How to Fix It

To rememdy this, we feel that some modifications can be made:

Eucaryotes

Don Fong and David Haussler have looked at ways to use multiple states. A publication should be forthcoming describing the equations which are necessary to do this.

False +

For elimination of false positives, we think that analyzing contigs for G/C content and modifying the gene codon usage tables might be successful [6]. Another way to eliminate false positives could be to strengthen the translational initiation sensor by studying the -35 regions as well as the -10 region.

tRNA & mRNA

For recognition of tRNA and rRNA genes, their sequences could be analyzed statistically to see if their recognition is compatible with mRNA genes. It may be that we only need to build a length histogram and codon usage table from a training dataset of all three types of genes, and weaken the start and stop codon sensors (tRNA and rRNA genes don't have the same start and stop codons as mRNA genes)