Seminars

Over the last dozen years, we have seen a number of information technology advances in our understanding of and tools to handle biological data. Work on genomic and proteomic databases had many successes, but it also exposed some limitations of ordinary databases and also the fragility of connections between different databases. Similarly, we have made great strides in defining molecular pathways by showing regulations schemes where one protein is regulated by a second which in turn regulates others. It is only within the last few years that a number of these models have become capable of providing quantitative prediction. This talk is about trying to put these two areas - Models and Databases - in a common informatics context such that they work together.

Dr. Califano's scientific interests lay in the investigation of Systems Biology, using a variety of physics and knowledge-based methods. Since 1998 he has been especially active in the development of integrative methodologies for the investigation of human B cell lymphomas. Specifically, he is interested in the reverse engineering of B cell cellular networks and in their use for the dissection of biological processes related to oncogenesis and tumor progression.

A fundamental problem in discriminant analysis concerns the classification/prediction of an entity into one of several a priori, mutually exclusive groups based upon specific measurable characteristics of the entity. Typically, a classification/predictive rule is formed from data collected on a sample of entities for which the group classifications are known. Then new entities can be classified based on this rule. Such an approach is important to medical diagnosis where often a definitive classification of a patient can be made only after exhaustive and invasive physical and clinical assessments. Hence, a challenging avenue of research involves developing sophisticated predictive models that allow accurate early diagnosis.

In this talk, we present novel optimization-based classification models that are general purpose and suitable for developing predictive rules for large heterogeneous biological and medical data sets. Our predictive model simultaneously incorporates (1) the ability to classify any number of distinct groups; (2) the ability to incorporate heterogeneous types of attributes as input; (3) a high-dimensional data transformation that eliminates noise and errors in biological data; (4) the ability to incorporate constraints to limit the rate of misclassification, including a reserved-judgment region that provides a safeguard against over-training (which tends to lead to high misclassification rates from the resulting predictive rule); and (5) successive multi-stage classification capability to handle data points placed in the reserved judgment region. The classification models have been applied successfully in a wide variety of applications ranging from finance, forensics, to industrial and biomedical applications. In this talk, application of the predictive model to a broad class of biological and medical problems is described. Applications include: genomic analysis and prediction of aberrant CpG island meythlation in human cancer; discimination of various localization sites of proteins; discriminant analysis of motility and morphology data in human lung carcinoma; prediction of ultrasonic cell disruption for drug delivery; identification of tumor shape and volume in treatment of sarcoma; differential diagnosis of erythemato-squamous dermatology disease; multistage discriminant analysis of biomarkers for prediction of early atherosclerois; and fingerprinting of native and angiogenic microvascular networks for early diagnosis of diabetes, macular degeneracy and tumor metastasis. In all these applications, the predictive model yields correct classification rates ranging from 85% to 100%. This provides motivation for pursuing its use as a medical diagnostic, monitoring and decision-making tool.

Sampling the Energy Landscape: Thermodynamics and Rates for Biomolecules,
Clusters and Glasses

Stationary points of the potential energy surface provide a natural way to
coarse-grain calculations of thermodynamics and kinetics, as well as a
framework for basin-hopping global optimisation. Thermodynamic properties
can be obtained from samples of local minima using the basin-sampling
approach, and kinetic information can be extracted if the samples are
extended to include transition states. Using statistical rate theory a
minimum-to-minimum rate constant can be associated with each transition
state, and phenomenological rates between sets of local minima that define
thermodynamic states of interest can be calculated using a new graph
transformation approach. Since the number of stationary points grows
exponentially with system size a sampling scheme is required to produce
representative pathways. The discrete path sampling approach provides a
systematic way to achieve this objective once a single connected path
between products and reactants has been located. In large systems such paths
may involve hundreds of stationary points of the potential energy surface.
New algorithms have been developed for both geometry optimisation and making
connections between distant local minima, which have enabled rates to be
calculated for a wide variety of systems.

D.J. Wales, "Energy Landscapes", Cambridge University Press (2003)
D.J. Wales and T.V. Bogdan, J. Phys. Chem. B, 110, 20765-20776 (2006)
D.J. Wales, Phil. Trans. Roy. Soc. A, 363, 357-377 (2005)

The electrical activity of neurons and neuronal networks arises from the
complex and non-linear interactions of many cellular and synaptic
players, such as voltage-dependent membrane currents and synaptic
transmitters and receptors. Computational modeling provides a valuable
tool for the study of these interactions. Astrid Prinz will describe the
use of brute-force computational exploration of neuron and network model
parameter spaces to investigate how individual cellular and synaptic
components shape electrical signaling in neural systems. Tim Hickey will
discuss how the high-dimensional datasets generated by such parameter
space explorations can be visualized and analyzed with tools from
computer science.

Over the last dozen years, we have seen a number of information technology advances in our understanding of and tools to handle biological data. Work on genomic and proteomic databases had many successes, but it also exposed some limitations of ordinary databases and also the fragility of connections between different databases. Similarly, we have made great strides in defining molecular pathways by showing regulations schemes where one protein is regulated by a second which in turn regulates others. It is only within the last few years that a number of these models have become capable of providing quantitative prediction. This talk is about trying to put these two areas - Models and Databases - in a common informatics context such that they work together.

Identifying genes underlying complex diseases or traits is a challenging problem in genetic studies. Traditional approaches include linkage and association studies and gene expression analysis. Many different sources of genomic data, such as protein-protein physical interactions, genetic interactions, regulation networks, and gene expression, are available. We developed several methods to integrate different data sources for gene discovery and for finding the pathways related to the traits. The approaches were validated and tested using genetic polymorphisms, gene expression profiles, and protein interaction in yeast. Several interesting pathways were obtained.

The combination of high-throughput methods of molecular biology with advanced mathematical and computational techniques has propelled the emergent field of systems biology into a position of prominence. Unthinkable only a decade ago, it has become possible to screen and analyze the expression of entire genomes, simultaneously assess large numbers of proteins and their prevalence, and characterize in detail the metabolic state of a cell population. While very important, the focus on comprehensive networks of biological components is only one side of systems biology. Complementing large-scale assessments, and sometimes at risk of being forgotten, are more subtle analyses that rationalize the design and functioning of biological modules in exquisite detail. This intricate side of systems biology aims at identifying the specific roles of processes and signals in smaller, fully regulated systems by computing what would happen if these signals were lacking or organized in a different fashion. In this presentation I will exemplify this type of approach with a detailed analysis of the regulation of glucose utilization in Lactococcus lactis. This organism is exposed to alternating periods of glucose availability and starvation. During starvation, it accumulates an intermediate of glycolysis, which allows it to take up glucose immediately upon availability. This notable accumulation poses a non-trivial control task that is solved with an unusual, yet ingeniously designed and timed feedforward activation system. The elucidation of this control system required high-precision in vivo data on the dynamics of intracellular metabolite pools, combined with methods of nonlinear systems analysis, and may serve as a paradigm for multidisciplinary approaches to fine-scaled systems biology.

The complete sequencing of a human reference genome was a remarkable technological event achieved by implementing an industrial model that has consistently produced ever-greater quantities of data at a reduced cost. However, it increasingly appears unlikely that current approaches are sufficiently scalable to fulfill the promise of individual genomic medicine, and the resequencing of multiple human genomes. Next generation DNA sequencing technologies, that require relatively few people and limited laboratory space, are being touted as possible solutions. These technologies will require improved computational algorithms to succeed. Resequencing arrays (RAs), a technology that offers the promise of higher sequencing throughput with dramatically reduced costs, will be discussed. The experience and lessons learned from developing the ABACUS computer algorithm, currently the industry standard for interpreting RA data, will be presented. Challenges facing next generation DNA sequencing technologies - that include the development of robust computational and data quality algorithms, will be examined.