Ket Fah Chong

Computing bounding volume hierarchies using model simplification

This paper presents a framework that uses the outputs of model simplification to guide the construction of bounding volume hierarchies for use in, for example, collision detection. Simplified models, besides their application to multiresolution rendering, can provide clues to the object’s shape. These clues help in the partitioning of the object’s model into components that may be more tightly bounded by simple bounding volumes. The framework naturally employs both the bottom-up and the topdown approaches of hierarchy building, and thus can have the advantages of both approaches. Experimental results show that our method built on top of the framework can indeed improve the bounding volume hierarchy, and as a result, significantly speedup the collision detection.

TUTORIAL ON DE NOVO PEPTIDE SEQUENCING USING MS/MS MASS SPECTROMETRY

This paper is a self-contained introductory tutorial on the problem in proteomics known as peptide sequencing using tandem mass spectrometry. This tutorial deals specifically with de novo sequencing methods (as opposed to database search methods). We first give an introduction to peptide sequencing, its importance and history and some background on proteins. Next we show the relationship between a peptide and the final spectrum produced from a tandem mass spectrometer, together with a description of the various sources of complications that arise during the process of generating the mass spectrum. From there we model the computational problem of de novo peptide sequencing, which is basically the reverse problem of identifying the peptide which produced the spectrum. We then present several major approaches to solve it (including reviewing some of the current algorithms in each approach), and also discuss related problems and post-processing approaches.

Modeling and Characterization of Multi-Charge Mass Spectra for Peptide Sequencing

Peptide sequencing using tandem mass spectrometry data is an important and challenging problem in proteomics. We address the problem of peptide sequencing for multi-charge spectra. Most peptide sequencing algorithms currently consider only charge one or two ions even for higher-charge spectra. We give a characterization of multi-charge spectra by generalizing existing models. Using our models, we analyzed spectra from Global Proteome Machine (GPM) [Craig R, Cortens JP, Beavis RC, J Proteome Res 3:1234-1242, 2004.] (with charges 1-5), Institute for Systems Biology (ISB) [Keller A, Purvine S, Nesvizhskii AI, Stolyar S, Goodlett DR, Kolker E, OMICS 6:207-212, 2002.] and Orbitrap (both with charges 1-3). Our analysis for the GPM dataset shows that higher charge peaks contribute significantly to prediction of the complete peptide. They also help to explain why existing algorithms do not perform well on multi-charge spectra. Based on these analyses, we claim that peptide sequencing algorithms can achieve higher sensitivity results if they also consider higher charge ions. We verify this claim by proposing a de novo sequencing algorithm called the greedy best strong tag (GBST) algorithm that is simple but considers higher charge ions based on our new model. Evaluation on multi-charge spectra shows that our simple GBST algorithm outperforms Lutefisk and PepNovo, especially for the GPM spectra of charge three or more.

EnzDP: Improved enzyme annotation for metabolic network reconstruction based on domain composition profiles

Determining the entire complement of enzymes and their enzymatic functions is a fundamental step for reconstructing the metabolic network of cells. High quality enzyme annotation helps in enhancing metabolic networks reconstructed from the genome, especially by reducing gaps and increasing the enzyme coverage. Currently, structure-based and network-based approaches can only cover a limited number of enzyme families, and the accuracy of homology-based approaches can be further improved. Bottom-up homology-based approach improves the coverage by rebuilding Hidden Markov Model (HMM) profiles for all known enzymes. However, its clustering procedure relies firmly on BLAST similarity score, ignoring protein domains/patterns, and is sensitive to changes in cut-off thresholds. Here, we use functional domain architecture to score the association between domain families and enzyme families (Domain-Enzyme Association Scoring, DEAS). The DEAS score is used to calculate the similarity between proteins, which is then used in clustering procedure, instead of using sequence similarity score. We improve the enzyme annotation protocol using a stringent classification procedure, and by choosing optimal threshold settings and checking for active sites. Our analysis shows that our stringent protocol EnzDP can cover up to 90% of enzyme families available in Swiss-Prot. It achieves a high accuracy of 94.5% based on five-fold cross-validation. EnzDP outperforms existing methods across several testing scenarios. Thus, EnzDP serves as a reliable automated tool for enzyme annotation and metabolic network reconstruction. Available at: www.comp.nus.edu.sg/~nguyennn/EnzDP .

A database search algorithm for identification of peptides with multiple charges using tandem mass spectrometry

Peptide sequencing using tandem mass spectrometry is the process of interpreting the peptide sequence from a given mass spectrum. Peptide sequencing is an important but challenging problem in bioinformatics. The advancement in mass spectrometry machines has yielded great amount of high quality spectra data, but the methods to analyze these spectra to get peptide sequences are still accurate. There are two types of peptide sequencing methods –database search methods and the de novo methods. Much progress has been made, but the accuracy and efficiency of these methods are not satisfactory and improvements are urgently needed. In this paper, we will introduce a database search algorithm for sequencing of peptides using tandem mass spectrometry. This Peptide Sequence Pattern (PSP) algorithm first generates the peptide sequence patterns (PSPs) by connecting the strong tags with mass differences. Then a linear time database search process is used to search for candidate peptide sequences by PSPs, and the candidate peptide sequences are then scored by share peaks count. The PSP algorithm is designed for peptide sequencing from spectra with multiple charges, but it is also applicable for singly charged spectra. Experiments have shown that our algorithm can obtain better sequencing results than current database search algorithms for many multiply charged spectra, and comparative results for singly charged spectra against other algorithms.

HOMOLOGOUS SYNTENY BLOCK DETECTION BASED ON SUFFIX TREE ALGORITHMS

A synteny block represents a set of contiguous genes located within the same chromosome and well conserved among various species. Through long evolutionary processes and genome rearrangement events, large numbers of synteny blocks remain highly conserved across multiple species. Understanding distribution of conserved gene blocks facilitates evolutionary biologists to trace the diversity of life, and it also plays an important role for orthologous gene detection and gene annotation in the genomic era. In this work, we focus on collinear synteny detection in which the order of genes is required and well conserved among multiple species. To achieve this goal, the suffix tree based algorithms for efficiently identifying homologous synteny blocks was proposed. The traditional suffix tree algorithm was modified by considering a chromosome as a string and each gene in a chromosome is encoded as a symbol character. Hence, a suffix tree can be built for different query chromosomes from various species. We can then efficiently search for conserved synteny blocks that are modeled as overlapped contiguous edges in our suffix tree. In addition, we defined a novel Synteny Block Conserved Index (SBCI) to evaluate the relationship of synteny block distribution between two species, and which could be applied as an evolutionary indicator for constructing a phylogenetic tree from multiple species instead of performing large computational requirements through whole genome sequence alignment.

A Merge-Decoupling Dead End Elimination algorithm for protein side-chain conformation

Dead End Elimination (DEE) is a technique for eliminating rotamers that can not exist in any global minimum energy configuration for the protein side chain conformation problem. A popular method is Simple Goldstein DEE (SG-DEE) which is fast and eliminates rotamers by considering single residues for possible elimination. We present a Merge-Decoupling DEE (MD-DEE) that further reduces the number of rotamers after SG-DEE. MD-DEE works by forming residue-pairs but is fast and, like SG-DEE, is practical even for large proteins. Our experiments show that MD-DEE achieves further reduction in residue elimination (up to 25%) after SG-DEE.

De Novo Peptide Sequencing for Mass Spectra Based on Multi-Charge Strong Tags.

This paper presents an improved algorithm for de novo sequencing of multi-charge mass spectra. Recent work based on the analysis of multi-charge mass spectra showed that taking advantage of multi-charge information can lead to higher accuracy (sensitivity and specificity) in peptide sequencing. A simple de novo algorithm, called GBST (Greedy algorithm with Best Strong Tag) was proposed and was shown to produce good results for spectra with charge > 2. In this paper, we analyze some of the shortcomings of GBST. We then present a new algorithm GST-SPC, by extending the GBST algorithm in two directions. First, we use a larger set of multi-charge strong tags and show that this improves the theoretical upper bound on performance. Second, we give an algorithm that computes a peptide sequence that is optimal with respect to shared peaks count from among all sequences that are derived from multi-charge strong tags. Experimental results demonstrate the improvement of GST-SPC over GBST.

Characterization of Multi-Charge Mass Spectra for Peptide Sequencing.

Sequencing of peptide sequences using tandem mass spectrometry data is an important and challenging problem in proteomics. In this paper, we address the problem of peptide sequencing for multi-charge spectra. Most peptide sequencing algorithms currently handle spectra of charge 1 or 2 and have not been designed to handle higher-charge spectra. We give a characterization of multicharge spectra by generalizing existing models. Using these new models, we have analyzed spectra with charges 1-5 from the GPM [8] datasets. Our analysis shows that higher charge peaks are present and they contribute significantly to prediction of the complete peptide. They also help to explain why existing algorithms do not perform well on multi-charge spectra. We also propose a new de novo algorithm for dealing with multi-charge spectra based on the new models. Experimental results show that it performs well on all spectra, especially so for multi-charge spectra.

An extension of dead end elimination for protein side-chain conformation using merge-decoupling

A two-phase strategy is widely adopted to solve the side-chain conformation prediction (SCCP) problem. Phase one is a fast reduction phase removing large numbers of rotamers not existing in the GMEC. Phase two (optimization phase) uses heuristics or exhaustive search to find a good/optimal solution. Presently, DEE (Dead End Elimination) is the only deterministic reduction method for phase one. However, to achieve convergence in phase two using DEE, the strategy of forming super-residues is used. This quickly leads to a combinatorial explosion, and becomes inefficient In this paper, an improvement of the DEE process by forming super-residues efficiently is proposed for phase one. The method basically merges residues into pairs based on some merging criteria. Simple Goldstein is then applied until no more elimination is possible. A decoupling process then reforms the original residues sans removed rotamers and rotamer pairs. The process of merging and elimination is repeated until no more elimination is possible. Initial experiments have shown the method, called Merge-Decoupling DEE, can fix up to 25% of the unfixed residues coming out of Simple Goldstein DEE.