Experimental / Data Technology

Transcription

1 Experimental / Data Technology Fumiaki Katagiri Univ. of Minnesota Dept. of Plant Biology Center for Microbial and Plant Genomics Adam Arkin Howard Hughes Medical Institue Department of Bioengineering Univ. of California, Berkeley WTEC Systems Biology Final Workshop March 11, 2005 Experimental (wet) and Theoretical (dry) work Experimental data for: Network Inference Parameters in a model Validation of models etc. Kitano H. (2002) Science 295,

2 Different stages in systems biology research When the network structure has not been well established: large amounts of data in many categories help inference of the network. When good network models are available: models point out specific, important measurements. More precise, directed, lower-throughput measurements may be more desirable Important criteria for experimental data in the absence of modeling constraints Exhaustive in each category (not missing anything important) Correlated data in many different categories (as many as possible) Sufficient resolution in space and time Quantitation with sufficient accuracy In combination with high-throughput genetic and other perturbations Implementation: Large high-throughput experimentation centers? 2

3 Examples General availability of OMICS technologies - Network inference using Bayesian approaches based on OMICS data (Koller, Stanford) Systematic transcriptome and metabolome analyses of E. coli KO mutants (Tomita and Mori, IAB) Protein chip-based HT kinase activity assay (JBIRC) Important issues for experimentation when guided by models What needs to be measured? (e.g., to discriminate among possible models) What is the most informative experimental designs for measurement? Does it require people with different expertise? (New device development?) Is there sufficient precision and resolution? Correlations with other data are important. Implementation: Ad hoc experimentation teams? 3

4 Examples Model-guided experimental design and model validation Cell cycle, bacteria chemotaxis studies directed genetics, imaging, protein activity measurement Circadian rhythm (Doyle, UCSB; Ueda, RIKEN CDB) Jak-Stat pathway, protein phosphorylation measurement (Klingmüller, Max-Planck- Institut für Immunbiologie) Technical challenges in specialized measurements Measuring methods in situ Quantitative abundance and activities of molecules, with the exception of Ca 2+ imaging dyes and perhaps fluorescence-tagged proteins and FRET Better controlled perturbation methods Caged molecules, etc. 4

5 Experimental technology in the US Broad availability of technologies for mrna profiling and proteomics Limited availability of metabolic profiling technology. Metabolite identification is slow. Methods for more precise measurements difficult and slow and limited to real experts (FRET, quantitative Westerns, RTPCR) Various phenomics (detailed biological phenotyping) approaches Some systems biology centers (ISB, PNNL) Advanced research in industry as well as academia Experimental Technology in Europe Profiling technologies appear to be more centralized. Large, focused operations at some Max-Planck- Institutes and Free University. Geographically spread consortia, such as EU projects and the Hepatocyte project, which need to emphasize standardization. Small programs, such as one at U. of Warwick, which emphasize integration of wet and dry research at the individual level. 5

6 Experimental Technology in Japan Strongly oriented toward high-throughput discovery research. Systems biology approaches (which involve modeling) are not prevalent. General availability of mrna profiling technologies is high. Generally, large new institutes we visited are equipped very well and emphasize development of new measurement technologies. Metabolite profiling technology is advanced at Institute for Advanced Biosciences and Kazusa DNA Research Institute. U Tokyo, LSBM and RIKEN Yokohama Institute is very strong in generation of materials and information in highthroughput research. Many government-led academia-industry collaborations. Important but not covered in our study Microfluidics and other micro- (or nano-) manufacturing technologies, which could vastly improve measuring technologies. Imaging technologies, especially in realtime at the single cell or single molecule level, to track single events. 6

7 Important database issues Large inclusive databases vs. small specialty databases. Automatic vs. manual curation Standardization vs. competition Maintenance/update: effectiveness, continuation of funding Sharing and (virtual) consolidation A Dawn of Real Large-Scale Sciences in Biology? One possible scenario 7

8 Correlated data in different categories Why are correlated data important? Data are context dependent. If we do not have the same context in data collection in multiple categories, we cannot efficiently integrate the data from other categories. Tracking of biological context will be VERY important Broad profiling methods provide highly detailed (but less precise) descriptions. If data are not highly correlated, a lot of information about the details is wasted. Correlated data in different categories Ideal Do all the measurements with the same samples. Automate all the procedures. mrna profiling, protein profiling, metabolite profiling, etc. 2 nd best Do all the experiments using the same facilities. Have them performed by the same people. A small number of large experimentation centers are favored. Typical compromise Do the experiments at multiple sites with multiple people, but with strict standards. 8

9 Cost issues Development, maintenance, and running costs of new state-of-the-art measuring devices are expensive. To run such devices cost-effectively, it is important to run them close to capacity all the time. Storage and dissemination of the data in standard form is also more costly than commonly held. Examples of large centers ISB, PNNL (USA) Max-Planck-Institutes (Germany) LSBM, RIKEN Yokohama Institute (Japan) 9

10 Ideal for an experimentation center A highly automated and controlled facility to generate consistent samples All sorts of state-of-the-art but established profiling technologies that are highly automated Spaces for ad hoc experimentation teams A highly versatile engineering team that supports ad hoc experimentation teams Can provide measurement services for external samples Supporting staff for all the activity Proximity of experimental and theoretical teams Close collaboration, not division of labor, between experimentalists and theoreticians. better understand what the other party is doing. Theoreticians, engineers, physical scientists and biologists should also be at the centers. 10

11 Impacts on data technology With a small number of major centers: It is easier to set standards for databases, software tools, etc. Standardization easier use, easier training It is easier to avoid unnecessary duplication of effort. Each center can host the databases for particular biological systems. This eases maintenance of the databases. Doesn t it sound more like large experimentation centers in high-energy physics? Not that large and not that few, though 11

12 Some social issues associated with large-scale sciences Change in the status of individual researchers: much more team-oriented. Evaluation-reward mechanisms, including publication, funding, and promotion Strong management skills in leaders Governing structure in the research community Project prioritization, etc. Sharing data in the research community Inter-program, inter-agency cooperation for funding International cooperation Influence of politics Although this is one extreme view, We are likely to see more or less of this trend. Many factors interact in a complex manner to determine the behavior of the research community system. 12

13 Summary The US and Japan are leading in experimental technologies in general. However, the mainstream research in Japan is not oriented toward systems biology. In the area of model-guided experimentation, the US and Europe have several success stories. Europe s large, geographically spread consortia approach does not seem very effective from the experimental viewpoint, although it raised the sense of importance of standardization and unity. Overall, the US is taking a leading role in the experimental area of systems biology at the global level. Importance of correlated data in many categories could push experimentation in systems biology toward establishment of large experimentation centers. This trend will bring new social issues within the research community. 13