With advances in high-throughput sequencing and relevant computational techniques, the genetic basis for a number of diseases is being revealed. The network of protein-protein interactions, or the interactome, allows us to understand the molecular mechanisms by which these genesrelate to the specific diseases. Here, I present our work on the discovery of the interactome using computational methods. There are massive volumes and varieties of partially characterized data related to this problem. Traditional approaches to modeling the data are not possible. We have developed suitable machine learning approaches to solve the open challenges in this domain, and predicted of interactions accurately for several genes. These predictions are advancing biology research: an example is the interaction between two genes OASL and RIG-I, which when studied further by experimental methods, led to the discovery that boosting a naturally occurring protein OASL may help the body to detect and fend off certain viral infections on its own. Such translation of the hundreds of predicted interactions to biology requires large scale coordinated human interactions through experts. We are using the web for collecting and authenticating such expert knowledge through crowd sourcing and for disseminating the novel predictions to suitable biologists.

This talk focuses on the iSECURE project, a research-in-progress that investigates methods and procedures for linking and integrating multi-media teaching materials (slides, videos and textbooks) for use in Information Security courses based on a security ontology. The semantic linking of multimedia materials allows students to search and compose multimedia and interactive course materials based on the digital contents, methods and learning styles, thus enabling flexible personalized learning. To achieve these goals, the project seeks to analyze, develop, and assess the following related research tasks: (1) Segmenting and annotating learning media based on their learning content, (2) Building a security ontology to be used for annotating and querying the learning objects, (3) Adapting lecture contents to student learning styles, and (4) Evaluating the usability and effectiveness of the linked multimedia learning system.

This NSF-funded research project involves collaboration among multiple institutions, namely, NJIT (Vincent Oria, Reza Curtmola, Jim Geller), CUNY College of Staten Island (Soon Ae Chun), and Montclair State University (Edina Renfro-Michel).

The problem of selecting the optimal packages of databases and related e-materials for a library is very complex due the ever changing nature of information system resources, the multiple conflicting goals, and the needs of the university faculty and students. In addition, library resources are intended for sharing to support faculty research network across multiple disciplines. Clearly, factors related to faculty collaboration in both face-to-face and virtual modes must be accounted for in the resource sharing model. Thus, problem not only involves multiple factors and multiple decision makers, but also multiple-decision levels for budget allocation and resource sharing constraints. This research proposes a modelling framework merging a multi-level/multi-divisional and multi-criteria models integrating capital budgeting, the knapsack problem, network optimization and AHP models to help multiple decision makers in their evaluation and selection decision. The framework offers a 3-step methodology based on the multifaceted combination of inter-related models to derive the best set of e-databases, e-journals and other applicable online resources subject to the budget, resources sharing rules and other university restrictions.

The emergence of online social networks and, with them, the phenomenon of "virality" wherein certain themes or pieces of content (or memes as they are known) spread explosively through the network, has attracted a lot of attention from computer science researchers in the last few years. The central problem in this area can be posed as a simple question: Can we predict which meme will go viral before it has actually gone viral? We began studying this problem by trying to identify metrics that discriminate between viral and non-viral topics and followed up by trying to actually solve the prediction problem. In this talk we will present some of the challenges that we encountered along the way and some of the results we obtained.