Analysis of the effectiveness of a family of GeneMark algorithms for the genomes annotation

In this paper, we critically survey a family of GeneMark algorithms used for automatic annotation of new prokaryotic, particularly, bacterial genomes without using comparisons (alignment) with known genes and proteins. It is shown that the GeneMark algorithm is arguably the best among others in the class of ab initio gene prediction algorithms. Still, the algorithm could be improved by a more in depth consideration of the regulatory sites located near translation initiation sites as well as by taking into account the local variations of the genomic GC composition frequently related to events of horizontal gene transfer from other microbes.