Current Events

Second Year Bioinformatics Seminars

Leonard Apeltsin
Thursday, September 27th, 2007, 4:00 PM - Genentech Hall Room S201
Categorizing Functionally Significant Residues in Protein Superfamily Sequence Datasets Using Network Based Algorithms

Given a few thousand sequences belonging to a completely uncharacterized protein superfamily, is it possible estimate the identity of the active site residues along all the sequences? My research aims at answer this question. I treat an input dataset as a protein similarity network. By filtering the network, I elucidate out the functionally identical families, as well as those linker sequences that serve as a functional boundary between families in sequence space. Next, I compute an alignment trace graph for pairs of neighboring families. Clustering the alignment data allows me to estimate the conservation of residues across sequence space. By measuring areas of densely packed conserved and non-conserved residues in the linker sequences, I obtain a set of functionally significant motifs. I then propagate these statistics down to all the sequences in the dataset in order to categorize all residues. Preliminary results for my algorithm look promising.

David Barkan
Monday, September 10, 2007, 10:00 AM - QB3 Room 212
Predicting protein-protein interaction specificity in the bacterial type three secretion system

A number of pathogenic bacteria, among them E. coli, S. typhimurium, and Y. pestis, deliver virulence proteins into the host cell through the Type Three Secretion System (TTSS). These effector proteins are injected through the host cell membrane by the TTSS 'nano-syringe' assembly, where they interfere with host signaling systems for the advantage of the pathogen. Within the bacteria, virulence proteins must bind to small, globular proteins known as chaperones that direct them to the appropriate secretion apparatus. Each virulence protein binds specifically to one chaperone, and solved chaperone-virulence protein complexes have shown a similar binding mode for all pairs. My research seeks to predict novel binding chaperone-virulence protein binding partners using comparative protein structure modelling and protein-protein interface assessment by a residue-based statistical potential. I will discuss progress towards recapitulating binding specificity in complexes for which the members are known and ways that the methods used are adjusted to take advantage of the knowledge of the biology of the system.

Michelle Dimon
Tuesday, August 14, 2007, 3:00 PM - QB3 Room 212
Evading the host: Regulation of antigenic variation in plasmodium falciparum

Worldwide there are approximately half a billion cases of malaria each year, resulting in one to three million deaths. The parasite Plasmodium falciparum causes the most deadly form of malaria and is responsible for almost all malarial deaths. PfEMP1, a protein exported to the surface of infected red blood cells, provides a key to P. falciparum's ability to establish long-term infections. The var gene family, which encodes PfEMP1, is regulated by a system of allelic exclusion ensuring that only a single var gene, out of repertoire of about sixty, is expressed at a time by a parasite. Over the course of an infection, the active var gene switches to continually evade the host immune system. This research uses Chromosome Conformation Capture (3C) to examine regulation in the var gene family.

Colin Smith
Tuesday, April 24th, 2007, 2:00 PM - QB3 Room 212
A new masseuse on the block: Using a backrub to simulate localized protein motion

Explicit side-chain rotamer sampling has proven to be an important element of protein design, rescoring docking results, loop prediction, homology modeling, and pKa prediction. However, treating backbones as rigid is clearly an artificial imposition that can significantly bias computational predictions. The focus of my work to date has been on implementing and applying an efficient full-atom Monte Carlo move, called Backrub, that represents one off the simplest possible motional models extending beyond side-chains. This particular combination of simplicity and efficiency distinguishes it from other backbone sampling methods, including molecular dynamics, gradient-based minimization, fragment insertion, and other more complex geometric algorithms.

In prior work, the Richardson group has qualitatively shown the Backrub motion to be a good model for transitions between alternate backbone conformations observed in high-resolution (<1Å) crystal structures. I have used simulations combining the Backrub move and side-chain rotamer sampling to partially recapitulate those correlated backbone/side-chain conformations. A generalization of this model of local backbone movement has proven to be useful in a number of other applications. I show preliminary data indicating that Backrub sampling at room temperature can capture conformational fluctuations (up to 7 Å) in active site loop of triosephosphate isomerase that are qualitatively similar to those observed in nature. Work by others in the lab have shown the Backrub move to generate structural ensembles that reproduce side-chain and backbone order parameters better than either static crystal structures or deposited NMR ensembles. In addition, sequence design on similar ensembles has shown better agreement with the sequence entropy observed in protein families.

Michael Keiser
Friday, October 20, 2006, 2:00 PM - QB3 Room 212
Relating proteins by ligand similarity

The identification of protein function based on biological information is an area of intense research. Here I consider a complementary technique that quantitatively groups and relates proteins based on the chemical similarity of their ligands. For robustness, a statistical model of random similarity was developed to rank the significance of these relationships. These significance scores are then expressed graphically, as networks and heat maps. Although no biological information is used in calculating these maps, biologically sensible clusters appear as an emergent property. Links among unexpected targets also emerged, including predictions of novel compound selectivity that were subsequently confirmed by biochemical and cell-based assays. Relating receptors by ligand chemistry organizes biology in new ways to reveal unexpected relationships that may be tested directly by the ligands themselves.

Kai-Yeung Lau
Monday, September 11, 2006, 3:00 PM - QB3 Room 212
Function poses constraints to networks architecture and dynamics: A case study for yeast cell cycle network

Systems biology studies biological networks composed of large numbers of interacting molecules. A central problem is to understand the relationship between network structure and its function. Previous studies have focused on how the structure of a biological network constrains its types of dynamical behaviors, such as its functions, and robustness against perturbations. A principal finding was a phase transition from ordered to chaotic dynamical behavior as connectivity increases. More recent studies found that the scale-free architecture, which is observed in many biological networks, are robust to random failures. However, the question on how functional requirement constrains the structural and more general dynamical properties of a network remains relatively unexplored. In this talk, I present techniques to address this question using a case study for the yeast cell cycle network.

Dan Mandell
Friday, September 8, 2006, 3:00 PM - Genentech Hall Room 201
Flexible Backbone Protein Design

Protein design promises to aid the development of proteins as highly specific therapeutics, agents for bioremediation, and systems-level regulators of cellular activity. Currently, most protein design methods sample amino acid side chains on a fixed backbone. Despite some seminal successes, fixed backbone methods insufficiently reproduce the sequence diversity available to a given protein fold, leaving a potentially useful portion of sequence space inaccessible. In this talk, I explore current methods for modeling backbone flexibility during design, and describe a new technique for sampling local backbone fluctuations. Prior work demonstrating improvements in conformational sampling is discussed, and a larger-scale study validating the method is proposed. Finally, an application of the new backbone flexibility methodology to a biophysically motivated engineering task is described.