What Gene Prediction in Transcriptomes, Metatranscrptomes, Metagenomes, and Microbial Genomes has in Common

NGS technology is instrumental in production of many different types of sequence data, including draft prokaryotic genomes, eukaryotic transcripts assembled from RNA-Seq reads, sequences of prokaryotic metagenomes and eukaryotic metatranscriptomes. All these sequences may carry complete or partial protein-coding regions with no introns. An accurate prediction of the protein-coding genes requires precise and robust estimation of parameters of the statistical models used in gene finding algorithms.
We describe a general method of model training that could be used for all these genomic datasets that differ significantly in terms of size and diversity of sequence population.
The method has already been used in multiple applications.