Protein complexes are fundamental for understanding principles of cellular organizations. However, most protein interactome maps are still essentially an in vitro scaffold. Furthermore, these protein interactome maps contain a significant amount of noise interactions, as well as missing many real interactions. It is thus an important challenge to reliably deduce in vivo protein interactions and to identify membership in the same protein complexes. In this talk, we describe recent progress in computational techniques for protein complex prediction from noisy protein interaction network data.

In the recent years there has been a growing interest in inferring the total order of genes or markers on a chromosome, since current genetic mapping efforts might only suffice to produce a partial order. Many interesting optimization problems were thus formulated in the framework of genome rearrangement. As an important one among them, the minimum breakpoint linearization (MBL) problem is to find the total order of a partially-ordered genome that minimizes its breakpoint distance to a reference genome whose genes are already totally ordered. It was previously shown to be NP-hard, and the algorithms proposed so far are all heuristic. In this talk, I will present an approximation algorithm for the minimum breakpoint linearization problem. This is the joint work with Dr. Cui Yun from NTU.

An area of much recent interest in Bioinformatics is developing strategies for combining information across different biological experiments in order to answer broad questions or create new and innovative hypotheses for further investigation. Microarray data is a major focus of these efforts, as it represents genetic information derived under many different conditions and tissue types, and it is readily accessible from a number of publicly accessible repositories. Combining data from diverse sources allows scientists to perform global studies such as identifying genes that are involved in different types of cancers, tissues or different stages of cancer progression. Such global studies are not generally possible from the standard results provided by individual microarray studies, in which lists of genes that are differentially expressed are provided for the specific experimental conditions, but no extensions are generally available to other relevant experiments. In order to address this problem, we propose a new technique that exploits relative information such as inter-gene dependencies. The proposed technique projects gene expression data onto a new information space where genes are represented by their relations to other landmark genes. In this talk, we also introduce a microarray integration framework and a new clustering technique based on this approach.

Proteins are always at the center of any biological systems. As the number of protein sequences with unknown function increases, it has become an important issue to assign accurate functions to those proteins. To understand their functions in a more complete sense, it is important to understand how they look, what they do, what they interact with, how they do all those things. It is also important to understand what the consequences of specific structures, functions and interactions of proteins are. In this presentation, we will discuss recent research developments on protein's structures, functions, and interactions. We will also discuss its implications on drug discovery. Primary tools for these studies are based on various types of machine learning approaches such as support vector machine and matrix factorization. We will discuss how these machine learning methods can be effectively utilized in bioinformatics research.

With availability of large scale biological data sets, it became important research topics to develop new computational methods for integrating biological information. In this talk, I will present our recent two studies about integrative topics: protein function predictions and disease gene predictions. First, decades of research in computational function prediction in model organisms such as yeast and worm became mature enough to apply them into mammals. To show the performance of these methods in mammals, mouse function prediction project has been organized and showed results with both breadth and high accuracy in predicting mouse functions. Second, DNA copy number aberrations (CNAs) and gene expression (GE) changes provide valuable information for studying chromosomal instability and its consequences in cancer. Our study suggest that CNAs are both directly and indirectly correlated with changes in expression and that it is beneficial to examine cancer related genes and biological pathways.

Determining isotopic clusters and their monoisotopic masses is a first step in interpreting complex mass spectra generated by high resolution mass spectrometers. We propose a mathematical model for isotopic distributions of polypeptides and an effective interpretation algorithm. Our method was applied to high resolution mass spectrometric data obtained from a Fourier transform ion cyclotron resolution (FT-ICR) mass spectrometer that is coupled to a reverse-phase liquid chromatography (RPLC). For polypeptides whose amino acid sequences were identified by MS/MS, we applied both THRASH-based software implementations and our method. Our method was observed to find more masses of known peptides when the numbers of the total clusters identified by both methods were fixed. Another advantage of our method is that it is very fast, since our method can calculate the scores of isotopic clusters much faster than THRASH that calculates the least-squares fit.