Mark Richard Gilder

Extraction of protein interaction information from unstructured text using a context-free grammar

MOTIVATION As research into disease pathology and cellular function continues to generate vast amounts of data pertaining to protein, gene and small molecule (PGSM) interactions, there exists a critical need to capture these results in structured formats allowing for computational analysis. Although many efforts have been made to create databases that store this information in computer readable form, populating these sources largely requires a manual process of interpreting and extracting interaction relationships from the biological research literature. Being able to efficiently and accurately automate the extraction of interactions from unstructured text, would greatly improve the content of these databases and provide a method for managing the continued growth of new literature being published. RESULTS In this paper, we describe a system for extracting PGSM interactions from unstructured text. By utilizing a lexical analyzer and context free grammar (CFG), we demonstrate that efficient parsers can be constructed for extracting these relationships from natural language with high rates of recall and precision. Our results show that this technique achieved a recall rate of 83.5% and a precision rate of 93.1% for recognizing PGSM names and a recall rate of 63.9% and a precision rate of 70.2% for extracting interactions between these entities. In contrast to other published techniques, the use of a CFG significantly reduces the complexities of natural language processing by focusing on domain specific structure as opposed to analyzing the semantics of a given language. Additionally, our approach provides a level of abstraction for adding new rules for extracting other types of biological relationships beyond PGSM relationships. AVAILABILITY The program and corpus are available by request from the authors.

p-PIC: Parallel power iteration clustering for big data

Power iteration clustering (PIC) is a newly developed clustering algorithm. It performs clustering by embedding data points in a low-dimensional subspace derived from the similarity matrix. Compared to traditional clustering algorithms, PIC is simple, fast and relatively scalable. However, it requires the data and its associated similarity matrix fit into memory, which makes the algorithm infeasible for big data applications. This paper attempts to expand PICs data scalability by implementing a parallel power iteration clustering (p-PIC). While this paper focuses on exploring different parallelization strategies and implementation details for minimizing computation and communication costs, we have also paid great attention to ensuring the algorithm works well on low-end commodity computers (COTS-based clusters and general purpose servers found at most commercial cloud providers). The experimental results demonstrate that the proposed p-PIC algorithm is highly scalable to both data and compute resources.

A GUI for Parallel Code Generation

Developing applications for parallel architectures is a very complicated and arduous task even for expert programmers. There are several issues that must be considered, i.e., the number of CPU’s available, vector processing capabilities, shared memory issues, process communications, and process synchronization, to name a few. Software developers have been trained to view the solution to a selected problem as a sequence of dependent steps or transitions which are applied to some input in an effort to produce the desired results. This approach to problem solving has been enforced by the traditional languages of C, Pascal, and Fortran. In this paper we describe the Interactive Visualization Tool (IVT) developed for the HICOR interactive parallelizing compiler. In particular, the IVT allows users to interactively manipulate a graphical representation of the program to be parallelized. Parameters describing the target architecture may be manipulated interactively to create what-if scenarios for architectural simulation.

Automatic source-code parallelization using HICOR objects

We show that by using an intermediate representation, which supports a formalized interface on which to construct parallelization tools, the mapping of the representation onto parallel architectures can be performed quickly and efficiently. An intermediate representation called HICOR (Hierarchical Intermediate Code Object Representation) is shown to facilitate the exploitation of parallel operations by providing an abstraction layer for performing high-level intermediate code analysis, scheduling, and code generation. An object-oriented design approach has been employed in the development of HICOR and associated tools. Source language constructs are transformed into specialized object classes. Inheritance properties provided by the object-oriented paradigm are utilized to provide a common interface to each object in the HICOR representation. It is this interface that provides the needed consistency and flexibility in which to construct tools that has since been lacking. In particular, a tool to performCompile-Time Scheduling is presented. The scheduling algorithm employed differs from traditional scheduling problems in that merging of tasks is performed to reduce both task creation and communication costs in determining the final schedule. Architectural parameters are provided as input to the heuristic allowing the scheduler to produce near-optimal results for a wide variety of MIMD architectures. Once the final schedule is determined theTarget Code Generator, also presented, is used to generate the corresponding target code. A prototype system has been implemented in C++ which incorporates the HICOR intermediate representation with the tools described. The target architectures include the Sun 630 MP/4, Sequent Symmetry S81, and Stardent Titan.