Wing-Kin Sung

Bio: Wing-Kin Sung is an academic researcher from National University of Singapore. The author has contributed to research in topics: Gene & Chromatin immunoprecipitation. The author has an hindex of 64, co-authored 327 publications receiving 26116 citations. Previous affiliations of Wing-Kin Sung include University of Hong Kong & Yale University.

Detection of generic spaced motifs using submotif pattern mining

Abstract: Motivation: Identification of motifs is one of the critical stages in studying the regulatory interactions of genes. Motifs can have complicated patterns. In particular, spaced motifs, an important class of motifs, consist of several short segments separated by spacers of different lengths. Locating spaced motifs is not trivial. Existing motif-finding algorithms are either designed for monad motifs (short contiguous patterns with some mismatches) or have assumptions on the spacer lengths or can only handle at most two segments. An effective motif finder for generic spaced motifs is highly desirable. Results: This article proposes a novel approach for identifying spaced motifs with any number of spacers of different lengths. We introduce the notion of submotifs to capture the segments in the spaced motif and formulate the motif-finding problem as a frequent submotif mining problem. We provide an algorithm called SPACE to solve the problem. Based on experiments on real biological datasets, synthetic datasets and the motif assessment benchmarks by Tompa et al., we show that our algorithm performs better than existing tools for spaced motifs with improvements in both sensitivity and specificity and for monads, SPACE performs as good as other tools. Availability: The source code is available upon request from the authors. Contact: ksung@comp.nus.edu.sg Supplementary information: Supplementary data are available at Bioinformatics online.

Indexing similar DNA sequences

Abstract: To study the genetic variations of a species, one basic operation is to search for occurrences of patterns in a large number of very similar genomic sequences. To build an indexing data structure on the concatenation of all sequences may require a lot of memory. In this paper, we propose a new scheme to index highly similar sequences by taking advantage of the similarity among the sequences. To store r sequences with k common segments, our index requires only O(n +N log N) bits of memory, where n is the total length of the common segments and N is the total length of the distinct regions in all texts. The total length of all sequences is rn + N, and any scheme to store these sequences requires Ω(n + N) bits. Searching for a pattern P of length m takes O(m + mlogN + mlog(rk)psc(P) + occlogn), where psc(P) is the number of prefixes of P that appear as a suffix of some common segments and occ is the number of occurrences of P in all sequences. In practice, rk ≤ N, and psc(P) is usually a small constant. We have implemented our solution and evaluated our solution using real DNA sequences. The experiments show that the memory requirement of our solution is much less than that required by BWT built on the concatenation of all sequences. When compared to the other existing solution (RLCSA), we use less memory with faster searching time.

Constructing compressed suffix arrays with large alphabets

Abstract: Recent research in compressing suffix arrays has resulted in two breakthrough indexing data structures, namely, compressed suffix arrays (CSA) [7] and FM-index [5]. Either of them makes it feasible to store a full-text index in the main memory even for a piece of text data with a few billion characters (such as human DNA). However, constructing such indexing data structures with limited working memory (i.e., without constructing suffix arrays) is not a trivial task. This paper addresses this problem. Currently, only CSA admits a space-efficient construction algorithm [15]. For a text T of length n over an alphabet Σ, this algorithm requires O(|Σ|nlogn) time and (2 H 0 + 1+e)n bits of working space, where H 0 is the 0-th order empirical entropy of T and e is any non-zero constant. This algorithm is good enough when the alphabet size | Σ| is small. It is not practical for text data containing protein, Chinese or Japanese, where the alphabet may include up to a few thousand characters.

Finding short right-hand-on-the-wall walks in graphs

Abstract: We consider the problem of perpetual traversal by a single agent in an anonymous undirected graph G. Our requirements are: (1) deterministic algorithm, (2) each node is visited within O(n) moves, (3) the agent uses no memory, it can use only the label of the link via which it arrived to the current node, (4) no marking of the underlying graph is allowed and (5) no additional information is stored in the graph (e.g. routing tables, spanning tree) except the ability to distinguish between the incident edges (called Local Orientation). This problem is unsolvable, as has been proven in [9,28] even for much less restrictive setting. Our approach is to somewhat relax the requirement (5). We fix the following traversal algorithm: “Start by taking the edge with the smallest labelx. Afterwards, whenever you come to a node, continue by taking the successor edge (in the local orientation) to the edge via which you arrived” and ask whether it is for every undirected graph possible to assign the local orientations in such a way that the resulting perpetual traversal visits every node in O(n) moves. We give a positive answer to this question, by showing how to construct such local orientations. This leads to an extremely simple, memoryless, yet efficient traversal algorithm.

Increasing confidence of protein-protein interactomes.

Abstract: High-throughput experimental methods, such as yeast-two-hybrid and phage display, have fairly high levels of false positives (and false negatives). Thus the list of protein-protein interactions detected by such experiments would need additional wet laboratory validation. It would be useful if the list could be prioritized in some way. Advances in computational techniques for assessing the reliability of protein-protein interactions detected by such high-throughput methods are reviewed in this paper, with a focus on techniques that rely only on topological information of the protein interaction network derived from such high-throughput experiments. In particular, we discuss indices that are abstract mathematical characterizations of networks of reliable protein-protein interactions--e.g., "interaction generality" (IG), "interaction reliability by alternative pathways" (IRAP), and "functional similarity weighting" (FSWeight). We also present indices that are based on explicit motifs associated with true-positive protein interactions--e.g., "new interaction generality" (IG2) and "meso-scale motifs" (NeMoFinder).

Fast and accurate short read alignment with Burrows–Wheeler transform

Abstract: Motivation: The enormous amount of short reads generated by the new DNA sequencing technologies call for the development of fast and accurate read alignment programs. A first generation of hash table-based methods has been developed, including MAQ, which is accurate, feature rich and fast enough to align short reads from a single individual. However, MAQ does not support gapped alignment for single-end reads, which makes it unsuitable for alignment of longer reads where indels may occur frequently. The speed of MAQ is also a concern when the alignment is scaled up to the resequencing of hundreds of individuals. Results: We implemented Burrows-Wheeler Alignment tool (BWA), a new read alignment package that is based on backward search with Burrows–Wheeler Transform (BWT), to efficiently align short sequencing reads against a large reference sequence such as the human genome, allowing mismatches and gaps. BWA supports both base space reads, e.g. from Illumina sequencing machines, and color space reads from AB SOLiD machines. Evaluations on both simulated and real data suggest that BWA is ~10–20× faster than MAQ, while achieving similar accuracy. In addition, BWA outputs alignment in the new standard SAM (Sequence Alignment/Map) format. Variant calling and other downstream analyses after the alignment can be achieved with the open source SAMtools software package. Availability: http://maq.sourceforge.net Contact: [email protected]

Ultrafast and memory-efficient alignment of short DNA sequences to the human genome

Abstract: Bowtie is an ultrafast, memory-efficient alignment program for aligning short DNA sequence reads to large genomes. For the human genome, Burrows-Wheeler indexing allows Bowtie to align more than 25 million reads per CPU hour with a memory footprint of approximately 1.3 gigabytes. Bowtie extends previous Burrows-Wheeler techniques with a novel quality-aware backtracking algorithm that permits mismatches. Multiple processor cores can be used simultaneously to achieve even greater alignment speeds. Bowtie is open source http://bowtie.cbcb.umd.edu.

An integrated encyclopedia of DNA elements in the human genome

Abstract: The human genome encodes the blueprint of life, but the function of the vast majority of its nearly three billion bases is unknown. The Encyclopedia of DNA Elements (ENCODE) project has systematically mapped regions of transcription, transcription factor association, chromatin structure and histone modification. These data enabled us to assign biochemical functions for 80% of the genome, in particular outside of the well-studied protein-coding regions. Many discovered candidate regulatory elements are physically associated with one another and with expressed genes, providing new insights into the mechanisms of gene regulation. The newly identified elements also show a statistical correspondence to sequence variants linked to human disease, and can thereby guide interpretation of this variation. Overall, the project provides new insights into the organization and regulation of our genes and genome, and is an expansive resource of functional annotations for biomedical research.

Machine learning

Abstract: Machine Learning is the study of methods for programming computers to learn. Computers are applied to a wide range of tasks, and for most of these it is relatively easy for programmers to design and implement the necessary software. However, there are many tasks for which this is difficult or impossible. These can be divided into four general categories. First, there are problems for which there exist no human experts. For example, in modern automated manufacturing facilities, there is a need to predict machine failures before they occur by analyzing sensor readings. Because the machines are new, there are no human experts who can be interviewed by a programmer to provide the knowledge necessary to build a computer system. A machine learning system can study recorded data and subsequent machine failures and learn prediction rules. Second, there are problems where human experts exist, but where they are unable to explain their expertise. This is the case in many perceptual tasks, such as speech recognition, hand-writing recognition, and natural language understanding. Virtually all humans exhibit expert-level abilities on these tasks, but none of them can describe the detailed steps that they follow as they perform them. Fortunately, humans can provide machines with examples of the inputs and correct outputs for these tasks, so machine learning algorithms can learn to map the inputs to the outputs. Third, there are problems where phenomena are changing rapidly. In finance, for example, people would like to predict the future behavior of the stock market, of consumer purchases, or of exchange rates. These behaviors change frequently, so that even if a programmer could construct a good predictive computer program, it would need to be rewritten frequently. A learning program can relieve the programmer of this burden by constantly modifying and tuning a set of learned prediction rules. Fourth, there are applications that need to be customized for each computer user separately. Consider, for example, a program to filter unwanted electronic mail messages. Different users will need different filters. It is unreasonable to expect each user to program his or her own rules, and it is infeasible to provide every user with a software engineer to keep the rules up-to-date. A machine learning system can learn which mail messages the user rejects and maintain the filtering rules automatically. Machine learning addresses many of the same research questions as the fields of statistics, data mining, and psychology, but with differences of emphasis. Statistics focuses on understanding the phenomena that have generated the data, often with the goal of testing different hypotheses about those phenomena. Data mining seeks to find patterns in the data that are understandable by people. Psychological studies of human learning aspire to understand the mechanisms underlying the various learning behaviors exhibited by people (concept learning, skill acquisition, strategy change, etc.).

Model-based Analysis of ChIP-Seq (MACS)

Abstract: We present Model-based Analysis of ChIP-Seq data, MACS, which analyzes data generated by short read sequencers such as Solexa's Genome Analyzer. MACS empirically models the shift size of ChIP-Seq tags, and uses it to improve the spatial resolution of predicted binding sites. MACS also uses a dynamic Poisson distribution to effectively capture local biases in the genome, allowing for more robust predictions. MACS compares favorably to existing ChIP-Seq peak-finding algorithms, and is freely available.