DETECTION OF GENES IN ESCHERICHIA-COLI SEQUENCES DETERMINED BY GENOMEPROJECTS AND PREDICTION OF PROTEIN-PRODUCTION LEVELS, BASED ON MULTIVARIATE DIVERSITY IN CODON USAGE

We used principal component analysis to develop measures (called Z-par ameters in this study) which reflect the diversity of codon usage in E scherichia coli genes. Protein production levels for 1500 CDSs (protei n-coding sequences) identified by E.coli genome projects in Japan and the US were estimated from a correlation equation between Z(1) and cel lular protein content obtained through analysis of the genes experimen tally characterized. Through the profile analysis of Z(1) for E.coli s equences obtained by;he Japanese Project, we predicted an additional 3 6 CDSs that had not been annotated in the International DNA Database. Thirtyone out of the 36 CDSs could be assigned to presumptive protein genes through a BLASTX search for recent protein databases in the Geno me Net in Japan. Detailed examination of the Z(1)-parameter profile le d us to assess sequencing errors which cause frame-shift.