SciLink Spotlight - Tom Plasterer: How a system’s approach will aid clinicians and pharma in getting to a diagnosis and treatment faster
Comments(0) Tom, Can you tell us a little about your background? how did you get into science and interested in a system’s approach?
As an undergrad at the University of Wisconsin I pursued dual interests in biology and English literature. I came across Lynn Margulis and Dorian Sagan’s Microcosmos, a wonderful treatise on the role of symbiosis in eukaryotic evolution, and wanted to pursue a few ideas on the putative bacterial origin of the structural protein tubulin. Fortunately, I was at Wisconsin during the late 1980’s, which was a great place to be during the seminal years of bioinformatics. Two pioneering sequence analysis software companies, the Genetics Computer Group (GCG, currently Accelrys) and DNASTAR, sprang out of efforts to sequence Escherichia coli at Wisconsin; GCG by John Devereaux and DNASTAR by Fred Blattner. With this backdrop I began to pursue my nascent ideas in sequence analysis and symbiosis.
I completed an undergraduate thesis on the molecular evolution of tubulin under the guidance of Dr. Gary Borisy (an expert in tubulin and cell division) and Dr. Ann Palmenberg (an expert in RNA picornaviruses). Ann taught and organized a seminar series in Bioinformatics, which was likely among the first classes in the country. She worked with John Devereaux to bring in many of the heavyweights in Bioinformatics; both for the seminar series and to collaborate with GCG. I took the class three times and was taught early sequence analysis theory by Michael Zucker (RNA folding), Steven Altschul (BLAST), Temple Smith (Smith-Waterman alignment), Walter Fitch (molecular phylogeny), James Crow (neutral theory of evolution) and many others. The undergraduate thesis did not lend itself to any spectacular results, presumably due to a combination of my inexperience, an exceedingly small database of protein sequences (Genpept was around 40,000 sequences in 1990) and the lack of a structure-based phylogenetic approach. Later, Linda Amos and Jan Löwe, would show that there was indeed a strong structural similarity between tubulin and bacterial ftsZ structural proteins, although symbiosis may not needed to explain this event as later transfer is just as likely.
My undergraduate work did, however, give me a taste of bioinformatics and sequence analysis which would lead me to Boston University and BG Medicine. After graduating Wisconsin, I took a role at DNASTAR in sales and technical support and shortly transitioned into technical writing. After a time I grew tired of describing bioinformatics tools and wanted to develop and apply them myself. With Fred Blattner’s (head of the E.coli genome project and DNASTAR’s founder/CEO) encouragement and support I joined Temple Smith’s group—the BioMolecular Engineering Research Center (BMERC)—in the fall of 1996.
Tell us a little about your experience in graduate school and your interest in bioinformatics
I was again extremely fortunate to be at the BMERC during the late 1990s. The lab had a number of research interests, anywhere from sequence analysis (dynamic programming, Bayesian prior-profiles) to protein structure prediction (homology modeling, threading) to molecular phylogeny. The Institute for Genomic Research (TIGR) was beginning to rapidly complete multiple bacterial genomes and eukaryotic genomes were in sight. This afforded a tremendous opportunity in comparative sequence analysis. Under Temple’s leadership and an extremely bright group of post-docs, students and collaborators we were able to accomplish a lot of guerilla science (get in, make your mark, get out before the big-boys arrive), including a technical comment in Science on Genome Excess plots in my second year in the lab.
The environment at BMERC was quite different than many graduate experiences. Temple encouraged debate at all levels and collaboration within and outside of the group. He can still beat most of his students with a gigantic library of existing codes (“I’ve already done that in Fortran…”). His piped combinations of Sed, Awk, Sort and Comm are still semi-legendary.
I finished my dissertation work applying prior-profile analysis to proteins involved in mitochondrial pathologies. This work highlighted the great degree of sequence and structural conservation among the mitochondrial proteome and the consequences for modifying key residues in critical locations.
After BMERC you joined Beyond Genomics now BG-Medicine what do you do there?
After graduating I joined Beyond Genomics, a Waltham, Massachusetts based systems biology start-up. In those days BG was more of a technology shop interested in developing mass spectroscopy approaches to measuring systems, primarily in plasma. We also did a lot of work developing the computational, statistical and bioinformatics architecture necessary to support such efforts. Eric Neumann (Clinical Semantics Group), Matej Oresic (VTT, Finland) and I created correlation networks as a way to take advantage of mathematical associations in cross-omics data. The key advantage of such an approach is the ability to annotate and understand poorly-characterized analytes as well as well-characterized analyte within the context of a single experiment. This approach became extremely useful in evaluating novel mechanisms of disease development and drug interaction, biomarker discovery and circulating biomarkers of tissue effect.
I took over leadership of the group from 2003-2006 and have since moved on to direct the project planning & data interpretation group at BG Medicine (Beyond Genomics changed its name to BG Medicine in the fall of 2004). In this role I have greater responsibilities for experimental design as well as the back end of bioinformatics and general biocontextualization (a fancy systems biology term for attempting to place findings into an appropriate biology context). This included primary scientist roles in our Liver Toxicity Biomarker Study (LTBS) and our High-Risk Plaque consortium (HRP).
What you believe is the definition of system’s biology and how a system’s approach will aid clinicians and pharma in getting to a diagnosis and treatment?
Hiroaki Kitano, Lee Hood and Doug Lauffenburger still have the best definitions for Systems Biology. From Kitano’s review in Science in 2002 he described systems biology in terms of four properties: system structures, system dynamics; the control method and the design method. This view is fairly complementary to Lauffenburgers’s 4M model: measure, mine, model, manipulate. The problem is that, with the exception of very narrowly defined systems, we’re still in the first two stages: measuring all of the omics data you can and trying to determine the relationship among the measured analytes. Having a good understanding of network control and manipulating networks is still a ways, off, at least for most of the problems that clinicians and the pharmaceutical industry is interested in. A few companies (Entelos, Genstruct) are focused on this problem but most of this work is still carried by academics. Some groups are combining systems biology alongside synthetic biology, for example Jim Collins’ and Tim Gardner’s work on mammalian switches and network inference, which look particularly promising.
BG Medicine has taken a more narrow approach in the systems biology space focusing on ‘Systems Pathology’ and ‘Systems Pharmacology’. Systems Pathology is loosely defined as the measurement and interpretation of molecular analytes that change in the disease state. Systems Pharmacology, then, is the measurement and interpretation of molecular analytes that change with a pharmacological perturbation. Under this model measuring system components and their interactions are the keys to elucidating a system under disease burden and drug treatment.
In terms of the ultimate utility of system approaches for medicine, I’d stress the importance of well-conceived experimental designs over sophisticated pathway, network and bioinformatics approaches—yes this is coming from a card-carrying bioinformaticist. So many of the omics studies today are underpowered, making results non-generalizable to larger contexts. Good experiments are costly, however, so this is why small sample sizes are the norm. We spend a lot of time reviewing project objectives prior to carrying out any experiments: is the study sufficiently powered to see univariate markers? Multivariate classifiers? Correlations among analytes? Are we recording the system with the right set of experimental platforms to measure perturbations in disease and drug intervention? Even addressing these concerns there are still a lot of open questions that can interfere with interpretation. Are you measuring at the right time-scale? In the right location? Do you need to fractionate tissues or cells? Does a cell-system accurately recapitulate its environment? Does an animal model accurately recapitulate clinical behavior? I could go on…
The first return on investment using a systems approach will probably occur at the intersection of systems biology with biomarker discovery, likely on the diagnostic front. Biomarkers derived from omics experiments are already in use for disease classification, such as Oncotype Dx® for breast cancer. Therapeutic response is another area where biomarkers can play a role. Predicting responders/non-responders when the disease and drug treatment is known is a much less parameterized space than disease classification and therefore a less daunting target. The holy grail for biomarker utility is an analyte, or small set of analytes, that can be used for diagnosis, prognosis and therapeutic monitoring, for example; this marker would be elevated in diseased individuals over healthy, higher levels indicate a poorer prognosis and drug therapy would decrease biomarker levels over time while outcome improves. Biomarkers of this type are exceedingly rare and to my knowledge, only Galectin-3, a prognostic marker of hear failure, comes close.
Who should reach out to you in SciLink?
I have two primary interests, one in network biology and pathway analysis and a second in biomarker-guided medicine (the ‘BG’ in BG Medicine). I
Some of my thoughts in network biology are better suited for an academic environment, and I hope to develop these further at Northeastern (these concepts were recently published in Drug Efficacy, Safety, and Biologics Discovery: Emerging Technologies and Tools: “Systems Biology, Biomarkers, and Biomolecular Networks”). I’m always interested in bouncing these ideas off like-minded colleagues.
Biomarker adoption throughout the entire therapeutic process is slowing becoming a standard approach. There is a chain of biomarker utility from disease prediction, disease diagnosis, disease prognosis, disease classification, therapeutic selection, therapeutic predictors of response/nonresponse and well as surrogate biomarkers. I’m also interested in discussions around the use of biomarker-guided therapies at both the scientific and business level.
