Science at the Interface

This series of lectures focuses on frontiers of research where the biological and mathematical sciences converge. Seminars are held in the Conference Room of the Mellon Institute at 12:30 on Wednesdays unless otherwise indicated.

October 31, 2007
Michael Domach, Ph.D.
Department of Chemical Engineering, Carnegie Mellon University
Host: Dr. Russell Schwartz

"Metabolic engineering of E. coli and B. subtilis to increase synthetic efficiency"
The aim of the work to be presented is to redirect the metabolism of E. coli and B. subtilis more fully into the synthesis of economically interesting products such a folic acid and recombinant proteins. Currently, these commonly used production platforms “waste” more than 40% of input carbon as by-products under aerobic conditions. From a more philosophical standpoint, showing that cells can be re-engineered towards achieving the technologist’s view of efficiency supports the contention of game theorists and others that microbes (and cancer cells) are “wired” to be inefficient. Such inefficiency is thought by some to be a vital component of a successful competition strategy. This seminar will provide an overview of the computational methodologies we have developed to enable metabolic engineering, but more emphasis will be placed on the experimental verification of predictions.

April 4, 2007
Jason Swedlow, Ph.D.
University of Dundee
Host: Dr. Robert Murphy

"Functional analysis of the mitotic centromere"
The assembly of mitotic chromosomes and kinetochores and the orientation and alignment of chromosomes on the mitotic spindle are critical for proper mitotic chromosome segregation. We recently discovered that a protein kinase critical for correct chromosome biorientation, Aurora B, phosphorylated an important microtubule depolymerase, MCAK. Our evidence suggests that this phosphorylation constitutes a functional switch that determines how MCAK functions in chromosomes. More recently, we have developed tools for a proteomic analysis of mitotic chromosomes. One of the novel proteins we identified, Bod1, is also required for proper chromosome biorientation and for the proper phosphorylation of MCAK by Aurora B. We are currently characterising the interaction between Aurora B, Bod1, and MCAK.

March 21, 2007
Newell Washburn, Ph.D.
Chemistry Department, Carnegie Mellon University
Host: Dr. Dannie Durand

"Adventures in biomaterials"
In this seminar, I will provide an overview of our research at the intersection of polymer science and biotechnology. Our primary focus has been on developing polymeric materials capable of interacting with native repair processes to promote wound healing and tissue regeneration. Toward this goal, we have been investigating the interactions of soluble signaling proteins, such as growth factors and cytokines, with polymeric and biological matrices and developing strategies for modulating these interactions. I will discuss our work on developing matrices based on hyaluronic acid and work in measuring and controlling the dynamics of the pro-inflammatory cytokine interleukin-1beta. In a related project, we have been working toward developing optical biosensors capable of detecting sub-nanomolar concentrations of cytokines and other markers of tumor growth. This biosensor uses a range of proteins and other biomolecules to capture DNA, pro-inflammatory cytokines, and other damage-associated molecular patterns known to correlate with cancer progression. Finally, we have begun preliminary work in engineering yeast to synthesize styrene, a commodity chemical that is usually derived from petroleum. For styrene biosynthesis, two genes are incorporated to convert L-phenylalanine enzymatically to styrene. Strategies for optimizing styrene yield will be discussed as well as the potential for making other polymerizable monomers.

October 18, 2006
Federica Brandizzi, Ph.D.
Michigan State University
Host: Dr. Robert Murphy

"Towards a dynamic analysis of protein trafficking in the plant early secretory pathway"
Secretory materials are synthesized on the surface of the endoplasmic reticulum (ER). They are then shipped from the ER to the Golgi apparatus to be sorted either back to the ER or to distal secretory compartments such as vacuoles and plasma membrane. The ER and Golgi are closely associated in plant cells. How these two organelles communicate with each other is an important question that remains largely unanswered. To provide understanding of the mechanisms of cargo export from the plant ER, we have explored the dynamics of protein transport between the ER and the Golgi apparatus using live cell imaging techniques. With this approach we found that the domains of the ER dedicated to the export of proteins, the ER export sites (ERESs), form secretory units that move along the surface of the ER together with the Golgi stacks. We also determined that the integrity of protein export from the ER as well as that of Golgi and ERESs is regulated by the activity of specific GTPases, such as Sar1 and Arf1, as well as by specific signals on cargo molecules. Our results indicate that in plant cells the ER and Golgi form a dynamic membrane system whose components continuously cycle through the ER via controlled membrane trafficking pathways.

February 22, 2006
Christopher Langmead, Ph.D.
School of Computer Science, Carnegie Mellon University

Mathematical models of biological processes are essential to simulation studies. There is, however, a growing body of literature devoted to the formal analysis of the models themselves. These analyses seek to prove properties about the model and thus identify shortcomings or suggest new experiments. In this talk, I will survey some recent advances in the area of biological model checking and present some preliminary results from an experiment validating an instance of the 2D HP model of protein folding.

February 15, 2006
Fernando de La Torre, Ph.D.
The Robotics Institute, Carnegie Mellon University
"Component analysis for classification, clustering and modeling high dimensional data"
Host: Dr. Javier López

Component Analysis (Principal Component Analysis, Linear Discriminant Analysis, Tensor Factorization, ...) have been successfully applied to numerous bioinformatics, visual and signal processing tasks over the last two decades. In this talk I will provide an overview of traditional component analysis methods and recent extensions useful for dimensionality reduction, modeling, classifying and clustering high dimensional data (e.g. images). In particular, I will describe in a unified framework four novel component analysis techniques:

1) Robust parameterized component Analysis (RPCA): Extension of principal component analysis (PCA) to build a linear model robustly to outliers and invariantly to geometric transformations.

2) Multimodal Oriented Component Analysis (MODA): Generalization of linear discriminant analysis (LDA) optimal for Gaussian multimodal classes with different covariances.

3) Representational Oriented Component Analysis (ROCA): Extension of OCA to improve classification accuracy when few training samples are available. (e.g. just 1 training sample).

4) Discriminative Cluster Analysis (DCA): Unsupervised low dimensional reduction method that finds a subspace better suited for k-means clustering.

Applications of these techniques to visual tracking, learning and recognition, and temporal segmentation of activities from multimodal data (audio, video, body sensors) will be discussed.

December 7, 2005
Chris Burge, Ph.D.
Department of Biology, Massachusettes Institute of Technology
"Towards an RNA splicing code"
Host: Christine Wang

I will describe my lab's progress toward understanding the rules for exon recognition by the RNA splicing machinery in mammals. Current efforts are focused on systematic identification and characterization of sequences that function as exonic and intronic splicing silencers (ESS, ISS) and enhancers (ESE, ISE), using a combination of cell-based and computational screens. The identified splicing regulatory elements are being integrated with statistical models of the core splice site motifs into computer algorithms that simulate RNA splicing specificity. Recently, we have shown that ESS sequences play general roles in splice site definition at both the 5' and 3' splice sites, and we are investigating the mechanisms of this activity. We have also obtained evidence that ESS sequences are likely to control alternative 5' and 3' splice site usage in many exons, a common type of alternative splicing in mammals.

February 23, 2005
Eric Xing, Ph.D.
Department of Computer Science, Carnegie Mellon University
"In silico motif detection under complex genomic and evolutionary context - new Bayesian models motivated from biological principles"
Host: Dr. Elizabeth Jones

A hallmark of the transcriptional regulatory sequences of higher eukaryotic genome is the presence of highly sophisticated deterministic and stochastic constraints on motif deployment and diverse categorization of motif structures, and the enormous size of the regulatory sequences in which motifs must be found. Most contemporary motif detection algorithms adopt simple assumptions on motif structure and organization, and are therefore incapable of identifying non-trivial regulatory structures such as enhancers out of a complex background from higher eukaryotic genome. In this talk, I discuss an expressive probabilistic framework for modeling the transcriptional regulatory sequences in complex genome. This approach uses a Bayesian formalism to capture the dependency structure of regulatory elements at two levels---the conservation dependencies between sites within motifs and the clustering of motifs into regulatory modules. It supports major queries related to in silico DNA motif detection, such as learning motif representations, model-based motif prediction, and de novo motif detection. I will also discuss some recent ideas on probabilistic models for motif and enhancer evolution, and outline a novel multi-resolution phylogenetic HMM model for comparative motif detection.

March 3, 2004
Bartlett Mel, Ph.D.
Department of Biomedical Engineering at the University of Southern California
"Modeling pyramidal neurons: from McCulloch-Pitts to multi-layer networks"
Host: Dr. Nathan Urban

One of the central goals of neuroscience is to arrive at a compact functional description of the individual neuron. Our work has focused on the pyramidal neuron, the principal cell of the hippocampus and neocortex. The prevailing view of the pyramidal neuron dates back to the single neuron models of McCulloch-Pitts and Rosenblatt from the mid 1900's, later reinforced by the PDP wave beginning in the 1980's. This view holds that the individual neuron functions as a global summing unit, with a single nonlinearity representing the cell's output spike-generating mechanism. This talk will describe our recent biophysical modeling studies, and in vitro experiments in collaboration with Alon Polsky and Jackie Schiller, that support the view that an individual pyramidal neuron actually functions as a 2-layer network, whose first layer consists of several dozen "sigmoidal" subunits in the thin basal and oblique dendrites of the cell. The functional implications of the 2-layer network model of the pyramidal neuron will be discussed, along with hints that even two layers may not be enough.

January 21, 2004
Ziv Bar-Joseph, Ph.D.
Department of Computer Science and the Center for Automated Learning and Discovery, Carnegie Mellon University
"Computational discovery of gene modules and regulatory networks"
Host: Dr. Elizabeth Jones

Recent advances in high-throughput experimental methods in molecular biology hold great promise. DNA microarrays have been used to measure the expression levels of thousands of genes, and more recently microarrays have been exploited to measure genome-wide protein-DNA binding events. While useful, these datasets present many computational challenges. High-throughput biological datasets are often very noisy and contain many missing points. In addition, each of these data source measures only one type of activity in the cell. Principled computational methods are required in order to make full use of each of these datasets, and to combine them to infer genetic interaction networks. In this talk I will describe an algorithm that efficiently combines complementary large-scale expression and protein-DNA binding data to discover co-regulated modules of genes. I will then present extensions to this algorithm by using time series expression data to automatically infer and validate a dynamic sub-network for the cell cycle system in yeast.

October 1, 2003
Kwabena Boahen
Department of Bioengineering, University of Pennsylvania
"Self-configuring neuromorphic chips through epigenesis"
Host: Nathan Urban

Biology handles complexity through developmental processes, elaborating a relatively simple starting recipe into a complex mature structure. By borrowing from nature, we have developed two self-configuring neuromorphic silicon chips. The first utilizes a model of activity-dependent axonal remodeling to automatically wire a topographic mapping based solely on input correlations. Its silicon growth cones migrate up neurotropin gradients, which are represented by charge diffusing in transistor channels. The second utilizes transistor heterogeneity---introduced by the fabrication process---to generate feature maps similar to those imaged in vivo. Its recurrently connected excitatory and inhibitory cells give rise to hot spots of activity that, when perturbed, cause distinct groups of cells to respond to stimuli of different orientations. Capturing the ability of epigenetic development to generate feature maps and autoroute connections between them provides a powerful alternative to handling complexity. Abstraction, which has been used until now, is becoming increasingly inadequate as silicon chips approach a billion transistors.