Hoong Kee Ng

Detecting hubs and quasi cliques in scale-free networks

Scale-free networks are believed to closely model most real-world networks. An interesting property of such networks is the existence of so-called hub and community structures. In this paper, we model hubs as high-degree nodes and communities as quasi cliques. We propose a new problem formulation called the ¿-list dominating set and show how this single problem is suited to model both the structures in real-world networks better than traditional problems like vertex cover and clique. Additionally, we provide a fixed-parameter tractable algorithm to this detect these structures and show experimental results on protein-protein interaction networks.

Finding Patterns in Biological Sequences by Longest Common Subsequencesand Shortest Common Supersequences

Patterns in biological sequences are important for revealing the relationship among biological sequences. Much research has been done on this problem, and the sensitivity and specificity of current algorithms are already quite satisfactory. However, in general, for problems on a set of sequences, the relationship among their patterns, their longest common subsequences (LCS) and their shortest common supersequences (SCS) are not examined carefully. Therefore, revealing the relationship between the patterns and LCS/SCS might provide us with a deeper view of the patterns of biological sequences, in turn leading to a better understanding of them. In this paper, we propose the PALS (PAtterns by Lcs and Scs) algorithms to discover patterns in a set of biological sequences by first generating the results for LCS and SCS of sequences by heuristic, and consequently derive the patterns from these results. Experiments show that the PALS algorithms perform well (both in efficiencies and in accuracies) on a variety of sequences

Efficient algorithm for path-based range query in spatial databases

We study the path-based range query (PRQ) for 2-dimensional spatial database defined as follows: given a sequence of query points, P = {p/sub 1/, p/sub 2/,..., p/sub n/}, and a search distance d, we want to report all points in the spatial database that are within a distance d of some point p/sub i/ in P. This query arises from traveler information systems where it is often a feature to report events that lie nearby a planned route. The simple method of performing repeated range query (RRQ), i.e. the standard range query for each query point p/sub i/ (1 /spl les/ i /spl les/ n) and combining the results is inefficient as it involves multiple searches on the database. We present an algorithm for the PRQ that uses only one pass of the R-tree while simultaneously process all the points in the query path P. We generalize pruning rules for the standard range query and also present new ones for efficient processing of PRQ. Extensive experiments on the PRQ and RRQ using various different datasets (real and randomly-generated), different R-tree variants (including bulk-loaded ones), over different query paths P, and search distances d. Experiments show that the algorithm for PRQ outperforms RRQ significantly and that this is consistent across the various problem parameters studied. Recently, we have also compared this basic PRQ algorithm with more refined techniques for the PRQ such as the sorted-path algorithm and the rectangle intersection method. Our simple algorithm performs well in comparison to these more sophisticated methods.

Path-Based Range Query Processing Using Sorted Path and Rectangle Intersection Approach

Path-based range query (PRQ) is a class of range query in which given a set of points P in two-dimensional plane that define a path, find the union of all points within a distance d from the points in P. The simple method of performing repeated range query (RRQ), i.e. the standard range query for each query point and combining the results is inefficient as it searches the spatial index multiple times. Current method, using two pruning rules PointOut and NodeIn, of solving PRQ in one pass while simultaneously processing all the points in P involves using minimum and maximum geometric distances of the query points in P was revisited. We further present two techniques for PRQ that improves the running time of the two pruning rules, called sorted path and rectangle intersection. Empirical results from real life datasets supported our theoretic results that PRQ using sorted path is better than the other approaches.

A new approach for similarity queries of biological sequences in databases

As biological databases grow larger, effective query of the biological sequences in these databases has become an increasingly important issue for researchers. There are currently not many systems for fast access of very large biological sequences. In this paper, we propose a new approach for biological sequences similarity querying in databases. The general idea is to first transform the biological sequences into vectors and then onto 2-d points in planes; then use a spatial index to index these points with self-organizing maps (SOM), and perform a single efficient similarity query (with multiple simultaneous input sequences) using a fast algorithm, the multi-point range query (MPRQ) algorithm. This approach works well because we could perform multiple sequences similarity queries and return the results with just one MPRQ query, with tremendous savings in query time. We applied our method onto DNA and protein sequences in database, and results show that our algorithm is efficient in time, and the accuracies are satisfactory.