185 clones with randomized ribosome binding sites, from position −11 to 0 preceding the coding region of β-galactosidase, were selected and sequenced. The translatlonal yield of each clone was determined; they varied by more than 3000-fold. Multiple linear regression analysis was used to determine the contribution to translation initiation activity of each base at each position. Features known to be important for translation initiation, such as the initiation codon, the Shlne/Dalgarno sequence, the identity of the base at position −3 and the occurrence of alternative ATGs, are all found to be important quantitatively for activity. No other features are found to be of general significance, although the effects of secondary structure can be seen as outliers. A comparison to a large number of natural E.coli translation initiation sites shows the information profile to be qualitatively similar although differing quantitatively. This is probably due to the selection for good translation initiation sites in the natural set compared to the low average activity of the randomized set.

Author notes

Present addresses: +Institute of Molecular Biology, University of Oregon, Eugene, OR 97403,
Present addresses: §Energy Biosystems Corp., 4200 Research Forest Drive, The Woodlands, TX 77381
Present addresses: πMedical University of South Carolina, Department of Psychiatry, 171 Ashley Avenue, Charleston, SC 29425-0742
Present addresses: National Cancer Institute, Frederick Cancer Research and Development Center, Laboratory of Mathematical Biology, PO Box B. Frederick, MD 21702-1201, USA