Granule cell precursors of the cerebellum undergo exuberant proliferation and subsequent differentiation during postnatal days 1-21 (Pl-21) in the mouse. Granule cells are also thought, to give rise to medulloblastoma, a brain tumor primarily affecting children. This proposal is aimed at characterizing gene expression profiles during the highly proliferative phase of cerebellar development in the postnatal mouse. Existing analytical methods for elucidating the functional relationships between genes expressed in cells typically fall to account for significant sources of variation (i.e., """"""""noise"""""""") inherent in the data. This problem will only be compounded as analysis progresses from unicellular organisms to complex tissues such as the developing CNS. The hypothesis driving the proposed research plan is that improved classification and clustering techniques in functional genomics can be obtained by incorporating sources of variation/noise into the analytic method. The proposed work will utilize a model system of the developing mouse cerebellum, from which improved methods for the interpretation of complex, DNA microarray-based gene expression data will be developed.
Aim 1 is to generate a gene expression data set (consisting of the analysis of 19,000 genes) from the newborn and neonatal mouse cerebellum at various developmental stages (P1, 4, 7, 10, 14, 18, 21, 30). Math-1 is a basic helix-loop-helix transcription factor with essential roles in cerebellar granule cell development. We will utilize transgenic mice in which selectable reporter genes (e.g., jellyfish green fluorescent protein, GFP) are expressed under control of the granule precursor cell-specific Math-1 enhancer. This will allow for further subdivision into proliferating, immature granule cell (Math-1-positive) and other cell types (Math-1-negative) by fluorescent cell activated sorting (FACS).
Aim 2 - The time series obtained in Aim 1 will be analyzed in both the time and frequency domain, using Fourier transforms of the expression data. Classification and clustering techniques that fully exploit the """"""""noise models"""""""" developed in Aim 1 will be applied to these time series.
In Aim 3 we will screen candidate genes/ESTs to confirm appropriate spatio-temporal-specific expression in developing cerebellar granule cells from P0-P21 using a high throughput in situ hybridization protocol. Deliverables from the work include an open source, public domain, web-accessible tool kit of noise-aware machine learning (classification and clustering) tools that are anticipated to be widely applicable to the analysis of time series of gene expression networks in complex biological systems. In addition, it is anticipated that candidate genes with possible roles in cerebellar granule cell proliferation, differentiation or possibly, tumorigenesis, will emerge that can be analyzed in future hypothesis-driven projects.
