Nexperiment: Computational Platform for Model-based Design of Experiments to Reduce Dynamical Uncertainty

Aditya Sai (Dr. Ann Rundell, advisor)

Abstract: Model-based design of experiments (MBDOE) is a methodology that can plan optimal experiments to extract maximally informative data. While there are a multitude of MBDOE algorithms that aim for model discrimination and parameter precision, there are very few methods to address dynamical uncertainty reduction. Our lab has been extensively studying MBDOE, and have now developed a computational tool to aid those researchers interested in applying such methodologies to their own experiments. Nexperiment is a novel Matlab-based software that provides users a platform to apply experiment design methodologies to determine optimal designs that reduce dynamical uncertainty of model dynamics. In addition to an optimized input experiment design, Nexperiment offers sequential and parallel experiment design options, conventional Fisher Information Matrix (FIM) based experiment design, global & local sensitivity analyses, data consistent dynamics identification and model simulation. It does so by providing an easy-to-use interface that extracts and records essential details about the model supplied by the user. The results presented by the GUI are concise and easily implementable by experimentalists. Nexperiment provides a vital resource to those in the systems biology community by allowing them to analyze their own models for optimal measurements that will support model development. We plan to release Nexperiment into the Matlab User Community in the near future.

Using Singular Value Decomposition to Analyze Biological Datasets of Skeletal Diseases

Andy Chen (Dr. Hiroki Yokota, advisor)

Abstract: Singular value decomposition (SVD) is a linear algebra technique useful in characterizing large matrices.  This technique has many applications, such as in data compression, signal processing, and pattern recognition.  We have used SVD to perform principal component analysis on biological data to derive biologically useful patterns and characteristics that may elude more naïve analyses of that data.  Here I present two examples of its use: in analyzing gene expression data from microarrays; and in helping to identify possible drug targets for breast cancer and bone metastasis.  We are currently trying to use SVD and other computational techniques to identify DNA mutations in cancer cells that promote bone metastasis, which we hope leads to the development of novel therapies to treat bone metastases.

***Bring your lunch to seminar – BMEGSA will provide snacks and drinks***

Also available via WebEx meeting in SL220A at IUPUI