1. What are the key areas where we need strength to support strategic biological science and clinical research goals?
2. Which of those are areas that Vanderbilt can be #1 from an informatics, computation and statistics perspective?
3. Can the areas where we need strength, but can not be #1, be provided as service functions and, if so, by faculty or staff?
4. Where should various people be situated, and what kind of recruitment teams and packages do we need?

1. Research in the basic information-centric science of informatics, mathematics, statistics, etc. Faculty spend 100% effort in basic research in their information-centric discipline.
2. Research in application of techniques from information-centric disciplines to biological, translational, and health services research domains. Faculty spend 80% effort in basic research in their information-centric discipline, but all of that time is devoted to techniques needed for their research domain applications, and the other 20% of their time is spent in the research domain where the techniques are being applied.
3. Application of proven techniques from information-centric disciplines to biological, translational, and health services research domains. Faculty spend 80% effort in application of information-centric techniques to their research domain applications, and 20% of their time in the information-centric disciplines where the techniques are being developed.
4. Biological, translational and health services research. Faculty spend 100% effort in basic research in their domain.
5. Infrastructure and support services to support #2, #3 and #4.

1. DNA microarray technology: Provides ability to measure gene expression for on the order of 10,000 genes via RNA products. RNA from two different specimens marked using two different fluorochromes, and matched against pre-arranged array of 10k genes (DNA). By observing which RNA samples hybridize against each specific gene, one can determine activity (and differences between specimens) at gene-by-gene level. The technology is useful to measure gene activity in multiple contexts: patterns of which genes up- and down- regulated as the result of a mutation, or as the result of drug administration; and, patterns of gene expression over time, for example, in embryogenesis, or various forms of biological differentiation. Technique to recognize patterns of expression efficiently in populations of 10K individuals with 10K gene determinations are not yet available, and a good area for bioinformatics research. Machine learning, neural nets, Bayesian nets may be applicable. Stanford is a leading center in microarray technologies, could learn/recruit from them. In addition, Oak Ridge also active in this area.

2. DNA sequencing: standard bioinformatics methods and tools used by leading centers in human genome project only used sparingly here. Potential VU participation in "Second Round" genome projects, involving mouse and human genomes, will lead to substantial primary sequencing work here at Vanderbilt. New NIH RFA for second-generation sequencing centers expected in near future. Technology to do this sort of informatics work exists elsewhere (Baylor is closest collaborator), but ability to scale up to handle reliably large volume of base pair sequences is not in place at VUcurrently.

3. RNA post-transcriptional editing: RNA sequences edited by enzymes post-transcription so that RNA sequence that generates amino acid sequence does not match DNA sequence. Enzyme specificity based on secondary, 3-D structure of RNA, not on its sequence. Algorithms capable of predicting/recognizing secondary, 3-D RNA structure are essential.

4. State-wide mouse genetics consortium (VU, Oak Ridge, UT Memphis, UT Knoxville, St. Judes Hospital, others) to bid on mouse mutagenesis project for NIH funding (late 4/99 submission deadline). Mice will be exposed to mutagens, genotyped, bred randomly, and observed for physical and behavioral differences. Goal is to identify genes controlling various physiological features, such as blood pressure, liver size, etc. After specific mutations are induced, one can study gene expression in populations with same mutations using DNA microarray technology, to determine, for example, which sets of genes are up- or down- regaulated synchronously as a result of a mutation, or as the result of administering a drug. Also would want to correlate measures of gene expression at phenotypic (e.g., physiological) level, with results of microarray gene analysis. Over 10K mouse lines with 100K data points each to be analyzed, correlated. Oak Ridge database skills will be relied upon, but local expertise and effort in bioinformatics also required. Similarly, consortium may bid on NIH proposal to sequence mouse genome.

5. Proteomics: The proteome is the expressed proteins resulting from the genome. "Proteomics is functional genomics at the protein level. Proteomics can be divided into expression proteomics, the study of global changes in protein expression, and cell-map proteomics, the systematic study of protein-protein interactions through the isolation of protein complexes." [Blackstock WP, Weir MP Trends Biotechnol 1999 Mar;17(3):121-7]. Proteomics studies involve relating patterns of protein fragments from mass spectrometry to the genes that encode them.

6. Genetic epidemiology: Genetic studies in large populations (100's to 10,000's) of patients and their families. Detailed disease, demographic, and risk factor information must be connected to large amounts (10,000's) of genetic data points. Intricate relationships between individuals (family trees) must be stored. Traditional methods of analysis of these data, particularly statistical analyses, do not scale well to this large volume of data. New methods of pattern recognition and data visualization are needed.

7. Genetic medicine: Applying risk information from genetic tests in current medical practice. How to generate and store this genetic information and relate it to an individual's clinical record is an unsolved problem. Ethical issues in confidentiality of genetic information and its impact on relatives of patients is a difficult and unsolved area. The role of MARS in addressing this problem should be evaluated.

8. Pharmacogenetics: The role of a genetic variation in drug efficacy. Most drugs are variably effective with variable side-effects; in many cases the variability will be genetically controlled. Large databases of patients, their genotypes, and their clinical outcomes will need to be linked to pharmacokinetic results. Both clinical and research applications must be considered.

1. 1. Mass spectrometry: molecular weight of fragments from large biomolecules (e.g., proteins) can be used to predict/reconstruct overall structure, and be related to genetic or other data. Over 300K data points can be generated per half-second on each of 2K-3K instances. Problems include efficient capture, labeling/indexing, transfer, storage, and retrieval of large data sets; linkage to national and international data sets; and, complex pattern recognition within data sets.

2. Xray crystallography: more established technology, labs self-contained, self-sufficient for the most part

3. High field NMR - computationally intensive. Of note, Walter Chazin is being recruited from Scripps Institute to head new interdisciplinary Structural Biology program; he intends to recruit a bioinformaticist if he comes. Space available on 5th floor MRB III to house this new program.

4. Predicting 3-D structures of DNA/RNA/proteins from linear sequences of base pairs or amino acids: unsolved, difficult computational problem.

Cancer biology intersects with many of the other areas above. Ability to take tumor cells and through analysis of DNA/RNA/protein develop complex "molecular fingerprinting" to recognize tumor (genetic susceptibility & early detection); predict tumor behavior (e.g., rapidity of spread); and, guide therapy (susceptibility of tumor to various therapies). There are important databases issues related to tracking individual patients over time based on molecular information, and related to cross-referencing massive amounts of local data with national/international databases.

1. Ability to utilize shared resources at National level (e.g., NCBI, others) is variable across campus, some groups doing better than others. Having core staff/facilities to help get faculty up to speed with these programs, and to coordinate locale efforts, might be helpful. Many large and important databases are now proprietary (e.g., pharmaceutical companies), and mechanisms to establish access for research are a problem.

2. Issues of interconnectivity locally and state-wide are problematic: how to enable other labs to use outputs of a given lab here.

3. Ability to share computing infrastructure to realize economies of scale a mixed proposition. Some labs/researchers are independent and must have integral computing to accomplish work (e.g., structural biology - xray diffraction crystallography); others may benefit from shared centralized resources. More discussion is required to determine actual benefit, if any, for VU research community.

4. Ability to do mapping of genes using local and national databases: for a given sequence, cross-referencing it to human and mouse genomes (among others) can be done in part via BLAST and similar methods, but expertise here is variable.

5. MARS might help in storing/retrieving mouse genome, mass spec, and other large data sets; it is especially relevant if data are patient-specific and patient has other clinical records in MARS.

• Clinical epidemiology
• Clinical economics
• Management sciences
• Decision sciences
• Medical informatics
• Clinical measurement, such as psychometry
• Medical Genetics
• Health Policy
• Technology Assessment
• Clinical ethics

Doctoral (PhD) Faculty with training and research foci in the following disciplines:

• Epidemiology
• Biostatistics
• Health economics
• Health services organization/management
• Medical sociology/measurement
• Evaluation methodology
• Psychology/measurement
• Medical Informatics
• Health policy
• Ethics

Vanderbilt University currently is recruiting and training clinicians with the relevant training to serve as clinical investigators within the Health Services Research Center. Current resources are available to continue this recruitment within the Department of Medicine. Other departments should consider recruiting faculty with training and career interests in health services research.

Vanderbilt University currently has senior faculty with expertise in epidemiology but additional faculty would be very helpful. Biostatistics faculty have too little current capacity to support the grant submissions which will be generated by the Center, but the medical center has already responded to this concern by providing resources for the recruitment of two new faculty. This support is to be acknowledged and applauded as a very important component to the success of the Health Services Research Center. The university does not currently have a health economist. The medical center, VIPPS, and the Dean of Arts and Sciences support the recruitment of a health economist but with the important of cost components to many studies, the recruitment of two such individuals would be helpful. The university does not currently have a medical sociologist with a focus on health services research. Such an individual would be of enormous benefit to many studies with regard to measurement including the design, implementation, and evaluation of new measurement tools for clinical medicine. The university should identify the resources and position for the immediate recruitment of a medical sociologist with experience in health services research. Faculty with knowledge of evaluation and measurement exist at Vanderbilt University, especially through the School of Education and through VIPPS. These faculty need to be brought into the research programs of the Health Services Research Center to serve as collaborative faculty. Specific new faculty recruitment may be needed in this area but only after the Center identifies gaps in current areas of expertise. The Health Services Research Center will provide an important opportunity for collaboration between the medical center and the Owen School. The need for additional faculty in the management or decision sciences is unknown at this time but will be discovered as the Center's activities evolve.

Interviews with NCBI researchers: Dennis Benson, Ph.D. (4/19/99), and Mark Boguski, Ph.D. (4/22/99)

The National Center for Biotechnology Information (NCBI), headed by Mark Boguski, is a NIH unit housed within the National Library of Medicine. It currently employs approximately 120-125 people, including 30 Senior Scientists in Computational Biology. Its appointments follow traditional the traditional NIH system/structure. Dennis Benson is a Division Chief at NCBI.

Dennis Benson's comments:

In developing a new Bioinformatics unit, an ongoing, close linkage to the user community is the key. Changes in basic biology and genetics occur too fast to allow developers to carefully "spec out" and develop computer applications. The key thing is to keep in touch with the researcher-users and to be flexibly and rapidly adaptive to their evolving needs.

Dr. Benson said that from experience, it is far easier to take individuals well-grounded in Biology (e.g., Ph.D. or greater level ofexperience) and to teach them how to develop computer programs, rather than to teach computer scientists how to function in the application area of Biology. The leaders of bioinformatics efforts should be biologists with strong computer backgrounds, who have programmers working under them. At NLM, and at Wash U St Louis, librarians have become involved and played key roles in bioinformatics support.

Dr. Benson recommends using generic, general purpose hardware (e.g., Sun Solaris or 4-CPU Intel Unix boxes running C/C++/Perl) rather than some of the more exotic, expensive dedicated systems that have appeared from some vendors. While many universities refer their BLAST and related searches to NLM, some have chosen to mount such capabilities locally in order to give better response time. GenBank currently 15 GB in size; all data at NCBI approximately 1.5 to 2 terabytes in size. In discussions about the VUMC computing environment, Dr. Benson noted that MARS might be able to play an important role in supporting bioinformatics research, although the exact nature of how to best leverage MARS would have to be worked out with the researcher-users.

The level of expertise required to function in the field has been steadily increasing. For example, the "help desk" at NCBI now has 25 of 27 employees with Ph.D. level training or beyond. These individuals help users (researchers) to interpret their results and provide insight greater than the mechanics of using software programs in a research context.

Role of commercial companies increasing; can now hire (at significant expense) a number of leading companies to help solve a University's biotechnology informatics problems. Also note that many important biotechnology databases are proprietary (e.g., owned by pharmaceutical companies), and collaborative or other ties may be needed to gain access.

NCBI has a "Visiting Program" for collaborating institutions (Oak Ridge is one such institution). The referring institution must be actively involved in collaborating with specific NCBI personnel on active, ongoing research, and can send 1-2 individuals to work at NCBI for periods of 1 week or longer. This is NOT a training program; the visiting scientists must have already contributed to the ongoing collaborative research from their home institution, the "visit" helps them to further the collaborative relationship with NCBI personnel.

Institutions/programs producing graduates of potential interest: (1) University of Pennsylvania (contact: Chris Overton); (2) Wash U St Louis (David States, Mark Frisse, others); (3) Montgomery County (MD) Community College - Masters in Biotechnology program; other programs with good reputations for PhDs include: Stanford, Baylor, MIT/Whitehead.

It is useful to talk to NIH/NLM program officers and to examine the NIH databases (CRISP) to review what bioinformatics activities are being funded in an ongoing manner.

Dr. Benson sees techniques to extract biotechnology information from the literature (and to correlate it with basic science research data) as an emerging important area for research. OMIM is an example of such an effort. He notes that the NIH Director, Harold Varmus, has recently made statements favoring electronic publishing of journals using existing editorial boards and bypassing the expensive middle-man costs imposed by publishers. In such an environment, knowledge aggregation methods will be important.

Mark Boguski's comments:

In developing a bioinformatics unit, it is important to have both strong research and service components; these should be somewhat separate. Researchers must be given 80% protected time to avoid "bottomless pit" of service obligations. In developing a faculty, one of the best methods is to groom younger individuals locally; biologists preferred to computer scientists; ideally would start with undergraduates who have dual majors in Biology and CS. Attempting to transform fully trained computer scientists more difficult (they tend to be more theoretical than applied); possibly better to transform experimental physicists and biomedical engineers, who are both applied and more likely to already have experience with computers and large datasets.

Some of the service component support can be purchased "off the shelf":

Useful contacts suggested; others concurrently developing Bioinformatics programs:

1. DNA/Protein sequence analysis/automated DNA sequence assembly for Mac/Wintel desktop computers: DNA Star (Emerson lab); MacVector - Oxford Molecular Inc. (Sutcliffe, Threadgill labs); Sequencher (Gene Codes/ABI Autoassembler (ABI)

2. GCG program - Web-based upgrade - mounted by ACIS on Sun Ultras; VU has had outmoded version of GCG program Charles Alexander recruited from Rochester because he has experience in using GCG as a resource there; program will be treated as Core Resource, Mark Magnuson coordinating this effort.

3. PEDIGENE: Glaxo-sponsored set of programs for genotype, DNA, and clinical data to be installed on Unix workstations in 519 Light Hall.

4. RasMol - used by Xray crystallographers -RasMol is free software for looking at molecular structures. It runs on Windows or Macintosh/PPCcomputers (also unix). You must download a free PDB data file for each molecule you wish to view. It is very fast: rotating a protein or DNA molecule shows its 3D structure. Chime shows molecules like RasMol, but unlike RasMol, Chime shows molecules inside a web page. Chime shows only the molecules written into the web page by its author. Chime is free, and runs on Windows or Macintosh/PPC computers (also Silicon Graphics). Once you have installed Chime, you can view DNA, protein secondary structure, hemoglobin, antibody, etc. directly on web pages.

5. Intelligenetics

6. Linkage analysis software - Genehunter - Haines lab

7. Current VUMC Web site for molecular biology stated to be out-of-date

1. Oak Ridge: Nationally prominent bioinformatics group; mouse genome expertise
2. Baylor: one of most successful first-generation human genome sites
3. NCBI

1. Wash U St. Louis
2. Stanford
3. MIT/Whitehead
4. NCBI
5. Baylor
6. Penn
7. Oak Ridge

Copyright © 1997, Vanderbilt University Medical Center
URL: http://www.mc.Vanderbilt.Edu/infocntr/recruitment.html
For More Information: <Kimberly.Lawrence@mcmail.Vanderbilt.Edu>
Last Modified: February 12, 2001
<webmaster@www.mc.Vanderbilt.Edu>