List of Student Projects
Image Processing applied on micro-arrays for molecular biology
Semester Project, February 2001
In recent years, DNA micro array technology has become the most important tool in the analysis of gene expression studies. DNA micro array is a technique that allows studies of thousands of genes in one experiment. The basis for the micro array technology is to robotically spot minute amounts of clonal DNA, which represent genes, onto a small glass slide in an ordered fashion.
Common to all array based approaches is the necessity to analyze digital images of the array. The ultimate image analysis goal is to automatically assign a quantity to every array element giving information about the hybridization signal.
The objective of the micro array image analysis is to extract probe intensities or ratios at each cDNA target location, and then cross link printed clone target information so that biologist can easily interpret the outcomes and perform further high-level analysis.
To achieve the goal of accurately estimating the volume, i.e. the amount of genetic material of every spot the algorithm must cope with the following three major problems:
In order to achieve a successful identification and evaluation of each spots position and volume on the micro-array images, we decided to adopt a methodology based on reproducing, with the smallest distortion possible, these images by numeric copy.
1. Identify spot Model (parametric, non parametric )
2. Reproduce numeric spot Model
3. Identify the best resemblance measure tool (convolution, Euclidean distance )
4. Identify the best method to find the spot fitting the image (scanning, optimization )
The spots evolve around a pre-defined position and around an average sigma and amplitude. Therefore it is interesting to use this technique with the Euclidean distance as measurement tool, and aim at generating gaussians minimizing the slope all the for parameters. Because we were searching for several minimums (because we have several parameters to be identified), we chose the method to be one of the descendant gradient family (only with numeric derivate).