2004-2005 Series

Pavel Pevzner
Despite some differences in appearance and habits, men and mice are genetically very similar.  In a pioneering paper, Nadeau and Taylor, 1984 estimated that surprisingly few genomic rearrangements (about 200) have happened since the divergence of human and mouse 75 million years ago.  The genomic sequences of human and mouse provide evidence for a larger number of rearrangements than previously thought and shed some light on previously unknown features of mammalian evolution. In particular, they provide evidence for extensive re-use of breakpoints from the same relatively short regions and reveal a great variability in the rate of micro-rearrangements along the genome.  Our analysis also implies the existence of a large number of very short "hidden" synteny blocks that were invisible in comparative mapping data and were ignored in previous studies of chromosome evolution. These results suggest a new model of chromosome evolution that postulates that breakpoints are chosen from relatively short fragile regions that have much higher propensity for rearrangements than the rest of the genome.

Gene Myers
The whole genome shotgun sequencing method with paired end-reads has proven rapid and economical, producing high-quality reconstructions of Drosophila (2000), Human (2001) and Mouse (2001), in quick succession. We discuss the overall algorithmic strategy, and the results one can expect by comparing the whole genome assembly of Drosophila against the recently finished sequence, and advances such as high-density solid state sequencing and single molecule detection systems. We anticipate having the euchromatic portions of the genomes of twelve species of Drosophila in the next year. We discuss the current state of the art in comparative gene finding, cis-control module finding, and possible improvements. The hope of these approaches is that we will be able to accurately identify the "parts lists" of the D. melanogaster genome, a basic prerequisite for systems biology. We conclude with a segment on the possibility of a program of high-throughput in-situ image analysis in Drosophila embryos. We describe what information we might collect and what we might be able to infer form it. It is our contention that this may be the best way to understand development from a systems perspective.

Stephen Smale
The talk will pass from some answers to the question, "What is learning theory?" to statements of recent results on error estimates. An attempt will be made to speak to a broad audience of scientists and maintain mathematical rigor at the same time.

Eric Lander
(No abstract available)

James Ferrell
yet to come...


UNDER CONSTRUCTION...

2003-2004 Series

Russ Altman
Bioinformatics and computational biology have emerged primarily in response to the deluge of data from DNA sequencing, mRNA expression analysis, proteomics, and other high throughput data collection techniques. In many ways, the field has been driven be a frenzied need to gather, store and do preliminary analyses on these data. As the field has matured, however, it has begun to set sights on longer term "grand challenges." I believe that the management of biological knowledge--models that are too complex for individual humans to track--will be a primary challenge for biomedical computation. In this talk, I discuss the move from building "power tools" for biologists, to building "assistants" for biologists (and then ultimately, perhaps, building biologists?). Some of the infrastructural needs are already clear, although others have not yet been defined.

Michael Arbib
This talk is a prospectus for an approach to computing that focuses on analyzing the architecture of the primate brain to extract Brain Operating Principles that may be translated into biologically-inspired operating systems and computer architectures. The approach will provide a principled combination of cooperative computation and learning with the integration of action, perception and computation , and will be motivated by computational neuroscience studies of system evolution ranging from low-level vision to how the interactions between frontal and parietal cortices serve action, action recognition (the mirror system), and language. But there is a vast difference between the slow biological evolution of diverse brains and the needs of computer technology for explicit tools for the design and testing of novel programs and architectures.This raises the challenge of developing a reflection methodology for wrapping modules with descriptions that can be automatically updated as the modules adapt through learning. Such "wrappings" provide the outer layer of both functional and structural modules abstracted from neuroscience, and should serve to enable the analysis of such modules, and their retrieval and adaptation for use in new assemblages to handle novel tasks. The talk will describe an "intelligent room" to ground a progress report on this work.

Mark Borodovsky
Expectation-maximization (EM) algorithms are quite popular in Bioinformatics. They can be used for solving various problems  from multiple sequence alignment, prediction of binding sites and promoters to building phylogenetic trees and gene networks. I will talk about use of the EM algorithms for gene identification in prokaryotic and eukaryotic genomes. These EM algorithms will generalize GeneMark and GeneMark.hmm gene finding algorithms which can be defined in terms of posterior decoding (forward and backward algorithm) and Viterbi algorithm respectively.

Fred Cohen
The transmissible encephalopathies including Kuru, Creutzfeldt-Jakob Disease and Bovine Spongiform Encephalopathy appear to be caused by a proteinaceous infectious agent or prion. I will describe theoretical and experimental studies of the prion protein that help to explain how one disease can present in infectious, sporadic and inherited modes and how a protein can "replicate" in vivo . Current work suggests that the transmissible encephalopathies are diseases of protein folding where the tertiary structures of the normal cellular isoform, PrPc, and the disease causing isoform PrPSc, are distinct while their covalent structures are identical. Work on peptides refolding into b-rich conformations that initiate neuropathology in transgenic models of prion diseases as well as work on the design of small molecule inhibitors of prion replication will be discussed.

C. Lee Giles
Internets and intranets are offering a revolution in information access and creation. More and more information is being digitized (exabytes according to Lyman, Varian [HV00]) and the microsensor revolution is just starting. Much of this digitized information will be available online and will be in many formats from raw data capture to text and tagged data to databases. Some of it will be dynamic and only available for a certain amount of time; some will be a mix of proprietary and open sources. For individuals and organizations, Moore’s law and its impact on storage and computation make digital immortality [GB01] a viable option for information preservation. How will this information or knowledge be stored, protected and accessed? Will databases and information retrieval systems have to become more intelligent? This talk will discuss new models of information access, mining and storage, such as web graphs [DP02], and discuss the roles of intelligent search and portals. CiteSeer [CG98] will be used as an example of intelligent portal and information preservation archive. CiteSeer’s future directions are discussed [YP03].

[CG98] C.L. Giles, K. Bollacker, S. Lawrence, “CiteSeer: An Automatic Citation Indexing System,” DL'98 Digital Libraries, Third ACM Conference on Digital Libraries, 89-98, 1998.
[DP02] D.M. Pennock, G.W. Flake, S. Lawrence, E.J. Glover, C.L. Giles, “Winners don't take all: Characterizing the competition
for links on the web.” Proc. Natl. Acad. Sci. USA 99(8): 5207-5211, 2002.
[GB01] G. Bell, J. Gray: Digital immortality. CACM, 44(3): 28-31, 2001.
[HV00] H. Varian, P. Lyman, “How Much Information,” SIMS, U. of California, Berkeley, CA, Oct, 2000.
[YP03] Y. Petinot, P.B. Teregowda, H. Han, C.L. Giles, S. Lawrence, A. Rangaswamy, N. Pal, “eBizSearch: An OAI-Compliant
Digital Library for eBusiness,” ACM/IEEE Joint Conference on Digital Libraries, JDCL 2003, 2003 (accepted).

Jonathan Moreno
In this talk I reconstruct the complex history of human experiments and how current regulations and guidelines have evolved. The origins of these standards date at least to the early 1900s. The Nazi medical experiments that resulted in the Nuremberg Code and the Cold War experiments, emphasize the largely unappreciated national security aspect of this history, as compared to the widely publicized syphilis study. The talk concludes by noting some of the current issues that continue to challenge the ethics of human experiments.

Clarence Peters
When the viruses come, do we look to science or society?   Contrary to generally held beliefs forged in the era of highly effective vaccines and antibiotics for common diseases, in the last two decades it has become apparent that infectious diseases such as SARS, West Nile virus, and mad cow disease will increasingly threaten humans and other species on the planet.  The single most important factor driving these emerging and re-emerging diseases is our society--we dominate the earth and profoundly change the environment.  Globalization not only affects markets, but also impacts disease characteristics and transmission in our age of intercontinental trade and travel.  Today we look to science to spare us from disease and death, but will this continue to be our talisman?  Peters argues that we must anticipate and respond through social means to combat the emergence and spread of new infectious diseases.  Examples of dangerous microbes recently reckoned with include uncontrollable dengue hemorrhagic fever world-wide, Ebola virus in Africa and the movies, annual battles with influenza, and the recent “close shave” with SARS.

Tomaso Poggio
The problem of learning is one of the main gateways to making intelligent machines and to understanding how the brain works. In this talk I will give a brief overview of our recent work on learning theory, including new results on predictivity and stability of the solution of the learning problem. I will then describe recent efforts in developing machines that learn in applications such as visual recognition, computer graphics and bioinformatics.

Relevant papers can be downloaded from http://www.ai.mit.edu/projects/cbcl/publications/all-year.html.