Reproducibility

One of the key developments in proteomics was the development of gels which deliver reproducible results. In a reproducibility experiment, three cultures of the same Staphylococcus aureus strain were grown to stationary phase. Protein was extracted from each and separated by 2D gel electrophoresis at Oxford GlycoSciences (OGS).

Black marks on the gel images indicate a protein or cluster of proteins and are referred to as “features.” The x-axis is the Isoelectric point (pI) which is analagous to pH, while the y-axis is molecular weight (Mw) or size. The average feature variance taking in gel-to-gel and biological triplicate was an impressive 23% indicating strong reproducibility, even across different cultures of the same strain. As you can see, the gel images are very similar.

The breakthroughs made by OGS in creating an accurate and reliable electophoresis system now enable tissue-specific gel signatures to be generated. These gel signatures display a range of proteins expressed for any cell type. Incyte and OGS are using this technology platform to determine the protein profiles of different tissues in both healthy and disease states, working toward the creation of the LifeProt™ proteome library.

A pulsar is an extremely dense and small neutron star believed to have been born as a result of a supernova, or explosion of a large star from a class ten or more times the mass of the sun. PSR1257+12 (one of only twenty such stars identified in the galaxy) is about ten miles in diameter, has 1.4 times the mass of the sun, and emits beams of radio waves that have been focused by the star’s extremely strong, approximately dipolar magnetic fields. These signals are analogous to rotating lighthouse beams periodically sweeping across the universe.

Just as a sailor sees flashes of the beacon from a lighthouse, a radio telescope on earth receives pulsar emissions as intermittent but remarkably regular pulses. Scientists are able to detect pulsars by searching for faint periodic radio pulses in the blizzard of information coming from space. Only recently have computers and analytical techniques evolved that are capable of analyzing the millions of samples required to yield results that will stand up to scrutiny.

When searching for pulsars, astronomers examine data collected from the radio sky, or outer space as it appears to a radio, as opposed to an optical telescope. The astronomer’s ability to detect new pulsars is limited by the speed and accuracy of data-recording instruments and by the availability of computing resources. Detecting millisecond pulsars requires tremendous computing power. Samples for research like Wolszczan’s are recorded every 0.3 milliseconds (over 300,000 bits per second). Detecting pulsars is like looking for a subtle repeating color pattern among the blades of grass in a prairie as the breezes wax and wane with the cloud patterns—it’s a job for a supercomputer.

As Wolszczan analyzed the data collected over a 486-day period, he looked for a pattern in the arrival time of the pulses, searching millions of bits of information. Using the supercomputer facilities at the Cornell Theory Center, he determined that the pulses were arriving every 6.2 milliseconds. Wolszczan detected unusual complexity in the pattern of the pulses’ arrival times: they periodically arrived early and bunched together and then spread apart, as they began to arrive later than the predicted time. This behavior suggested that the pulsar’s motion is affected by the presence of other orbiting objects. Instead of moving steadily, PSR1257+12 was being pulled around a point in space called the barycenter—the center of mass of a system—by the gravitational interaction with these objects.

When Wolszczan tries to account for the forces causing the star to orbit, he comes up with a remarkable explanation. Wayne Lytle, visualization specialist at the Cornell Theory Center, has produced an animation of Wolszczan’s data that presents the pulsar orbiting around the barycenter of a system having two planets. These two planets are themselves orbiting every 66.6 days and 98.2 days respectively. If you allow for the gravitational pull of these two planets in the calculations, the rest of the problematic pattern is explained.

Since its founding in 1985, the Cornell Theory Center has been a leader in national and international high performance computing. Through partnerships with government, industry, and other academic institutions, the center and its users continue to advance the limits of high performance computing and to extend its application. Scientists and engineers advance and refine their research using the resources of the Cornell Theory Center. Incorporating the feedback of these researchers allows the computer industry to improve supercomputing technology itself, thus fueling a cycle of progress that has reached levels unimaginable ten years ago. Such rapid developments in computational technologies contribute to the more effective approaches to problem solving, to new and improved products and services, and to an enhanced overall national competitiveness.

Growing from a small base of Cornell users in 1985, the Cornell Theory Center’s research community now includes more than 2,300 users, representing more than 150 institutions nationwide. The research engages virtually every scientific and engineering discipline, from simulations of the Earth’s vibrations to visual imaging of Jupiter’s clouds; from the dynamics of ecosystems to the dynamics of the world economy; from discovering pulsars to analyzing plasma. More than 500 research projects from a variety of disciplines—including biological, behavioral, and social sciences; computer and information sciences; engineering; geosciences; and mathematical and physical sciences—are conducted using Cornell Theory Center resources.

The Theory Center receives major funding from the National Science Foundation and New York State. Additional funding comes from the Advanced Projects Research Agency, the National Institutes of Health, IBM Corporation, and other members of the center’s Corporate Research Institute.

The Biomolecular Modelling Laboratory employs molecular structure to assist in the ICRF objective of research into the cause, prevention, treatment and cure of cancer. Our approach is based on the use of computer algorithms to model the structures of biological molecules and their interactions.

The key modelling areas are:

(1) the prediction of protein structure from sequence;
(2) the docking of molecules especially protein/protein complexes;
(3) the relationship of structure to activity for pharmaceutical molecules.

A variety of approaches are used to develop algorithms. The key techniques for development of algorithms in the above three modelling areas are respectively:

(1) analysis of the structures of known protein to obtain empirical rules
(2) better understanding and representations of energetic effects in proteins
(3) applications of inductive logic programming

The Laboratory aims both to develop algorithms in these areas and to apply modelling to systems studied by other groups in the ICRF.

SWISS- PROT differs from other protein databases by three fundamental criteria: A) It contains a wealth of information which are called annotations and which consist, for a given protein, in the description of its function, its post- translational modification(s), the extent of specific domains and sites of biological interest, its secondary and quaternary structures, its similarities to other proteins, the disease(s) associated with deficiencie(s) in the protein, as well as various other topics. B) Minimal redundancy.

SWISS-PROT provides a single entry for each individual sequence. Data derived from different literature reports are merged thus providing a synthetic view of the current knowledge with respect to a given protein. C) It is integrated with other databases. Cross-references are provided to many other databases including nucleotide sequence databases, the protein 3D structure database (PDB), as well as specialized data collections

The rapid changes in computational environments in general, and in high performance computing in particular, compel us to rethink the ways in which Computational Molecular Biology (CMB) is done. The Resource is taking major new steps to address the new challenges. We offer a wide range of software tools developed in our core research projects that will run on diverse computing platforms. We also focus on making computation more accessible to non-computational scientists through a unique combination of software and hardware that make simulations and computations easier to use.

We continue to:

1. Develop core theories, algorithms, and software;
2. Make emerging software tools available for widely accessible computing platforms, including desktop PCs, as well as high-performance computers;
3. Provide straightforward and user-friendly interfaces to a hierarchical computing environment, starting at the desktop and ending at high performance computers;
4. Make web (e-computing) and network computing (net-computing) an integral part of the resource.

Arabidopsis thaliana is a small, fast-growing plant in the mustard family that has its genetic code stored in a conveniently compact package. Thus it became the model system for plant genome research. However, even this modest plant has a complex evolutionary history, and researchers have determined that Arabidopsis has doubled its genetic makeup repeatedly over time, with most of the activity occurring during the age of the dinosaurs.

Vision has devised a method of data analysis that makes it possible to see the faded tracks of these dramatic events in a representation of the modern genetic map, and then to reconstruct the ancient images.

“If you can reverse history by reconstructing what the map looked like in the distant ancestors of Arabidopsis, then you are well on your way to figuring out where each gene should be in the map of all the important crop plants that descended from that same ancestor. This is important because we don’t have nearly as detailed a map for most crops as we do for Arabidopsis,” says Vision. Herein lies the key: based on the map of the modern Arabidopsis genome, researchers will be able to track the genes through time to their locations in modern crops. As biologists determine the function of known genes in arabidopsis, they can use this system to find and test these genes for their functions in other species. Some will have retained the same functions; many others will have changed, but the information about their heritage will be very valuable.

The figure shows the positions of duplicated genes.Red indicates that both paired genes are oriented in parallel, blue that they are oriented in opposite directions. Duplicated blocks of genes are largely composed of runs of one color since a change in color between adjacent matches within a block suggests an inversion. The relative strength of the match is shown by color intensity; brighter matches are of higher rank than duller ones.

Vision and his colleagues used cluster computing resources at Cornell, including a high-performance Dell/Intel/Windows cluster at the Cornell Theory Center (CTC) funded by the USDA, to conduct his research. The team’s method combines standard software in a novel way that allows the computer to see patterns in the genomic sequence. The core of their method is a graph theoretic algorithm implemented by Brown in MATLAB that repeatedly identifies duplicated segments in the genome from thousands of automatically generated random samples. While the method is very computationally intensive, it allows researchers to assign statistical confidence to the patterns that they see for the first time. This tool is embedded in a powerful combination of bioinformatics tools including BLAST (Basic Local Alignment Search Tool), a program that searches gene and protein databases for similar sequences.

The methods developed through this collaboration will likely be picked up and used by other researchers for finding related chromosome segments both within individual genomes and between related genomes. For example, Tanksley has been studying the relationships between the genomes of Arabidopsis and tomatoes.

“Vision’s work demonstrates that basic inquiry, in this case into genomic evolution, can yield useful tools for the broad research community,” says CTC executive director Linda Callahan.

The CAB is supported by the U.S. Department of Agriculture, Agricultural Research Service, in partnership with the College of Agriculture and Life Sciences and CTC. CTC is a high-performance computing and interdisciplinary research center located at Cornell University. CTC receives funding from Cornell University, New York State, a number of federal agencies, and Corporate Program members.

The polypeptide fragments for each protein are then separated by capillary electrophoresis and analyzed using rapid-throughput mass spectrometry. At this point, we know the amino acid sequence of the polypeptide fragments, their mass, as well as post-translational modifications that occured such as glycosylation and phosphorylation.

Genomics Enters the Picture

Once the amino acid sequence is determined for each protein excised from the gel, these protein “fingerprints” can be checked against existing protein and DNA sequence databases. Incyte’s LifeSeq® database is essentially a vast library of predicted peptide (protein) sequences. When a matching sequence is found, it can be used to annotate and help predict the full amino acid sequence of the protein. This is where the power of genomics and proteomics comes together to create a “best of both worlds” discovery platform.

LifeSeq® database is one of the world’s largest sources of genomic data. This commercial database of human gene expression and sequence information is used daily by scientists at more than 50 pharmaceutical research and development sites worldwide to identify therapeutic targets, to develop new approaches for diagnosing disease, and to understand the pharmacological and toxicological impact of new drugs on human tissues.

Point-and-Click Biology

Today, the path from gene discovery to drug development starts at the computer, where scientists can study biology thousands of genes at a time. LifeSeq software provides the data and the enabling tools to explore this wealth of genomic information, including the electronic equivalents of biological experiments such as Northern Blots and Library Subtractions. Using simple point-and-click commands, researchers can navigate through the database to retrieve vital information in just seconds, literally “biology in silico,”eliminating weeks of work in a traditional laboratory.

The sequence analysis and assembly tools built into LifeSeq sofware also allow scientists to compare gene sequences, to assign putative functional characteristics, to assemble consensus sequences, to identify polymorphisms, and even to “clone” new genes. New Java-enabled template viewers make it easier than ever to quickly assess results from BLAST searches for putative homologs and splice variants allowing the user to drill down to nucleotide level differences.
Bioinformatics Excellence

Incyte combines high-throughput DNA sequencing with state-of-the-art bioinformatics to identify and characterize the expressed genes of the human genome. This proprietary data is integrated with publicly available EST data, including sequences from the Washington University-Merck EST Project, The Institute for Genomic Research (TIGR), and the Cancer Genome Anatomy Project (CGAP) to create a powerful, information-based tool for drug discovery and development.

LifeProt contains annotated protein expression data for numerous tissues. Researchers can investigate 2D gel images on screen, looking at identified proteins, drilling down to amino acid sequence data or linking to matching ESTs in the LifeSeq® human gene sequence database. LifeProt is an integrated gene expression analysis system, bringing together the best of proteomics and genomics.

The Promise of Protein Profiles

As more and more reference gel images for different tissues are added to LifeProt, it should be possible to identify most proteins by their position on the gel. In this selected portion of a 2D gel, two features were annotated as differentially modified forms of the same protein.