Analysis of mRNA folding patterns reveals several characteristics that vary widely among natural genes, with potential biological and evolutionary implications that could lead to new classification schema. The predicted pairing of bases within the coding region of a folded mRNA molecule can be assigned one of three possible alignment frames. Two of these three coding frame alignments can also be used to define a 'family' relationship which partitions the genetic code into subsets of the twenty amino acids, in the sense that bases belonging to member codons of the same subset can pair with one another in the particular frame that defines that partition, but bases belonging to codons in two different subsets cannot participate in such a pairing.
The GCG MFOLD program (adapted from Zuker) was used to fold the coding regions of 65 genes, and a program was written to analyze the resulting output in terms of the above criteria, along with the more usual measures such as base composition. Many of these genes were divided rather evenly over the several amino acid subsets, but some were chosen initially for a strong bias in favor of one family or another, excluding genes composed of mostly one or very few amino acids as too obviously biased in terms of family membership. Several genes exhibited significant biases in pairing alignment frames, with one particular frame or another being either preferred or avoided, both among the genes selected for a family bias and among those with fairly average distributions of family membership.
The results from known genes are being compared with the same analysis conducted on non-coding open reading frames, as a means to develop new criteria for identifying genes.