Progress 09/15/07 to 09/14/09
Outputs
OUTPUTS: Muir and Howard (1999, 2001, 2002) developed a method to estimate environmental risk of genetically modified (GM) organisms by predicting the fate of transgenes introduced into wild populations by escaped GM organisms. The objectives of this research project were to: 1) Expand the Muir and Howard model to incorporate stochasticity and uncertainty and develop a user-friendly computer program that incorporates these additions. 2) Estimate fitness components using transgenic zebrafish as a model organism. 3) Test model predictions using replicated populations of wild-type and GM zebrafish and examine changes in fitness component values through time. During the current (and final) reporting year, the data relative to the experiments for estimation of the fitness components were analyzed, including a bootstrap approach for assessing uncertainly on the parameter estimates. Such fitness component estimates (and the variability associated to them) were then used to feed the risk assessment model developed on this project to generate model predictions regarding transgene fate. The results were finally compared with experimental results obtained from replicated populations of wild-type and GM zebrafish to test the model, as indicated in Objective 3 of the project. PARTICIPANTS: Dr. Guilherme J. M. Rosa: Dr. Rosa was the PI of the project. He worked directly on the development of the methodology related to the risk assessment, and supervised the whole project. Dr. Ashok Ragavendran: Mr. Ragavendran was a PhD student at Michigan State University, and he has helped Dr. Rosa on the development of the methodology related to the risk assessment, and developed the C++ software code for implementing the methods. Dr. William Muir: Dr. Muir was a Co-PI for the project, and he was responsible (together with Dr. Howard) for the development of all experiments. He also participated actively on the development of the methodology and software related to the project. Dr. Richard Howard: Dr. Howard was a Co-PI for the project and, together with Dr. Muir, was responsible for the development of all experiments. During the development of the project on the reporting year, a graduate student (Mr. Ashok Ragavendran) had the opportunity to be trained in risk assessment, modeling, and computational statistics. In addition, two undergraduate students (Ms. Kelly Connell and Mr. Srdan Gajic), Biology majors at the University of Wisconsin - Madison, were trained in basic concepts of population genetics, specifically on forces that change gene frequencies. TARGET AUDIENCES: So far results and outcomes have been communicated by presentations at scientific meeting and publications in scientific journals. Hence, the target audience at this stage refers to academic researchers and regulators working on biotechnology and GMO matters. PROJECT MODIFICATIONS: Nothing significant to report during this reporting period.
Impacts
During the reporting period, additional sources of variability on model predictions were studied, such as the uncertainty on estimates of fitness components used as input on model implementations. Fitness components parameters were estimated from experimental data using genetically modified Zebrafish using different bootstrap approaches, and the results indicated that the estimates for the wildtype and transgenic genotypes presented considerable variation, both in absolute and relative values, leading to extra variation in model predictions. Results also showed that even moderate variation in estimates of individual components can generate large effects at the population level due to non-linearity and the interactions among genotypes. The model predictions showed good agreement with results observed in the mesocosm experiments, indicating that realistic prediction in the case of GEO risk assessment can be made using carefully constructed experiments and modeling approaches. Also within this context, we developed a Copula methodology to incorporate dependencies among fitness components due to life-history trade-offs, to assess their effects on model predictions. Results showed that assuming independence among fitness components can produce overly conservative predictions. Further, correlations among absolute fitness components differed from that among the relative values of fitness components, which are the final determinants of transgene fitness. As an additional study we developed a global sensitivity analysis to examine the effect of parameters on model predictions. Meta-modeling using a Bayesian Gaussian Process (BGP) was employed to improve the efficiency of the sensitivity analysis and thus reduce the overall computational burden without sacrificing model complexity. The predictions from the BGP model was shown to provide satisfactory performance as an emulator. The choice of a Bayesian approach is deliberate as this framework is flexible to incorporate model extensions as well combine outputs across multiple models and this choice will facilitate integration with other areas in future.
Publications
- Muir, W. M., Rosa, G. J. M., Pittendrigh, B. R., Xu, S., Rider, S. D., Fountain, M. and Ogas, J. 2008. A mixture model approach for the analysis of small exploratory microarray experiments. Journal of Computational Statistics and Data Analysis 53: 1566-1576.
- Ragavendran, A. Improving Risk Assessment of Transgene Invasion: Assessing the Role of Uncertainty in Predictions. PhD Thesis, Michigan State University, East Lansing. MI, August 2009.
- Van Eenennaam, A. and Muir, W. M. 2009. Animal Biotechnologies and Agricultural Sustainability. In: The Role of Biotechnology in a Sustainable Food Supply. Eds (in press)
|
Progress 09/15/07 to 09/14/08
Outputs
OUTPUTS: Muir and Howard (1999, 2001, 2002) developed a method to estimate environmental risk of genetically modified (GM) organisms by predicting the fate of transgenes introduced into wild populations by escaped GM organisms. The objectives of this research project are to: 1) Expand the Muir and Howard model to incorporate stochasticity and uncertainty and develop a user-friendly computer program that incorporates these additions. 2) Estimate fitness components using transgenic zebrafish as a model organism. 3) Test model predictions using replicated populations of wild-type and GM zebrafish and examine changes in fitness component values through time. During the current reporting year, all experiments necessary for estimation of fitness components related to objective (2) were finalized. In addition, all the methodology for incorporating stochasticicty in the fitness model was developed, and a software code (written in C++) to implement the methodology was generated and extensively tested. We are now implementing another extension of the model, to incorporate uncertainty regarding fitness parameters estimates into the model predictions of risk. PARTICIPANTS: Dr. Guilherme J. M. Rosa: Dr. Rosa is the PI of the project. He worked directly on the development of the methodology related to the risk assessment, and supervised the whole project. Mr. Ashok Ragavendran: Mr. Ragavendran is a PhD student at Michigan State University, and he has helped Dr. Rosa on the development of the methodology related to the risk assessment, and developed the C++ software code for implementing the methods. Dr. William Muir: Dr. Muir is a Co-PI for the project, and he was responsible (together with Dr. Howard) for the development of all experiments. He also participated actively on the development of the methodology and software related to the project. Dr. Richard Howard: Dr. Howard is a Co-PI for the project and, together with Dr. Muir, was responsible for the development of all experiments. During the development of the project on the reporting year, a graduate student (Mr. Ashok Ragavendran) had the opportunity to be trained in risk assessment, modeling, and computational statistics. In addition, two undergraduate students (Ms. Kelly Connell and Mr. Srdan Gajic), Biology majors at the University of Wisconsin - Madison, were trained in basic concepts of population genetics, specifically on forces that change gene frequencies. TARGET AUDIENCES: At this stage of the project development, results and outcomes have been communicated exclusively by presentations at scientific meeting and publications in scientific journals. Hence, the target audience at this stage refers to academic researchers and regulators working on biotechnology and GMO matters. PROJECT MODIFICATIONS: Nothing significant to report during this reporting period.
Impacts
During the reporting period, a comprehensive Monte Carlo study developed to assess the consequences of ignoring or accounting for demographic stochasticity into model predictions showed that considerable variability in prediction distributions is hidden when one uses a deterministic model. In addition of being more realistic, the stochastic model permits also expressing predictions as estimated probabilities, as opposed to point estimates of trasgene fate. The demographic stochastic model encompasses variability associated with the risk of trasgene fixation, which is not accomplished by the deterministic model. More importantly, accounting for stochasticity can increase the persistency of the transgene beyond what the deterministic model predicts, even where the trasgene would have been lost early. Thus, a more realistic and informative assessment of the possible risk of trasgene invasion into natural populations can be obtained by using a stochastic implementation of the net fitness components model.
Publications
- Ragavendran A, Muir WM, Howard RD and Rosa GJM. Risk assessment of transgene invasion in natural populations: effect of demographic stochasticity on model predictions. Transgenic Research 16(6): 862-863, 2007.
- Rosa GJM, Ragavendran A, Muir WM and Howard RD. Extending the Models for Prediction of Transgene Fate to Incorporate Uncertainty and Validation of the Model. In: 2nd Symposium for Agricultural Biotechnology Risk Analysis Research, FDA Wiley Building, College Park - MD, December 5-6, 2007.
- Rosa GJM. Experimental design in genetical genomics. Brazilian Journal of Animal Science 36(SE): 211-218, 2007.
- Muir WM, Rosa GJM, Pittendrigh BR, Xu S, Rider SD, Fountain M and Ogas J. 2008. A mixture model approach for the analysis of small exploratory microarray experiments. Journal of Computational Statistics and Data Analysis. In Press doi:10.1016/j.csda.2008.06.011
|
|