Enzyme-specific profiles for genome annotation: PRIAM

Abstract

The advent of fully sequenced genomes opens the ground for the reconstruction of metabolic pathways on the basis of the identification of enzyme-coding genes. Here we describe PRIAM, a method for automated enzyme detection in a fully sequenced genome, based on the classification of enzymes in the ENZYME database. PRIAM relies on sets of position-specific scoring matrices ('profiles') automatically tailored for each ENZYME entry. Automatically generated logical rules define which of these profiles is required in order to infer the presence of the corresponding enzyme in an organism. As an example, PRIAM was applied to identify potential metabolic pathways from the complete genome of the nitrogen-fixing bacterium Sinorhizobium meliloti. The results of this automated method were compared with the original genome annotation and visualised on KEGG graphs in order to facilitate the interpretation of metabolic pathways and to highlight potentially missing enzymes.

Figure 1

Distribution of the number of modules selected for each enzyme collection in PRIAM (logarithmic scale).

Figure 2

Example of modular enzymes implying an ‘AND’ rule. Modules detected for homocitrate synthase (EC 4.1.3.21) are displayed using the XDOM program (35). Both black and red striped modules are required in order to infer the presence of homocitrate synthase.

Figure 3

Calibration of PRIAM methodology on complete genomes. The specificity and sensitivity of enzyme predictions were calculated using as standards the enzyme sets from the complete genome annotation found in SWISS-PROT for B.aphidicola, E.coli, H.influenzae, M.genitalium and M.pneumoniae (19). The mean and standard deviation of PRIAM specificity (squares) and sensitivity (triangles) were calculated for different RPS-BLAST E-values.

Figure 4

Example of a colour-coded KEGG metabolic chart used for interpreting S.meliloti metabolism (http://genopole.toulouse.inra.fr/bioinfo/priam/). Green boxes correspond to EC numbers found by both annotation and PRIAM (dark or pale green: E-value below or above 10^–30, respectively). Orange and yellow boxes correspond to enzymes not annotated, yet detected by PRIAM with an E-value below or above 10^–30, respectively. Red boxes indicate annotated enzymes not detected by PRIAM. Blue boxes correspond to EC numbers for which no sequence information was available in the ENZYME database, hence for which no PRIAM profile could be built. Yellow, orange and red boxes indicate a discrepancy between PRIAM analysis and current genome annotation, suggesting metabolic steps that should be reinvestigated.