Shotgun: getting more from sequence similarity searches

Motivation: As genomic sequencing reveals the range of structural classes g enerated through the evolution of proteins, analysis of the superfamilies t o which they belong can contribute important insights for understanding the ir structure-function relationships. Current database search techniques fal l short of identifying the majority of distant sequence relationships at st atistically significant levels. We developed the Shotgun program in an effo rt to enhance the sensitivity and utility of current database search output . Results: We have developed and used the Shotgun program to identify both ne w superfamily members and to reconstruct several known enzyme superfamilies using BLAST database searches. An analysis of the false-positive rates gen erated in the analysis and other control experiments provides evidence that high Shotgun scores indicate real evolutionary relationships. Shotgun is a lso a useful tool for identifying subgroup relationships within superfamili es and for testing hypotheses about related protein families.