Estimating a property of a structure in data employing a prior probability distribution

摘要

A method of estimating a property of a structure in data comprising the steps of establishing a prior probability distribution, selecting values which characterise at least part of an initial approximation to a posterior distribution, updating a factor of the approximation in accordance with the other approximation factors and a likelihood function. The method is repeated for all approximation factors. An embodiment embraces microarray DNA data which is captured by image analysis 4. Estimates are made of spot radius and location of a spot centre 6. Expectation values of the posterior distribution are initialised at 8. The posterior function is updated by parameterising proposal distributions 10, obtain samples from the proposal distributions 12. An importance weight is determined 14, which is repeated for all samples. The importance weights are normalised 16. The importance weights are used to compute estimates of expectation values 18, and subsequent estimates 20. A convergence criterion is tested at 22 which if met causes the spot intensities and corresponding expression levels 24 to be determined. Finally the variance of the intensity values is calculated 26 which enables error bars to be determined.