About me

I am a senior graduate student at MIT, affiliated with CSAIL, RLE and Broad. I am working with Prof. Muriel Medard and Prof. Manolis Kellis. I received my M.Sc. degree at MIT in 2010, with the best master of science thesis award in the EECS department. I received my B.Sc. at Sharif University of Technology in 2008, with the highest GPA in the university.

MORE

My research

My research interests include mathematical understanding of functions and structures of networks with applications in Computational Biology, Social Networks and Communications. I design network algorithms for fundamental network science problems, borrowing tools/concepts from Machine Learning, Optimization, Signal Processing, and Information Theory.

MORE

Current Projects

My current projects include "Network Infusion" to infer information sources in networks, "Spectral Alignment of Networks" to find optimal mappings across multiple networks, "Network Deconvolution" to remove transitive noise, "Network Integration", "Module Discovery across Multiple Networks" and "Biclustering using Message-Passing".  

MORE

Latest Projects

Case Thumb

Network Infusion
Several models exist for diffusion of signals across biological, social, or engineered networks. However, the inverse problem of identifying the source of such propagated information seems on the surface difficult, even in the presence of multiple network snapshots, and especially for the single-snapshot case, given the many overlapping paths in real-world networks. Mathematically, this problem can be undertaken using a diffusion kernel that represents diffusion processes in a given network, but computing this kernel is generally intractable. Here, we introduce an efficient algorithm to solve the inverse diffusion problem, which we term Network Infusion (NI). We prove the mean-field optimality of the proposed method in different setups, under the SI diffusion model. We apply Network Infusion (NI) to identify infusion hubs of several human diseases including T1D, Parkinson’s, MS, SLE, CVD, CAD, psoriasis, and schizophrenia, and show that, NI infers candidate infusion hubs that are biologically relevant, and often unidentifiable using the raw p-values. In a second application, we identify the news sources for 3553 stories in the Digg social news network, and validate our results based on annotated information that was not provided to our algorithm. 



Case Thumb

Spectral Network Alignment
Network alignment refers to the problem of finding a bijective mapping across vertices of two or more graphs, to maximize the number of overlapping edges and/or to minimize the number of mis-matched interactions across networks. Finding an optimal alignment mapping across networks is closely related to a quadratic assignment problem (QAP) which is NP hard. However, owing to numerous applications of network alignment in different areas, several algorithms have been designed to solve this problem approximately using linear or semidefinite relaxations, Bayesian frameworks, message passing, or other heuristics. Here, we introduce a network alignment algorithm called EigenAlign, inspired by eigenvector analysis which creates a relaxation for the underlying quadratic assignment problem. We show that, when certain technical conditions hold, the relaxation given by EigenAlign is asymptotically exact over Erdos-Renyi graphs, with high probability. Through simulations, we show that, our algorithm performs better than existent top performing network alignment methods based on message-passing, spectral decomposition and Lagrange relaxation. We use EigenAlign over gene regulatory networks across human, fly and worm species and infer conserved regulatory interactions across these species despite large evolutionary distances spanned. We find strong conservation of centrally-connected genes and some biological pathways, especially for human-fly comparisons.



Case Thumb

Biclustering using Message-Passing
Biclustering is the analog of clustering on a bipartite graph. Existent methods infer biclusters through local search strategies that find one cluster at a time; a common technique is to update the row memberships based on the current column memberships, and vice versa. We propose a biclustering algorithm that maximizes a global objective function using message passing. Our objective function closely approximates a general likelihood function, separating a cluster size penalty term into row- and column-count penalties. Because we use a global optimization framework, our approach excels at resolving the overlaps between biclusters, which are important features of biclusters in practice. Moreover, Expectation-Maximization can be used to learn the model parameters if they are unknown. 



Case Thumb

Network Deconvolution
Recognizing direct relationships between variables connected in a network is a pervasive problem in biological, social and information sciences as correlation-based networks contain numerous indirect relationships. We formulate this problem as the inverse of matrix transitive closure (in real field), and introduce an algorithm that solves it efficiently by exploiting eigenvector and eigenvalue decomposition and infinite-series sums. We demonstrate the effectiveness of our approach in several network applications: distinguishing direct targets in gene expression regulatory networks; recognizing directly interacting amino-acid residues for protein structure prediction from sequence alignments; and distinguishing strong collaborations in co-authorship social networks using connectivity information alone.



Case Thumb

Network Integration
Rewiring of regulatory networks is a primary driver of evolutionary change, but the paucity of comprehensive catalogs of regulatory genomics datasets has hindered its study in animal genomes. Here, we leverage genome-wide functional genomics datasets from ENCODE and modENCODE to infer and compare regulatory networks across human, fly, and worm. We use rank-based and likelihood-based integration approaches to combine feature-specific input networks which are constructed using genome-wide transcription factor binding profiles, conserved sequence motif instances and gene expression levels for multiple cell types. We found that, inferred regulatory interactions have significant overlap with known interactions in TRANSFAC, REDfly and EdgeDB benchmarks, for human, fly and worm species, indicating the robustness and accuracy of the used network inference pipeline.



Case Thumb

Inference of Conserved Modules across Networks
One robust way of network representation is to use their system-level information such as module structures. Modules are subsets of densely connected nodes in the network. To find conserved modules across networks, traditional methods first infer modules independently in each network and then identify conserved modules as the ones with significant overlaps across networks. However, these methods are likely to miss partly-rewired conserved modules, specially across the networks with low similarity because of the uncertainty in module assignments to nodes within individual networks. Here, we have developed a method based on a max-sum algorithm to infer modules over multiple networks, by maximizing their joint connectivity density.