Over time, the work could help doctors diagnose the seriousness of melanoma cases more quickly and with better accuracy. In addition, this work could lead to new tools for outcome prediction, thus assisting doctors in determining best treatment approaches.

Nancy E. Thomas, an associate professor and M.D. in the School of Medicine’s dermatology department and a researcher with the Lineberger Comprehensive Cancer Center, will lead the research project, called Image Analysis to Assess Melanoma Heterogeneity. The project will examine imagery from more than 1,300 melanoma patients worldwide, including 214 from North Carolina. A $190,000 award from the University Cancer Research Fund will support the project. The North Carolina General Assembly created the fund in 2007 to accelerate the battle against cancer at UNC Chapel Hill’s School of Medicine and its Lineberger Comprehensive Cancer Center.

RENCI specialists in bioinformatics, image analysis and data mining will work with Dr. Thomas and melanoma researchers at UNC Chapel Hill to develop algorithms or computational methods that can identify cancerous and healthy tissue in high-resolution images. Once cells in the images are identified as cancerous or healthy, the researchers will collect information on physical details, such as cell size, shape, and color of melanoma cells.  All these details will be used to develop evidence-based models of melanoma cell descriptions.

In addition, advanced computational methods will be used to identify the physical characteristics of the melanoma cells–features uncovered through image analysis–which are associated with tumor classification, based on known information about somatic genetic mutations in melanomas, and survival.

“What’s exciting about our approach is that we may be able to uncover relationships between the physical qualities of melanoma cells, tumor genotype, and patient outcomes” said Charles Schmitt, manager of biological science programs at RENCI and a co-investigator on the project. “It’s the kind of work that can only be done by bringing together expertise in multiple areas.”

Nearly 60,000 new cases of melanoma were diagnosed in the U.S. in 2007, according to estimates from the American Cancer Society, and more than 8,000 patients died of the disease. Melanoma cases continue to increase and prognosis is poor for patients with advanced melanoma tumors.

“The objective of this proposal is to utilize image analysis to uncover associations between the physical characteristics of melanoma, somatic mutations in melanoma, and survival” said Thomas. “But ultimately, our long-term goal is to help patients.  We want to use image analysis to improve melanoma classification, which we would expect to improve diagnosis and guide treatment recommendations.”

RENCI…Catalyst for Innovation
The Renaissance Computing Institute brings together computer and discipline scientists, artists, humanists, industry leaders, entrepreneurs, state leaders and educators for collaborations designed to reshape science, the economy, the state of North Carolina and the world. RENCI leverages its expertise and resources in leading edge computing, networking and data technologies to ignite innovation and find solutions to previously intractable problems. Founded in 2004 as a major collaborative venture of Duke University, North Carolina State University, the University of North Carolina at Chapel Hill and the state of North Carolina, RENCI is a statewide virtual organization.  For more, see www.renci.org.

Even though the buzz about eScience often focuses on massive hardware, user interfaces, storage capacity and other technical issues, in the end, the ability of eScience to serve the needs of research teams boils down to people: the ability of infrastructure builders to communicate with scientific communities and understand their needs and the realities of their work cultures.

The builders of eScience infrastructure “need to talk about fostering, rather than building infrastructure,” said Alex Voss of the National Center for e-Social Science in Manchester, UK, and research theme leader at the e-Science Institute in Edinburgh, UK. There are social aspects to research that must be recognized—from understanding how research teams work and interact to realizing that research often does not involve the kinds of large, interdisciplinary projects engaged in by virtual organizations, but rather individual work and ad-hoc, flexible forms of collaboration within wider communities.

Voss was one of four panelists who discussed how to reduce the barriers that still inhibit scientists from becoming e-scientists. The discussion was part of the 2007 Microsoft eScience Workshop, hosted by the Renaissance Computing Institute (RENCI), in Chapel Hill, NC, Oct. 21 – 23. Also offering their thoughts on the barriers to broad eScience adoption were Ian Foster, director of the Computation Institute at the University of Chicago and Argonne National Laboratory, Phil Papadopoulus, director of grid and cluster computing at the San Diego Supercomputer Center, and May D. Wang of the Georgia Institute of Technology/Emory University School of Medicine and director of Biocomputing Core at the Emory/Georgia Tech Center for Cancer Nanotechnology Excellence.

All the panelists agreed that scientific communities must have easy-to-use applications and interfaces and easy access to stored data to become users of eScience grids. And they concurred that both grid researchers and users must work on cross-disciplinary and cross-cultural communications.

“We need applications, that’s obvious,” said Foster. “But perhaps we need to put more effort into communicating how these applications work. That’s probably the single thing we can do that will make the biggest difference: go out and tell the story about successes when applications work, and also tell them when applications don’t work, so they can avoid the pitfalls.” 

Foster noted wryly that in the past, “science advanced one funeral at a time,” allowing new ideas to take hold when those who advocated older paradigms passed on. On a more positive note, he said the ubiquitous connectivity offered by the Internet, the Web and grids “allows us to reach out, to share our interests, make discoveries, and apply new methods more rapidly and effectively than in the past.”

Papadopoulus pointed to both technical and social barriers to the adoption of eScience. Although raw storage is cheap, access to data isn’t, he said, and the eScience community must address questions about how to access data that is stored remotely, stored offline or behind firewalls.  Papadopoulus also challenged infrastructure creators to develop systems that are repeatable. A set of software tools should be transferable to any user’s work environment, without the aid of a systems administrator. The steps of a workflow should be repeatable and easy to communicate to another user.

In addition, he noted that the social realities of scientific communities can inhibit the adoption of eScience. Scientists in some domains have only recently started to share their data, a process that is the norm in well-established eScience domains, such as high-energy physics. The grid research community also has its customs that can inhibit broader adoption, according to Papadopoulos.

“Grid research is research, and researchers are rewarded for their research, for coming up with new ideas on how to use network technology and for writing papers, not really for easing the use of software,” he said.

Wang, an expert in biocomputing and bioinformatics, speculated on why the biomedical community has been relatively slow to adopt eScience practices. She stressed that eScience tools must be more intuitive for the biomedical community to use them. These researchers—often doctors with clinical practices—have little in-depth knowledge of computing and no time to learn it, said Wang. They are problem driven and will turn to eScience only if they see that it will help them address the big questions in medicine. In addition, the medical community would likely feel more at home with eScience if some general computer science were part of their education.

“Teaching the basics of computer science, learning some of the computer science languages and how to use computer tools to solve problems would help to overcome some of the barriers,” said Wang. “Now, many of our scientists wouldn’t even know how to begin a dialogue with a computer scientist. But they can learn by doing if they start at a young age.”

More than 260 scientists, industry and university-based grid researchers, faculty and administrators with funding agencies attended the Microsoft eScience Workshop, which was co-chaired by RENCI Director Dan Reed and Microsoft’s Vice President of External Research Tony Hey. Participants came from across the U.S., Europe, Canada, South America and Australia.

In the long run, the lasting effects of high-speed networks, data stores, computing systems, sensor networks, and collaborative technologies that make eScience possible will be up to the people who create it and use it, said Reed in his address to attendees.

“The instrumented life—in which we have biomarkers for disease risks, real-time monitoring of our food intake and exercise routines, analysis of air quality and other environmental factors—could seem like 1984 rather than 2010,” said Reed. “On the other hand, it could have enormous implications for improving our health and our lives. Is it good or bad? Probably a little of both.”

The conference wrapped up on Tuesday with a keynote session featuring Hey and David Heckerman, also of Microsoft Research. Heckerman told the audience about research that applies his machine-learning technologies to computational biology and personalized medicine. The work could play a role in developing effective vaccines for HIV and AIDS. Heckerman’s statistical models, sometimes called graphical models or Bayesian networks, can also be used for genome-wide association studies—the search for connections between human DNA and disease.

Hey’s talk, called eScience and Digital Scholarship, looked towards tools and technologies required for the whole eScience Data Life Cycle and a coming revolution in scholarly communication. He concluded that the future of eScience will be a mix of software and services “in the cloud.”

Enter workflows, technology that automates many redundant processes used in analyzing biological problems, such as launching a group of applications and moving data among applications. Workflows aim to take the “busy work” out of science. Instead of reentering data in countless different formats to accomplish different steps in problem solving, the scientist simply loads the data into the front end of the workflow and lets the underlying infrastructure handle the busy work.

Workflows are one of the newest features of the North Carolina/TeraGrid Bioportal, a shared, extensible web portal environment developed by RENCI that brings together more than 140 computational tools and applications and many standard biological data sets. Responding to the needs of Bioportal users for broader, more sophisticated biological research capabilities, RENCI research programmers began integrating workflows into the Bioportal last year.

When using a workflow, the Bioportal user sees a single meta application wrapped in the familiar Bioportal interface. Under the hood, however, ther’s a whole lot of analysis going on. The Bioprtal’s first workflow, called Gene2Life, captures a sequence of bioinformatics analysis steps into a single meta step, saving the user time and, quite possibly, frustration. Below is a simplified breakdown of Gene2Life operations. Over the next few months, RENCI researchers will add additional workflows to the Bioportal and finetune the Bioportal infrastructure to allow users tocreate and deploy their own customized workflows.

Gene2Life overview: Gene2Life uses as input a DNA or protein sequence and compares that sequence to current databases of known sequences to determine its closest relatives, generating a “tree” to depict relationships.

• User logs on to the Bioportal and specifies a DNA or protein sequence of interest
• Sequence is supplied to BLAST (Basic Local Sequence Alignment Tool), a tool for rapid searching of nucleotide and protein databases
• Sets of results are analyzed for their degree of matching to the starting sequence—the good matches are retained,
• Workflow divides the BLAST from the input data and fetches BLAST-identified sequence identities from the Internet.
• Fetched and input sequence are globally aligned using a clustalW process on the grid.
• Globally aligned results are sent to a parsimony program (dnapars or protpars) and then to a drawing program (drawgram) to display the phylogenetic tree.

The panel is part of an ongoing discussion series on data related issues. The series, which began in July, was developed to stimulate a dialogue and develop strategies for meeting the growing data storage, management, processing and analysis needs of faculty at Triangle area universities. It grew out of collaboration among RENCI, the Triangle Universities Center for Advanced Studies, Inc. (TUCASI) and the Triangle Area Research Libraries Network (TRLN). That collaboration is working to develop and deploy a large-scale, high performance data storage system that will include large scientific and medical data sets and interactive real-time databases. It also explores retrieval and search capabilities that work with a wide variety of data sets and file formats.

Panelists for the Oct. 26 discussion are:

• Dahlia Nielson, research assistant professor, statistics, North Carolina State University Bioinformatics Research Center.

Nielson develops techniques for fine-scale genetic mapping in human populations, using population history to help locate genes involved in the expression of hereditary diseases.  Her work in the Bioinformatics Research Center pulls together research and data from colleagues in a cross-section of departments.

• Charles Perou, assistant professor, pathology and genetics, Lineberger Comprehensive Cancer Center, UNC School of Medicine.

Perou’s laboratory characterizes the biological diversity of human tumors using DNA microarrays, molecular genetics, and cell biology. The goal is to classify tumors into subtypes of clinical relevance.

• Saianand Balu, manager, Bioinformatics Group, Lineberger Comprehensive Cancer Center, UNC School of Medicine.

Balu’s group provides support to Perou’s laboratory and to the biostatistics and data management groups at the Lineberger Center.

• Terrence Furey, assistant research professor, biostatistics and bioinformatics and computer science, Institute for Genome Sciences and Policy, Duke University.

Furey’s research focuses on genome sequence analysis and has included computational analysis and validation of the human genome sequence. He is co-creator of the UCSC Genome Browser (link: genome.ucsc.edu)

Additional discussions in the series will be held Nov. 2 and Nov. 16; check the RENCI calendar page (link http://www.renci.org/news/calendar.php) for details. Those interesting in attending the Oct. 26 panel discussion should contact Leesa Brieger at leesa@renci.org. RENCI…Catalyst for Innovation
The Renaissance Computing Institute brings together computer and discipline scientists, artists, humanists, industry leaders, entrepreneurs, state leaders and educators for collaborations designed to reshape science, the economy, the state of North Carolina and the world. RENCI leverages its expertise and resources in leading edge computing, networking and data technologies to ignite innovation and find solutions to previously intractable problems. Founded in 2004 as a major collaborative venture of Duke University, North Carolina State University, the University of North Carolina at Chapel Hill and the state of North Carolina, RENCI is a statewide virtual organization.

Dr. William Kaufmann, a professor of pathology and laboratory medicine at UNC’s School of Medicine, is working with a RENCI team led by Dr. Xiaojun Guan on overcoming some of the thorniest technical challenges in understanding melanoma; their work could lead to better treatments for the deadly disease. Those challenges include pulling together biological data from disparate databases and experiments, merging data sets into comprehensive data visualizations, and developing standard methods for visualizing genetic pathways, which is the series of interactions that take place among related genes that lead to mutations and the development of cancer cells.

“These scientists work with data from their own research subjects and from databases of protein interactions that are accessed through the Internet,” said Guan. “Very often the data are in different formats and use their own nomenclatures, which makes it difficult to do the large-scale comparative studies that are needed in this field.”

Kaufmann’s research looks at the genetic changes in skin cells brought about by ultraviolet radiation. Radiation damage, often triggered by something as common as sunburn, is a known risk factor in developing melanoma. Researchers know that prolonged or extreme exposure to radiation can trigger protein interactions that change the structure and function of skin cells. If they can zero in on the genes involved in these interactions, they will be a step closer to developing more effective strategies for prevention and treatment of melanoma.

“The sophisticated computational tools being generated at RENCI will provide us with the ability to rapidly monitor the system of response to DNA damage, which is known to suppress the development of cancer and is often a target of chemotherapy,” said Kaufman. “By comparing the system in normal and malignant melanocytes, we will be able to identify the molecular changes in cells that underlie development of the disease and that make melanomas resistant to standard chemotherapy.”

With funding from the National Center for Supercomputing Applications (NCSA) at the University of Illinois, the RENCI group is developing software that will allow the researchers to merge data obtained from their own research subjects and from databases distributed over the Internet into one unified format, a process called data federation. Guan also is incorporating Cytoscape, an open source visualization tool, into the new set of research tools. Cytoscape is a collaborative project of the Institute for Systems Biology, the University of California at San Diego, Memorial Sloan-Kettering Cancer Center and the Institut Pasteur.

Once the RENCI team develops a way to merge and visualize the data, the researchers will be able to take advantage of a suite of data mining tools developed at NCSA called Data to Knowledge (D2K). D2K can perform a wide range of data analysis functions, from sifting through the “noise” in large datasets to find meaningful patterns to showing correlations to developing predictive models. According to Guan, the RENCI software developed for Kaufmann’s group will be the prototype for a toolkit that in time will be made available to research teams worldwide to help them integrate and visualize data.

“This project is a good example of what RENCI is all about,” said Dr. Daniel A. Reed, director of the institute. “We bring together people who can benefit from cross-disciplinary collaborations. In this case at RENCI, the Carolina medical school and NCSA to solve problems that couldn’t be resolved by one group working alone. The end result, we hope, is real progress in the effort to understand and treat a serious disease.”