Assessing ancestral genome reconstruction methods by resampling

Abstract

Inference of ancestral or extinct genomes is important in evolutionary biology, cancer research and many other research areas. Whole-genome data has become readily available due to advances in large-scale sequencing technology. During the past decades, a number of evolutionary models and related algorithms have been designed to infer ancestral genome sequence or gene order. Since it is so hard or even impossible to know the true scenario of the ancestral genomes, there must be some tools used to test the robustness of the adjacencies obtained from existing methods. However, till now, there is still no systematic work being conducted to tackle this problem. On the other hand, it is a common practice in phylogenetic analysis to assess the confidence rate of the inferred branches, using techniques such as bootstrapping and jackknifing, which have been evaluated as good resampling tools for phylogenetic reconstruction methods. Some of these resampling methods can be potentially used as a robustness test for ancestral genomes. In this paper, we conduct large-scale experiments by using three existing best ancestral genome inference methods and four different resampling techniques. Experimental results show that resampling techniques are useful for assessing the quality of ancestral genomes, while different methods have different preferred resampling techniques. It is suggested the cut-off threshold of 75% should be used to filter bad gene adjacencies.