2000 , Volume 5, Special issue, p.16-24

The methods designed for extracting information about gene structure on the basis of EST (Expressed Sequence Tags) analysis are being widely exploited in recent years. However, mRNA sequences can be reconstructed from ESTs with low accuracy. This fact may be explained both by numerous errors in the course of EST sequencing (nucleotide substitutions, insertions, and deletions) and by low accuracy of the programs reconstructing open reading frames on the basis of EST analysis. The programs used for this purpose (e.,g., GRAIL, GeneFinder, GeneScan) were initially designed not for EST analysis but for gene structure prediction within extended DNA regions. In this connection, it is necessary to develop special tools for the EST analysis. We have developed the program ORFScan that reconstructs the open reading frame by means of dynamic programming and multiple ESTs alignment. It consists of the following stages: (i) multiple alignment of ESTs is performed by Cap2 [21]; (ii) construction of the EST-consensus; (iii) reconstruction of translational open reading frame (ORF) on the basis of this consensus; (iv) reconstruction of amino acid sequence corresponding to this frame. The distinctive feature of the program presented is the usage of not only the resulting consensus in the course of the ORF reconstruction, but the EST alignment process in a whole. Application of the program developed for the ORF reconstruction provides the false positive and false negative estimates equaling to 6 % and 4 %, respectively.