Nonparametric Identification of Regulatory Interactions from Spatial and Temporal Gene Expression Data

by Claire Tomlin (Berkeley and KTH)

Tuesday 2 Feb 2010, 13:30 → 13:30 Europe/Stockholm

122:028

The correlation between the expression levels of transcription factors and their target genes can be used to infer interactions within animal regulatory networks. Here we describe a novel approach which uses nonparametric statistics to generate ordinary differential equation (ODE) models. In contrast to current methods that use preexisting knowledge of the network, our approach requires minimal information about the mathematical structure of the ODE or qualitative descriptions of interactions within the network. Our approach also uses new statistics to protect against over-fitting and generates spatio-temporal maps of factor activity, highlighting the times and spatial locations at which different regulators might affect target gene expression levels. We use our technique to study eve mRNA pattern formation in the Drosophila melanogaster blastoderm. The results suggest that protein factors acting on a single cis-regulatory module frequently have the potential to behave as both an activator and inhibitor depending on the factor's concentration. Our method also distinguishes different modes of activation and repression. We compare our model to other models that use gene expression data to examine regulatory interactions, and discuss the advantages our method has over these existing methods.

This is joint work with Anil Aswani (EECS, UC Berkeley), Peter Bickel (Statistics, UC Berkeley) and Mark Biggin (Genomics and Life Sciences Division, Lawrence Berkeley Labs).