Curve alignment

The occasional need to align gene expression profiles comes from the fact that the rates of biological processes may be different across biological or environmental conditions, or the sampling times are different between two time course datasets to be compared. Aach and Church (79) suggested a time-warping algorithm with or without interpolation, while Bar-Joseph et al. (43) gave a B-spline approach also for the same task, assuming that each gene's profile is fitted with gene-specific and class-specific parameters. As with so much we have mentioned, these approaches have only been illustrated on the yeast cell-cycle data. Again, this idea is well-suited to longer time series, but may not be suitable for shorter time series. There is clearly room for more work here.

0 0

Post a comment