Process Pathway Inference via Time Series Analysis
C. H. Wiggins, I. Nemenman
2003
Experimental Mechanics
Motivated by recent experimental developments in functional genomics, we construct and test a numerical technique for inferring process pathways, in which one process calls another process, from time series data. We validate using a case in which data are readily available and we formulate an extension, appropriate for genetic regulatory networks, which exploits Bayesian inference and in which the present-day undersampling is compensated for by prior understanding of genetic regulation.
more »
... on The last decade has witnessed stunning advances in experimental biology, particularly in the fields of neuroscience and genomics, which have made possible "data-driven" biological investigations. As examples, the quantitative revolution of genomics has provided terabytes of transcriptome data; and neuroscientists routinely record for hours or even days from multiple neurons simultaneously. This transformation stands as a challenge to theorists who hope to advance understanding by making a connection between experiment and first principles models. † In genomics, for example, we are presented with the expression levels of thousands of genes, but our ability to model is limited not only quantitatively, in that there are myriad unknown rate constants and binding parameters, but qualitatively, in that a sizable fraction of proteins and genes remain of uncharacterized function. 1 Similarly, in neuroscience, we can model patches of cellular membranes, synapses, and (at least electro-physiologically) entire cells. 2 However, this modeling hinges on numerous unknown parameters, and even if we can perform massive computations involved in the study of even rather small biological neural networks, the sensitivity to these parameters still makes the whole approach intractable. The astronomical amounts of experimental data are troubling † The mathematization of such models are referred to below as the "microscopic equations"; consider, for example, those of fluid dynamics which govern, yet certainly fail to encapsulate, such phenomena as turbulence and the tumbling of a falling leaf.
doi:10.1177/00144851030433016
fatcat:ggevx34teralhgypzer6juzbne