A Simple and Fast Homology-Based Gene Prediction in Mitochondrial Genomes

Abstract

With the abundance of genomic data after the Human Genome Project, the need for analysis, and annotation of these data arise. Annotation of genomes helps us understand the functionality of different parts of the genomes of various species. In this thesis, we propose a simple, and fast homology-based gene prediction method called Exon Hunter (EH) that achieves a performance comparable with state-of-the-art methods in mitochondrial genomes. Mitochondria are crucial for a eukaryotic cell, and mutation in its DNA has connections with disorders such as Alzheimer and cancer. We used Hidden Markov Model (HMM) Protein Profile of a number of genes to search for protein-coding genes in different genomes. Our method forms every subset of the hit set, and calculates a score for each subset according to an objective function. Then it chooses the subset with the\ highest score. Finally, we analyze the codon usage bias of our dataset, and we discuss how it can help us improve this prediction. ExonHunter is written in Python and is publicly available on github.com/amirh-hajianpour/ExonHunter.