Abstract
The development of next generation sequencing facilitates the study of metagenomics. Computational gene prediction aims to find the location of genes in a given DNA sequence. Gene prediction in metagenomics is a challenging task because of the short and fragmented nature of the data. Our previous framework minimum redundancy maximum relevance-support vector machines (mRMR-SVM) produced promising results in metagenomics gene prediction. In this paper, we review available metagenomics gene prediction programs and study the effect of the machine learning approach on gene prediction by altering the underlining machine learning algorithm in our previous framework. Overall, SVM produces the highest accuracy based on tests performed on a simulated dataset.