Large-Scale Comparative Annotation of Bacterial Genomes

The MSc thesis describes the design and evaluation of a comparative annotation system for prokaryotic genomes. The proposed system makes use of a combination of comparative and abinitio gene finding techniques for accurate annotation. The three main modules described in the system are the markov-model based gene finding module, the alignment based profiler and the comparative module that combines the results from both modules. The system works by first creating a multiple sequence alignment of the input sequences and doing independent gene finding through a combination of markov models of order one to five. These two operations are carried out in parallel and there are no interdependencies. The multiple sequence alignment is then used to create a profile that emphasizes regions of high conservation within the sequences. The results from the markov model module are then processed through the profile to give a final comparative annotation. The system was evaluated by carrying out a series of tests with an increasing number of sequences in each test set. The results obtained were quite encouraging, in that the gene of interest was found 100% of the time, with the false positives being removed very accurately due to the comparative technique. Moreover the results indicated as expected, that increasing the number of sequences in the test sets has a very positive impact on resulting predictions and confidence scores.