Frontier Grants

UMIACS has recently started a Research Frontiers Grants program to support collaborative faculty research teams to develop and initiate research on new topics in interdisciplinary applications of computing that hold significant promise. The faculty teams are required to have UMIACS researchers work with researchers from outside UMIACS. The proposals undergo a rigorous, competitive, internal peer-review process for funding. The following proposals have been judged to be the most innovative and promising and are being currently funded by UMIACS in FY 12.

Enabling High-Throughput Macromolecular Structure Determination via Small Angle Scattering by David Fushman (Chemistry and Biochemistry), Ramani Duraiswami (Computer Science and UMIACS) , Nail A. Gumerov (UMIACS) and Konstantin Berlin (Chemistry and Biochemistry)
Abstract: We propose a completely new interdisciplinary project between researchers in UMIACS, Computer Science and Chemistry and Biochemistry that aims to develop high-throughput macromolecular structure determination methods based on Small Angle Scattering (SAS). The main computational bottleneck in the structure determination methods based on SAS is the computational complexity of ab initio SAS prediction. The project’s main goal is to reduce the computational complexity of this prediction by applying Fast Multipole Methods (FMM) and using GPU/hybrid architectures. These methods have successfully been used to tackle the mathematically similar problem of acoustic scattering. Improvement in the computational complexity of the prediction methods would have an immediate effect on the performance of most SAS-based algorithms. This in turn would have a major impact on the applicability of these methods in the fields of structural biology and high-throughput drug discovery by allowing scientists to tackle larger macromolecular systems, speed up experiments, and improve accuracy of molecular models. We anticipate that these improvements would extend our understanding of molecular mechanisms of cellular processes and foster new cures for diseases.

Computational approaches for identifying folding principles of eukaryotic genomes by Michelle Girvan (Physics), Sridhar Hannenhalli (Cell Biology and Molecular Genetics and CBCB/UMIACS), and Carl Kingsford (Computer Science and CBCB/UMIACS)
Abstract: The spatial arrangement of the molecules that encode an organism’s genome plays a significant role in its operation. Eukaryotic DNA, packed in chromosomes, occupies defined structures at different points in the cell cycle, and chromosomes preferentially locate both near other chromosomes and various regions of the nucleus. The investigation of these spatial relationships has led to insights about DNA replication, gene regulation, and cell differentiation and deviations from healthy spatial organization have been implicated in cancer. Reliable reconstruction of 3D chromosomal structure and its analysis is thus an important long-term goal. The goal of this 1-year proposal is to begin to develop better computational techniques for inferring chromosome structure from the “chromosome conformation capture” (3C) pairwise distance data and to create analysis methods that find biologically meaningful patterns in the inferred structures.

Correlation Analysis of Gene Expression and Programmed -1 Ribosomal Frameshifting and Their Effect on Predicting Survival Rates of Human Hepatocellular Carcinoma by Wael AbdAlmageed (UMIACS), Jonathan Dinman (Cell Biology and Molecular Genetics), and Larry Davis (Computer Science and UMIACS)
Abstract: Liver cancer is the third leading cause of cancer death worldwide with about 680,000 people projected to have died from it in 2007. The five-year survival rate of this devastating disease is less than 11% even in developed countries. Human Hepatocellular Carcinoma (HCC) is the most frequent type of liver cancer accounting for approximately 90% of all liver cancer cases. HCC is most prevalent in developing countries. Yet, its incidence is increasing in developed countries due to chronic infection with Hepatitis C virus (HCV) and resulting liver cirrhosis. Although, significant progress has been made in the HCC field, the molecular mechanisms and signaling pathways underlying HCC development and progression are still poorly understood. Therefore, it is critically important to better understand the molecular mechanisms underlying HCC and to develop new mathematical tools, which will allow analyzing large data sets without significant reduction of iterated data. The objective of the proposed research activity is to investigate the effect of DNA methylation on HCC survival rates and the correlation (if any) between the detected gene signatures and ribosomal frameshifting.

Directed Self-Organization in Multi-Robot Environments by Dana S. Nau (Computer Science) and Jim Reggia (Computer Science and UMIACS)
Abstract: Autonomous robot teams are becoming increasingly important for a wide variety of situations that involve interactions among autonomous agents. Generating robot plans for such scenarios is much more challenging than single-robot motion planning, because it requires reasoning about the potential strategies and future behaviors of other autonomous agents whose objectives may often conflict with one’s own. In two separate research projects, Dana Nau (Maryland Robotics Center) has investigated top-down techniques (game theory, planning) for multi-robot strategy generation, and Jim Reggia (UMIACS) has investigated bottom-up methods (neural computation, swarm intelligence) to provide fast collective reactivity to events as they occur. However, neither kind of approach is sufficient by itself for complicated interactions among teams of robots that have conflicting objectives. We propose exploratory research on systematic ways of integrating both top-down and bottom-up techniques for planning and carrying out multi-robot interactions.

Semantic Active Vision and Manipulation by Cornelia Fermüller (UMIACS), Yiannis Aloimonos (Computer Science and UMIACS), Hal Daumé (Computer Science and UMIACS), Jaydev P. Desai (Mechanical Engineering)
Absract: The goal of this project is to develop an interdisciplinary approach to advance our understanding of cognitive robotics, which is concerned with endowing robots with intelligent behavior by giving them perception, reasoning and manipulation capabilities to execute complex tasks in natural environments. We propose to develop an architecture that integrates in one unified system: vision, natural language processing and manipulation capabilities. Our approach consists of three components: Vision, Action, and Cognition. Within this framework we will develop modules that integrate Vision and Natural Language Processing to achieve complex visual recognition and reasoning, and modules that integrate Vision and Manipulation to achieve manipulation in unstructured environments. To demonstrate our framework we plan to assemble a hand-eye robotic system by combining existing hardware from our laboratories.