BACKGROUND: In the era of precision oncology and publicly available datasets, the amount of information available for each patient case has dramatically increased. From clinical variables and PET-CT radiomics measures to DNA-variant and RNA expression profiles, such a wide variety of data presents a multitude of challenges. Large clinical datasets are subject to sparsely and/or inconsistently populated fields. Corresponding sequencing profiles can suffer from the problem of high-dimensionality, where making useful inferences can be difficult without correspondingly large numbers of instances. In this paper we report a novel deployment of machine learning techniques to handle data sparsity and high dimensionality, while evaluating potential biomarkers in the form of unsupervised transformations of RNA data. We apply preprocessing, MICE imputation, and sparse principal component analysis (SPCA) to improve the usability of more than 500 patient cases from the TCGA-HNSC dataset for enhancing future oncological decision support for Head and Neck Squamous Cell Carcinoma (HNSCC). RESULTS: Imputation was shown to improve prognostic ability of sparse clinical treatment variables. SPCA transformation of RNA expression variables reduced runtime for RNA-based models, though changes to classifier performance were not significant. Gene ontology enrichment analysis of gene sets associated with individual sparse principal components (SPCs) are also reported, showing that both high- and low-importance SPCs were associated with cell death pathways, though the high-importance gene sets were found to be associated with a wider variety of cancer-related biological processes. CONCLUSIONS: MICE imputation allowed us to impute missing values for clinically informative features, improving their overall importance for predicting two-year recurrence-free survival by incorporating variance from other clinical variables. Dimensionality reduction of RNA expression profiles via SPCA reduced both computation cost and model training/evaluation time without affecting classifier performance, allowing researchers to obtain experimental results much more quickly. SPCA simultaneously provided a convenient avenue for consideration of biological context via gene ontology enrichment analysis.

UNLABELLED: Cordova is an out-of-the-box solution for building and maintaining an online database of genetic variations integrated with pathogenicity prediction results from popular algorithms. Our primary motivation for developing this system is to aid researchers and clinician-scientists in determining the clinical significance of genetic variations. To achieve this goal, Cordova provides an interface to review and manually or computationally curate genetic variation data as well as share it for clinical diagnostics and the advancement of research. AVAILABILITY AND IMPLEMENTATION: Cordova is open source under the MIT license and is freely available for download at https://github.com/clcg/cordova.

Next-generation sequencing technology provides greater sequencing capacity to identify disease-causing variations. Currently there are many alignment tools and variation detection techniques, multiple SNP data sets used to filter out benign polymorphisms with the intent of enriching for disease-causing variations, and many tools and websites for annotating variations and performing pathogenicity prediction of variations. There is no consensus or standard with regards to which tools and corresponding parameter settings for each tool is appropriate for the different application of next-generation sequencing technologies for disease-causing variation identification. We have implemented an annotation pipeline that can easily use the output from different alignment tools (BLAT, BWA, BFAST, Bowtie) typically used to align sequence from next-generation sequencing and Sanger sequencing experiments. The Automated Sequence Analysis Pipeline (ASAP) reports variations using the HGVS standardized nomenclature[1] to facilitate comparison across tools and publications. To date, we have annotated the sequence from 13 exomes in an effort to identify disease-causing mutations.

AudioGene is a software system developed at the University of Iowa to classify and predict gene mutations that indicate causal or increased risk factors of disease. We focus on a concise example — the most likely genetic causes of a particular form of inherited hearing loss — ADNSHL. Whereas the cost and throughput involved in the collection of genomic data have advanced dramatically during the past decade, gathering and interpreting clinical information regarding disease diagnosis remains slow, costly and error-prone. AudioGene employs machine-learning techniques in an iterative procedure to prioritize probable genetic risk factors of disease, which are then verified with a molecular (wet lab) assay. In our current implementation AudioGene achieves 67% first-choice accuracy (versus 23% using a majority classifier). When the top three choices are considered, accuracy increases to 83%. This has numerous implications for reducing the cost of genetic screening as well as increasing the power of novel gene discovery efforts. While AudioGene is focused on hearing loss, the design and underlying mechanisms are generalizable to many other diseases including heart disease, cancer and mental illness.

MOTIVATION: Despite recent improvements in high-throughput or classic molecular biology approaches it is still challenging to identify intermediate resolution genomic variations (50 bp to 50 kb). Although array-based technologies can be used to detect copy number variations in the human genome they are biased to detect only the largest such deletions or duplications. Several studies have identified deletions or duplications occurring within a gene that directly cause or predispose to disease. We have developed a novel computational system, SPeeDD (system to prioritize deletions or duplications) that utilizes machine learning techniques to predict likely candidate regions that delete or duplicate exon(s) within a gene. RESULTS: Data mining and machine learning methods were applied to identify sequence features that were predictive of homologous recombination events. The logistic model tree (LMT) method yielded the best results. Sensitivity varied from 20% to 71.6% depending on the specific machine learning model used, but specificity exceeded 90% for all methods evaluated. In addition, the SPeeDD system successfully predicted and prioritized a recently published novel BRCA1 mutation. CONCLUSIONS: Results suggest that the SPeeDD system is effective at prioritizing candidate deletions and duplications within a gene. Use of SPeeDD enables more focused screening, which reduces the labor and associated costs of the molecular assays and may also lead to targeted design of new array-based screens to focus on candidate areas to accelerate the process of mutation discovery.

MOTIVATION: High accuracy of data always governs the large-scale gene discovery projects. The data should not only be trustworthy but should be correctly annotated for various features it contains. Sequence errors are inherent in single-pass sequences such as ESTs obtained from automated sequencing. These errors further complicate the automated identification of EST-related sequencing. A tool is required to prepare the data prior to advanced annotation processing and submission to public databases. RESULTS: This paper describes ESTprep, a program designed to preprocess expressed sequence tag (EST) sequences. It identifies the location of features present in ESTs and allows the sequence to pass only if it meets various quality criteria. Use of ESTprep has resulted in substantial improvement in accurate EST feature identification and fidelity of results submitted to GenBank. AVAILABILITY: The program is freely available for download from http://genome.uiowa.edu/pubsoft/software.html

MOTIVATION: In gene discovery projects based on EST sequencing, effective post-sequencing identification methods are important in determining tissue sources of ESTs within pooled cDNA libraries. In the past, such identification efforts have been characterized by higher than necessary failure rates due to the presence of errors within the subsequence containing the oligo tag intended to define the tissue source for each EST. RESULTS: A large-scale EST-based gene discovery program at The University of Iowa has led to the creation of a unique software method named UITagCreator usable in the creation of large sets of synthetic tissue identification tags. The identification tags provide error detection and correction capability and, in conjunction with automated annotation software, result in a substantial improvement in the accurate identification of the tissue source in the presence of sequencing and base-calling errors. These identification rates are favorable, relative to past paradigms. AVAILABILITY: The UITagCreator source code and installation instructions, along with detection software usable in concert with created tag sets, is freely available at http://genome.uiowa.edu/pubsoft/software.html CONTACT: [email protected]

New methods of electronic collaboration are needed to manage and reconcile the vast scientific knowledge made available through the experience of diverse experts. Fundamental research in capturing, managing, analyzing, and explaining information and making it available for use is seen as a high priority for software research worldwide. Therefore, we have developed a method and software for intelligent-agent assisted knowledge fusion systems to address these needs. This system utilizes intelligent agents to facilitate the construction and fusion of knowledge from distributed human experts working on a common knowledge-sharing task. We describe our multi-agent system prototype, the Internet-Based Distributed Ophthalmology Consensus System (I-DOCS) and how we are employing it for finding disease genes. We also describe some results of using I-DOCS with clinicians attempting to develop a common clinical nomenclature and classification system for disease subtypes.

Increasingly, feedback of measured run-time information is being used in the optimization of computation execution. This paper introduces a model relating the static view of a computation to its run-time variance that is useful in this context. A notion of uncertainty is then used to provide bounds on key scheduling parameters of the run-time computation. To illustrate the relationship between fidelity in measured information and minimum schedulable, grain size, we apply the bounds to three existing parallel architectures for the case of run-time variance caused by monitoring intrusion. We also outline a hybrid static-dynamic scheduling paradigm-SEDIA-that uses the model of uncertainty to optimize computation for execution in the presence of run-time variance from sources other than monitoring intrusion.

Execution monitoring plays a central role in most software development tools for parallel and distributed computer systems. However, such monitoring may induce delays that corrupt event timing. If this corruption can be quantified, it may be possible to determine the intrusion-free behavior. In this paper, we describe an algorithm that, given a safe timed Petri net model of the monitored software, can determine the uncorrupted timestamp values, i.e. those that would have been observed had the delays not been present. Monitoring conditions sufficient to ensure correct operation of the algorithm, and examples illustrating the algorithm's applicability to message-passing systems are also presented. This work is part of a larger effort aimed at identifying cost-effective software alternatives to custom hardware monitoring.>

Very large problems with high resource requirements of both computation and communication could be tackled with large numbers of workstations. However for LAN-based networks, contention becomes a limiting factor whereas latency appears to limit communication for WAN-based networks, nominally the Internet. We describe a model to analyze the gain of communication latency hiding by overlapping computation and communication. This model illustrates the limitations and opportunities of communication latency hiding for improving speedup of parallel computations that can be structured appropriately. Experiments show that latency hiding techniques increase the feasibility of parallel computing in high-latency networks of workstations across the Internet as well as in multiprocessor systems.

Execution monitoring plays a central role in most software development tools for parallel and distributed computer systems. However, such monitoring may induce delays that corrupt trace event timing. Recently we demonstrated that, given a safe timed Petri net model of monitored software, timestamp values that would have been observed had the delays not been present can sometimes be recovered from the corrupted trace. In this paper we discuss the architectural implications of applying this result to multi-processing systems.

The authors focus on techniques for enhancing the feasibility of using graphic visualization in analyzing the complexities of parallel software. The central drawback to applying such visual techniques is the overhead in developing analysis tools with flexible, customized views. The PARADISE system alleviates some of this design overhead by providing an abstract, object-oriented, visual modeling environment which expedites custom visual tool development. PARADISE is a visual tool which is used to develop other visual tools, or a meta-tool. The authors complement previous work on PARADISE by describing the philosophy behind its design, and how that philosophy leads to a methodology for constructing visual models which characterize parallel systems in general. The emphasis is on the crucial issues in utilizing visualization for parallel software development, and on how PARADISE deals with these issues.>

The Seamless parallel system being developed at the University of Iowa ECE Department provides a method for providing latency tolerance in physically-distributed memory systems utilizing “off-the-shelf” RISC CPU's without incurring the overhead of multithreading. Seamless encompasses an evolutionary new programming model emphasizing data locality that views communication as data movement rather than message passing I/O. A hardware Locality Manager is added to each processing element to perform this data movement concurrently with computation.

An approach for visually specifying parallel/distributed software using Petri nets (PNs) extend with transition enabling functions (TEFs) is investigated. The approach is demonstrated to be useful in the specification of decision-making activities that control distributed computing systems. PNs are employed because of their highly visual nature that can give insight into the nature of the controller of such a system and because of their analytical properties. In order to increase the expressive power of PNs, the extension of TEFs is used. The main focus is the specification and analysis of parallel/distributed software and systems. A key element of this approach is a set of rules derived to automatically transform such an extended net into a basic PN. Once the rules have been applied to transform the specification, analytical methods can be used to investigate characteristic properties of the system and validate correct operation.>

As multiprocessor software solutions find their way into critical applications such as avionics as a means of increasing performance and reliability, the need for both debugging and testing tools increases dramatically. The authors focus on the gathering of accurate run-time information providing support for both debugging and testing of embedded multiprocessor systems by using run-time identification of simple events and postexecution identification of compound events. Additionally, they propose a way of combining specification-based (black-box) system testing and automatic monitoring to substantially reduce the complexity, tedium, and cost of complete structural (white-box) testing.>

The problem of using the idle cycles of a number of high-performance workstations, interconnected by a high-speed network, for solving computationally intensive tasks is discussed. The classes of distributed applications examined require some form of synchronization among the sub-tasks, hence the need for coscheduling to guarantee that sub-tasks start at the same time and execute at the same pace on a group of workstations. A model of the system that allows the definition of an objective function to be maximized is presented. Then a quadratic time and linear space algorithm is derived for computing the optimal coscheduling.>

This paper discusses the following: effective instrumentation as the key to effective monitoring; distributed monitoring systems as a basis for general purpose distributed multiprocessors; problems and prospects; performance monitoring, fact and fancy; and software aspects.>

A novel approach for visually specifying parallel/distributed software using extended Petri-nets (PNS) is investigated. The approach is demonstrated to be useful in the specification of decision-making activities that control distributed computing systems. PNs are employed because of their highly visual nature that can give insight into the nature of the controller of such a system and because of their well-known analytical properties. In order to increase the expressive power of PNs, the extension of transition enabling functions is used. The main focus is a set of automatic transformations derived to transform such an extended net into a basic PN. Then, well-known analytical methods can be used to investigate characteristic properties of the system and validate correct operation. A graphical programming tool (XPAT) has been built to automate the visual programming process as well as the transformations.>

The design of a debugging and performance analysis system that includes a specification language for process-level events and hardware for nonintrusive identification of these events during the execution of parallel and distributed application for a nonshared memory system is presented. The design is based on a formal event/action model and a layered architecture model that have been previously presented. Background, related work, and specification, and identification of events are discussed.>

Experimentation aimed at determining the potential benefit of mixed-mode SIMD/MIMD (single instruction, multiple data/multiple instruction, multiple data) parallel architectures is reported. The experimentation is based on timing measurements made on the PASM (Partitionable SIMD/MIMD) system prototype utilizing carefully coded synthetic variations of a well-known algorithm. The synthetic algorithms used to measure and evaluate this system were based on bitonic sorting of sequences stored in the processing elements. This computation was mapped to both the hybrids of the SIMD and MIMD modes. The computations were coded in these four ways and experiments were performed that explore the tradeoffs among them. The results of these experiments are presented and discussed, with special consideration of the effects of the system's architecture. The goal is to obtain implementation-independent analyses of the attributes of mixed-mode parallel processing with respect to the computational characteristics of the application being examined.>

A modeling technique for distributed computation based on a combination of directed graphs and finite automata is described. The paradigm of distributed decision-making (DDM) is used to illustrate the technique for its two primary purposes: providing a standard specification mechanism for different algorithms for solving the same problem and providing a common mechanism for objective quantitative evaluation and comparison of alternative DDM algorithms. This is accomplished through the definition of the terms performance and efficiency as they relate to the domain of DDM. The two terms, which have precise meanings with respect to the analysis of sequential algorithms, currently lack a common interpretation in the environment of DDM. In particular, they need to be expressed in terms of the information movement necessary to share state information. The method has been used extensively to conduct analyses of several distribution scheduling algorithms. This paper focuses on the model specification properties.>

An experimental analysis of the architecture of an SIMD/MIMD parallel processing system is presented. Detailed implementations of parallel fast Fourier transform (FFT) programs were used to examine the performance of the prototype of the PASM (Partitionable SIMD/MIMD) parallel processing system. Detailed execution-time measurements using specialized timing hardware were made for the complete FFT and for components of SIMD, MIMD, and barrier-synchronized MIMD implementations. The component measurements isolated the effects of floating-point arithmetic operations, interconnection network transfer operations, and program control overhead. The measurements allow an accurate extrapolation of the execution time, speedup, and efficiency of the MIMD, SIMD, and barrier-synchronized MIMD programs to a full 1024-processor PASM system. This constitutes one of the first results of this kind, in which controlled experiments on fixed hardware were used to make comparisons of these fundamental modes of computing. Overall, the experimental results demonstrate the value of mixed-mode SIMD/MIMD computing and its suitability for computational intensive algorithms such as the FET.>

Several aspects of the multidimensional problem of providing monitoring support for the debugging and performance analysis of distributed and parallel systems and parallel applications are presented. The key elements of the solution to the problem are a formal event-action model at the process level and a layered architectural model. The application of the event-action model to the development of the abstract layered architectural model is shown, as well as a proposed implementation sketch of a nonintrusive monitor for the PASM parallel processing system.>

The problem of efficient task migration in hypercube multiprocessors is addressed. Task migration is useful for subcube reallocation, dynamic task sizing, load balancing, and support of fault tolerance. Efficient routing algorithms for large messages can increase the benefits obtained from migration. While pipelined communication techniques allow a maximum reduction in communication time, hardware costs for these solutions may be extreme. Current technology limits the extent to which communication links can be used simultaneously, and therefore the architectural support of pipelined communications in hypercubes. The proposed centralized algorithm applies an optimization heuristic in an idealized, static, nonpipelined setting and further uses retrograde motion to maximize the utilization of available bandwidth. The algorithm is applied to a dynamic system, and simulation studies evaluate the actual performance. The results provide a basis for extension and enhancement of the existing algorithm to a distributed routing algorithm.>

A detailed study is presented of parallel fast Fourier transform programs executing on the prototype of the PASM parallel processing system. Detailed execution time measurements using specialized timing hardware were made for complete fast Fourier transforms (FFTs) and for SIMD, MIMD, and hybrid SIMD/MIMD implementations. The component measurements isolated the effects of floating-point operations, interconnection network transfer operations, and program control overhead. Using these detailed measurements, an expression for projecting the execution time for an M-point FFT executing on M/2 PASM processing elements is derived. This expression is then used to obtain an accurate extrapolation of the execution time and speedup of the MIMD, SIMD, and hybrid programs to a full 1024-processor PASM system. Overall, the experimental results demonstrate the value of the multimode capability of PASM and the stability of PASM for computationally intensive algorithms such as the FFT.>

It is suggested that, while high-level-language and OS abstractions have evolved gradually and intelligently over the years with respect to single-instruction, single-data stream machines, these abstractions should not necessarily be provided in early generations of parallel supercomputers. Experience with low-level programming must precede the definition and evolution of abstractions for such machines. Future systems will naturally evolve to the correct level of abstraction, but experimentation with current machines must proceed as much as possible at levels that allow direct examination and evaluation of machine characteristics. Further, new software and hardware techniques should be developed to provide services that will be necessary, such as secure access to multiuser devices.>

Experimental analyses of an implementation of an SIMD algorithm for recursive digital filtering using the partitionable SIMD/MIMD (PASM) parallel processing system prototype at Purdue University are presented. The algorithm used easily generalizes to use N processing elements (PEs). Timing-based analyses are made based on a four-PE version by examining the following constituent execution times: microcontroller execution time, PE execution time, broadcast communication time, and the execution time of five additional phases in the recursive digital filtering summation calculation. Broadcast execution time was found to account for roughly 44% of the total execution time, and the implication of this is discussed for larger problem sizes and machine sizes. The total measured execution time is verified through summation of execution times for the various components of the algorithm.>

One measure of the usefulness of a general-purpose distributed computing system is the system's ability to provide a level of performance commensurate to the degree of multiplicity of resources present in the system. A taxonomy of approaches to the resource management problem is presented in an attempt to provide a common terminology and classification mechanism necessary in addressing this problem. The taxonomy, while presented and discussed in terms of distributed scheduling, is also applicable to most types of resource management.>

An examination is made of the effects of response and stability on scheduling algorithms for general-purpose distributed computing systems. Response characterizes the time required, following a perturbation in the system state, to reach a new equilibrium state. Stability is a measure of the ability of a mechanism to detect when the effects of further actions will not improve the system state as defined by a user-defined objective. These results have implications for distributed computations in general. Analysis is based on formal communicating finite automata models of two distinct approaches to the scheduling problem, each using the objective of global optimal load balancing. The results indicate that absolute stability is not always necessary in dynamic systems for the same reasons that relatively small amounts of instability are tolerated in the design of analog control systems. It is shown that response is a very important first-order metric of dynamic scheduling behavior, and that response and stability are related.>