| Due to the volume and complexity of data arising from high throughput biological assays, computational analysis becomes increasingly important to assist biologists in forming and testing hypotheses. In the current study, bioinformatics is applied to the fields of microbiology and toxicogenomics in analyzing chIP-chip data to study the thyroid hormone receptor conducted by Health Canada. This data analysis requires normalization and signal detection. A survey of contemporary methods was performed in order to find the most appropriate model for each step, given our experimental platform. Proof of concept experiments using high quality benchmark data revealed that normalization for chIP-chip data didn't improve the accuracy of subsequent peak finding algorithms. Splitter was used to detect peaks, which revealed 230 regions for which the thyroid hormone receptor is believed to be bound to DNA. Once signal detection was complete, the identified DNA segments were examined to model the degenerate sequence motif. Motif finding algorithms (MFAs) from a number of underlying statistical models were also applied to find occurrences of novel motifs not previously known to be linked to the thyroid hormone receptor. In total, 105 thyroid hormone receptor binding sites (thyroid response elements) were identified with an expected false discovery rate of 20%. MFAs found motifs which are very similar to known binding sites for proteins which could interact with the thyroid hormone receptor, such as SP-1, PAX and KROX binding sites. A wet laborary validation of theses sites is now needed in order to reveal the functionality of these sites, Le. whether the identified motifs truly exhibit a gene regulation function. |
