Signaling and metabolic pathways, which consist of a series of reactions producing target molecules from source compounds, are cornerstones of cellular biology. The cellular reaction networks containing such pathways can be precisely modeled by directed hypergraphs, where each reaction corresponds to a hyperedge, directed from its set of reactants to its set of products. Given such a network represented by a directed hypergraph, inferring the most likely set of reactions that produce a given target from a given set of sources corresponds to finding a shortest hyperpath, which is NP-complete. The best methods currently available for shortest hyperpaths either offer no guarantee of optimality, or exclude hyperpaths containing cycles even though cycles are abundant in real biological pathways. We derive a novel graph-theoretic characterization of hyperpaths, leveraged in a new formulation of the general shortest hyperpath problem as an integer linear program that for the first time handles hyperpaths containing cycles, and present a novel cutting-plane algorithm that can solve this integer program to optimality in practice. This represents a major advance over the best prior exact algorithm, which was limited to acyclic hyperpaths (and hence fails to find a solution for the many biological instances where all hyperpaths are in fact cyclic). In comprehensive experiments over thousands of instances from the standard NCI-PID and Reactome databases, we demonstrate that our cutting-plane algorithm quickly finds an optimal hyperpath, with a median running-time of under ten seconds and a maximum time of around thirty minutes, even on large instances with many thousands of reactions. Source code implementing our cutting-plane algorithm for shortest hyperpaths in a new tool called Mmunin is available free for research use at http://mmunin.cs.arizona.edu .

Cell signaling pathways, which are a series of reactions that start at receptors and end at transcription factors, are basic to systems biology. Properly modeling the reactions in such pathways requires directed hypergraphs, where an edge is now directed between two sets of vertices. Inferring a pathway by the most parsimonious series of reactions then corresponds to finding a shortest hyperpath in a directed hypergraph, which is NP-complete. The state of the art for shortest hyperpaths in cell-signaling hypergraphs solves a mixed-integer linear program to find an optimal hyperpath that is restricted to be acyclic, and offers no efficiency guarantees. We present for the first time a heuristic for general shortest hyperpaths that properly handles cycles, and is guaranteed to be efficient. Its accuracy is demonstrated through exhaustive experiments on all instances from the standard NCI-PID and Reactome pathway databases, which show the heuristic finds a hyperpath that matches the state-of-the-art mixed-integer linear program on over 99% of all instances that are acyclic. On instances where only cyclic hyperpaths exist, the heuristic surpasses the state-of-the-art, which finds no solution; on every such cyclic instance, enumerating all possible hyperpaths shows that the solution found by the heuristic is in fact optimal.

Protein secondary structure prediction is a fundamental task in computational biology, basic to many bioinformatics workflows, with a diverse collection of tools currently available. An approach from machine learning with the potential to capitalize on such a collection is ensemble prediction, which runs multiple predictors and combines their predictions into one, output by the ensemble. We conduct a thorough study of seven different approaches to ensemble secondary structure prediction, several of which are novel, and show we can indeed obtain an ensemble method that significantly exceeds the accuracy of individual state-of-the-art tools. The best approaches build on a recent technique known as feature-based accuracy estimation, which estimates the unknown true accuracy of a prediction, here using features of both the prediction output and the internal state of the prediction method. In particular, a hybrid approach to ensemble prediction that leverages accuracy estimation is now the most accurate method currently available: on average over standard CASP and PDB benchmarks, it exceeds the state-of-the-art Q3 accuracy for 3-state prediction by nearly 4%, and exceeds the Q8 accuracy for 8-state prediction by more than 8%. A preliminary implementation of our approach to ensemble protein secondary structure prediction, in a new tool we call Ssylla, is available free for non-commercial use at ssylla.cs.arizona.edu.

Protein secondary structure prediction is a fundamental precursor to many bioinformatics tasks. Nearly all state-of-the-art tools when computing their secondary structure prediction do not explicitly leverage the vast number of proteins whose structure is known. Leveraging this additional information in a template-based method has the potential to significantly boost prediction accuracy. We present a new hybrid approach to secondary structure prediction that gains the advantages of both template-based and template-free methods. Our core template-based method is an algorithmic approach that uses metric-space nearest neighbor search over a template database of fixed-length amino-acid words to determine estimated class-membership probabilities for each residue in the protein. These probabilities are then input to a dynamic programming algorithm that finds a physically-valid maximum-likelihood prediction for the entire protein. Our hybrid approach exploits a novel accuracy estimator for our core method, that estimates the unknown true accuracy of its prediction, to discern when to switch between template-based and template-free methods. On challenging CASP benchmarks, the hybrid approach boosts the state-of-the-art Q8 accuracy by as much as 10%, and Q3 accuracy by as much as 3%, yielding the most accurate method currently available for both 3- and 8-state secondary structure prediction.

We describe the scientific goals and capabilities of the ANTARES project. This is a software infrastructure system designed to process time-domain alerts at the scale the Large Synoptic Survey Telescope will produce. This system will allow access to large-scale time-domain streams with real-time filters, machine learning, and other tools.

The unprecedented volume and rate of transient events that will be discovered by the Large Synoptic Survey Telescope (LSST) demand that the astronomical community update its follow-up paradigm. Alert-brokers-automated software system to sift through, characterize, annotate, and prioritize events for follow-up-will be critical tools for managing alert streams in the LSST era. The Arizona-NOAO Temporal Analysis and Response to Events System (ANTARES) is one such broker. In this work, we develop a machine learning pipeline to characterize and classify variable and transient sources only using the available multiband optical photometry. We describe three illustrative stages of the pipeline, serving the three goals of early, intermediate, and retrospective classification of alerts. The first takes the form of variable versus transient categorization, the second a multiclass typing of the combined variable and transient data set, and the third a purity-driven subtyping of a transient class. Although several similar algorithms have proven themselves in simulations, we validate their performance on real observations for the first time. We quantitatively evaluate our pipeline on sparse, unevenly sampled, heteroskedastic data from various existing observational campaigns, and demonstrate very competitive classification performance. We describe our progress toward adapting the pipeline developed in this work into a real-time broker working on live alert streams from time-domain surveys.

SummaryMeasuring execution time is one of the most used performance evaluation techniques in computer science research. Inaccurate measurements cannot be used for a fair performance comparison between programs. Despite the prevalence of its use, the intrinsic variability in the time measurement makes it hard to obtain repeatable and accurate timing results of a program running on an operating system. We propose a novel execution time measurement protocol (termed EMP) for measuring the execution time of a compute‐bound program on Linux, while minimizing that measurement's variability. During the development of execution time measurement protocol, we identified several factors that disturb execution time measurement. We introduce successive refinements to the protocol by addressing each of these factors, in concert, reducing variability by more than an order of magnitude. We also introduce a new visualization technique, what we term ‘dual‐execution scatter plot’ that highlights infrequent, long‐running daemons, differentiating them from frequent and/or short‐running daemons. Our empirical results show that the proposed protocol successfully achieves three major aspects—precision, accuracy, and scalability—in execution time measurement that can work for open‐source and proprietary software. Copyright © 2017 John Wiley & Sons, Ltd.

The Arizona-NOAO Temporal Analysis and Response to Events System (ANTARES) is a joint effort of NOAO and the Department of Computer Science at the University of Arizona to build prototype software to process alerts from time-domain surveys, especially LSST, to identify those alerts that must be followed up immediately. Value is added by annotating incoming alerts with existing information from previous surveys and compilations across the electromagnetic spectrum and from the history of past alerts. Comparison against a knowledge repository of properties and features of known or predicted kinds of variable phenomena is used for categorization. The architecture and algorithms being employed are described.

A1 Highlights from the tenth ISCB Student Council Symposium 2014 Farzana Rahman, Katie Wilkins, Annika Jacobsen, Alexander Junge, Esmeralda Vicedo, Dan DeBlasio, Anupama Jigisha, Tomás Di Domenico Genomics and Computational Biology Research Group, University of South Wales, UK; Computational Biology Department, Cornell University, USA; Centre for Integrative Bioinformatics (IBIVU), VU University Amsterdam, Netherlands; Center for non-coding RNA in Technology and Health, University of Copenhagen, Denmark; Department of Bioinformatics and Computational Biology, Fakultät für Informatik, Germany; Department of Computer Science, University of Arizona, USA; University of Geneva, Switzerland; Wellcome Trust/Cancer Research UK Gurdon Institute, University of Cambridge, UK E-mail: tdido@tdido.com.ar BMC Bioinformatics 2015, 16(Suppl 2):A1

The Arizona-NOAO Temporal Analysis and Response to Events System (ANTARES) is a joint project of the National Optical Astronomy Observatory and the Department of Computer Science at the University of Arizona. The goal is to build the software infrastructure necessary to process and filter alerts produced by time-domain surveys, with the ultimate source of such alerts being the Large Synoptic Survey Telescope (LSST). The ANTARES broker will add value to alerts by annotating them with information from external sources such as previous surveys from across the electromagnetic spectrum. In addition, the temporal history of annotated alerts will provide further annotation for analysis. These alerts will go through a cascade of filters to select interesting candidates. For the prototype, `interesting' is defined as the rarest or most unusual alert, but future systems will accommodate multiple filtering goals. The system is designed to be flexible, allowing users to access the stream at multiple points throughout the process, and to insert custom filters where necessary. We describe the basic architecture of ANTARES and the principles that will guide development and implementation.

Automatic design of synthetic gene circuits poses a significant challenge to synthetic biology, primarily due to the complexity of biological systems, and the lack of rigorous optimization methods that can cope with the combinatorial explosion as the number of biological parts increases. Current optimization methods for synthetic gene design rely on heuristic algorithms that are usually not deterministic, deliver sub-optimal solutions, and provide no guaranties on convergence or error bounds. Here, we introduce an optimization framework for the problem of part selection in synthetic gene circuits that is based on mixed integer non-linear programming (MINLP), which is a deterministic method that finds the globally optimal solution and guarantees convergence in finite time. Given a synthetic gene circuit, a library of characterized parts, and user-defined constraints, our method can find the optimal selection of parts that satisfy the constraints and best approximates the objective function given by the user. We evaluated the proposed method in the design of three synthetic circuits (a toggle switch, a transcriptional cascade, and a band detector), with both experimentally constructed and synthetic promoter libraries. Scalability and robustness analysis shows that the proposed framework scales well with the library size and the solution space. The work described here is a step towards a unifying, realistic framework for the automated design of biological circuits.

A central focus of computational biology is to organize and make use of vast stores of molecular sequence data. Two of the most studied and fundamental problems in the field are sequence alignment and phylogeny inference. The problem of multiple sequence alignment is to take a set of DNA, RNA, or protein sequences and identify related segments of these sequences. Perhaps the most common use of alignments of multiple sequences is as input for methods designed to infer a phylogeny, or tree describing the evolutionary history of the sequences. The two problems are circularly related: standard phylogeny inference methods take a multiple sequence alignment as input, while computation of a rudimentary phylogeny is a step in the standard multiple sequence alignment method. Efficient computation of high-quality alignments, and of high-quality phylogenies based on those alignments, are both open problems in the field of computational biology. The first part of the dissertation gives details of my efforts to identify a best-of-breed method for each stage of the standard form-and-polish heuristic for aligning multiple sequences; the result of these efforts is a tool, called Opal, that achieves state-of-the-art 84.7% accuracy on the BAliBASE alignment benchmark. The second part of the dissertation describes a new algorithm that dramatically increases the speed and scalability of a common method for phylogeny inference called neighbor-joining; this algorithm is implemented in a new tool, called NINJA, which is more than an order of magnitude faster than a very fast implementation of the canonical algorithm, for example building a tree on 218,000 sequences in under 6 days using a single processor computer.

When aligning biological sequences, the choice of scoring scheme is critical. Even small changes in gap penalties, for example, can yield radically different alignments. A rigorous way to learn parameter values that are appropriate for biological sequences is through inverse parametric sequence alignment. Given a collection of examples of biologically correct reference alignments, this is the problem of finding parameter values that make the scores of the reference alignments be as close as possible to those of optimal alignments of their sequences. We extend prior work on inverse parametric alignment to partial examples, which contain regions where the reference alignment is not specified, and to an improved formulation based on minimizing the average error between the scores of the reference alignments and the scores of optimal alignments. Experiments on benchmark biological alignments show we can learn scoring schemes that generalize across protein families, and that boost the accuracy of multiple sequence alignment by as much as 25 percent.

MOTIVATION:Multiple sequence alignment is a fundamental task in bioinformatics. Current tools typically form an initial alignment by merging subalignments, and then polish this alignment by repeated splitting and merging of subalignments to obtain an improved final alignment. In general this form-and-polish strategy consists of several stages, and a profusion of methods have been tried at every stage. We carefully investigate: (1) how to utilize a new algorithm for aligning alignments that optimally solves the common subproblem of merging subalignments, and (2) what is the best choice of method for each stage to obtain the highest quality alignment.RESULTS:We study six stages in the form-and-polish strategy for multiple alignment: parameter choice, distance estimation, merge-tree construction, sequence-pair weighting, alignment merging, and polishing. For each stage, we consider novel approaches as well as standard ones. Interestingly, the greatest gains in alignment quality come from (i) estimating distances by a new approach using normalized alignment costs, and (ii) polishing by a new approach using 3-cuts. Experiments with a parameter-value oracle suggest large gains in quality may be possible through an input-dependent choice of alignment parameters, and we present a promising approach for building such an oracle. Combining the best approaches to each stage yields a new tool we call Opal that on benchmark alignments matches the quality of the top tools, without employing alignment consistency or hydrophobic gap penalties.AVAILABILITY:Opal, a multiple alignment tool that implements the best methods in our study, is freely available at http://opal.cs.arizona.edu.

For as long as biologists have been computing alignments of sequences, the question of what values to use for scoring substitutions and gaps has persisted. While some choices for substitution scores are now common, largely due to convention, there is no standard for choosing gap penalties. An objective way to resolve this question is to learn the appropriate values by solving the Inverse String Alignment Problem: given examples of correct alignments, find parameter values that make the examples be optimal-scoring alignments of their strings.We present a new polynomial-time algorithm for Inverse String Alignment that is simple to implement, fast in practice, and for the first time can learn hundreds of parameters simultaneously. The approach is also flexible: minor modifications allow us to solve inverse unique alignment (find parameter values that make the examples be the unique optimal alignments of their strings), and inverse near-optimal alignment (find parameter values that make the example alignments be as close to optimal as possible). Computational results with an implementation for global alignment show that, for the first time, we can find best-possible values for all 212 parameters of the standard protein-sequence scoring-model from hundreds of alignments in a few minutes of computation.

We consider the problem of aligning sequences related by a given evolutionary tree: given a fixed tree with its leaves labeled with sequences, find ancestral sequences to label the internal nodes so as to minimize the total cost of all the edges in the tree. The cost of an edge is the edit distance between the sequences labeling its endpoints. In this paper, we consider the case when the given tree is a regular d-ary tree for some fixed d and provide a d+1/d−1-approximation algorithm for this problem that runs in time O(d(2kn) d + n 2 k 2d ) where k is the number of leaves in the tree and n is the maximum length of any of the sequences labeling the leaves. We also consider a new bottleneck objective in labeling the internal nodes. In this version, we wish to find the labeling of the internal nodes that minimizes the maximum cost of any edge in the tree. For this problem we provide a simple 2δ+ 1-approximation algorithm where δ is the depth of the given undirected tree defined as the maximum over all internal nodes of the number of edges from the internal node to a closest leaf. For phylogenetic trees on n nodes that have no internal nodes of degree two, δ ≤ lg n.

We define a new problem in multiple sequence alignment, called maximum weight trace. The problem formalizes in a natural way the common practice of merging pairwise alignments to form multiple sequence alignments, and contains a version of the minimum sum of pairs alignment problem as a special case. Informally, the input is a set of pairs of matched characters from the sequences; each pair has an associated weight. The output is a subset of the pairs of maximum total weight that satisfies the following property: there is a multiple alignment that places each pair of characters selected by the subset together in the same column. A set of pairs with this property is called a trace. Intuitively a trace of maximum weight specifies a multiple alignment that agrees as much as possible with the character matches of the input. We develop a branch and bound algorithm for maximum weight trace. Though the problem is NP-complete, an implementation of the algorithm shows we can solve instances on as many as 6 sequences of length 250 in a few minutes. These are among the largest instances that have been solved to optimality to date for any formulation of multiple sequence alignment.