011) Algorithms for Detecting Significantly Mutated Pathways in Cancer. Journal of Computational Biology. 18(3):507-22.

Moreover, we have used HotNet in the following publications:

  • The Cancer Genome Atlas Research Network (2011). Integrated genomic analyses of ovarian carcinoma. Nature. 474:609-615.
  • F. Vandin, P. Clay, E. Upfal, and B. J. Raphael (2012) Discovery of Mutated Subnetworks Associated with Clinical Data in Cancer. In Pacific Symposium on Biocomputing (PSB) 2012.
  • C. Grasso, Y.Wu, D. Robinson, X. Cao, S. Dhanasekaran, A. Khan, M. Quist, X. Jing, R. Lonigro, J.C. Brenner, I. Asangani, B. Ateeq, S. Chun, J. Siddiqui, L. Sam, M. Anstett, R. Mehra, J. Prensner, N. Palanisamy, G. Ryslik, F. Vandin, B. Raphael, L. Kunju, D. Rhodes, K. Pienta, A. M. Chinnaiyan, S.A. Tomlins. The Mutational Landscape of Lethal Castrate Resistant Prostate Cancer. Nature, 2012.
  • The Cancer Genome Atlas Research Network. Genomic and Epigenomic Landscapes of Adult De Novo Acute Myeloid Leukemia. New England Journal of Medicine, May 1st 2013.
  • The Cancer Genome Atlas Research Network. Comprehensive molecular characterization of clear cell renal cell carcinoma. Nature, 2013 Jul 4; 499(7456):43-9.


A new version of HotNet (v1.1.0; August 25th, 2013) is now available:

Detailed instructions for running HotNet are provided in the included README. For support, please see the HotNet Google Group.

You can also download previous versions of HotNet here.


Detailed instructions for running HotNet are provided in the README file in the corresponding release. Here we give instructions for buildling the influence matrix for a custom interaction network.

Building the influence matrix

HotNet uses an influence measure between two genes to find significant subnetworks. Since the influence depends only on the network, there is no reason to compute the influence everytime HotNet is run. Therefore, the influence matrix must be provided as input to HotNet. Pre-computed influence matrices are provided above for two common interaction networks, HPRD and iRefIndex. However, you can also compute the influence matrix for a custom network as follows:

Let A be the adjacency matrix of the undirected network (i.e, the entry of row i and column j is =1 if there is an interaction between i and j). The graph Laplacian L is then given by L = D-A, where D is a diagonal matrix with D(i,i)=degree of i in the network (and D(i,j)=0 if i different from j). To compute the influence at time t, you need to find the exponential of the matrix L*t (that corresponds to the heat kernel at time t). For HotNet to work, you need to assign the matrix the name 'Li', and then save it on file in Matlab format. In Matlab, given the laplacian matrix L and the time t, you can use:

save name_file.mat Li

You will also need to generate a gene name to index in this matrix mapping file, associating the row/column index to the gene name.

In the original publication [Vandin et al., JCB 2011] a different diffusion process was used to derive the influence, that in our tests gives results similar to the diffusion process currently used (described above). This diffusion kernel is described in:

  • Qi, Y, Suhail, Y, Lin, YY, Boeke, JD, Bader, JS (2008). Finding friends and enemies in an enemies-only network: a graph diffusion kernel for predicting novel genetic interactions and co-complex membership from yeast genetic interactions. Genome Res., 18, 12:1991-2004.

That kernel requires the inversion of a shifted version of the laplacian matrix (refer to the original publication or to [Vandin et al., JCB 2011] for details).


New visualization tools are coming soon! Please follow the HotNet Google Group for updates.