Constructing optimal convolutional code models for prokaryotic translation initiation

Rapid advances in both genomic data acquisition and computational technology have encouraged the development and use of engineering methods in the field of bioinformatics and computational genomics. Several researchers are encouraging the use of error-correction coding in analyzing genetic data [1][2]. Using information theory, coding theory specifically, the translation of messenger RNA (mRNA) into amino acid sequences is functionally paralleled to the decoding of noisy, convolutionally encoded parity streams. The ribosome is modeled as a table-based convolutional decoder. This work presents a genetic algorithms (GAs) method for the design of optimal table-based convolutional coding models for prokaryotic translation initiation sites using Escherichia coli K-12 as the model organism. We explore and compare several categories of error-control codes, including: horizontal, vertical, equal and unequal error protection (UEP) codes. Results show that UEP code models recognize the non-random and Shine-Dalgarno domain of mRNA leaders better than equal error protection models. Codes whose decoding masks (gmasks) have high similarity to the 3′ end of the 16S ribosomal RNA (rRNA) were discovered. Additional results are presented.