On 2015 Jun 26, Donald Forsdyke commented:

In the light of new reviews (Zhang J, 2015 and Forsdyke DR, 2015), the following email to the senior author (Feb 6 2013) may be of interest:

'I was delighted to see a new paper from your laboratory in this week’s PNAS Early Edition Park C, 2013 and, as expected, it did not disappoint me. The convergence of our thinking can be seen in the earlier emails and the attached articles on "Functional Constraint" (ELS where I squeezed in a last-minute reference to your work) and on "Collective Gene Functions" (BioEssays). Since I differ from you and your coauthors in some current interpretations, I hope you will not mind my again offering comments, which I hope will be helpful.

1. I think you would agree that your statement "Amino acid substitutions are slower as the mRNA folding strength increases" could be restated as "Amino acid substitutions are faster as the mRNA folding strength decreases". The latter was the main observation of my paper in Molecular Biology and Evolution (1995 12, 1157-1165, http://post.queensu.ca/~forsdyke/introns.htm). In other words, under strong Darwinian positive selection (high dN), amino acid substitutions (codon changes) primarily serve the needs of protein function, rather than of nucleic acid structure. When the protein is conserved (evolving slowly) then the needs of nucleic acid structure are more readily accommodated. This approach can even be used to assay positive Darwinian selection (see Immunogenetics 1996 43, 182-189, http://post.queensu.ca/~forsdyke/introns.htm). The point was recently reiterated (see J. Biol. Systems 2007 15, 95-108, http://post.queensu.ca/~forsdyke/speciat3.htm).

2. While your Introduction implies the possibility of "a selective pressure at any level (DNA, mRNA, or protein)," you assume that there is "a major role of natural selection at the mRNA level in constraining protein evolution." You do not mention the possibility that mRNA structure is by default, because the encoding DNA needs structure (i.e. DNA has the potential to extrude single strands as stem-loops). Since such structure is pervasive in DNA (in exons and introns and in extragenic locations), it appears that DNA needs structure, whether it encodes proteins or not. I have considered the rationale for this elsewhere.

3. You note that "most of the correlations reported in this study are strong" when using enzymic (Fig. 1a) rather than computational (Fig. 1b) RNA structure determination. And you point out that mRNA folding strength may be impacted by "different amino acid, nucleotide, and synonymous codon frequencies." Here it might have been better first to look at the nucleotide component by comparing folding energies with randomly shuffled sequences. Your use of "pseudo-mRNAs" as a basis for comparison was useful for eliminating the possibilities of roles for "specific protein sequences or synonymous codon usages," but would have greatly reduced the statistical significance of your results.

4. You call for "further improvement of the computational prediction." Base composition tends to be a genome-wide, or segment-wide, characteristic; it tends not to be a local characteristic. On the other hand, base order is a local characteristic. Nucleic acid structure depends on both base composition and order, but for many purposes, base order provides a more sensitive measure. If you follow my method for dissecting out the base order-dependent component of the folding energy, I believe you will obtain more satisfactory results with the computational method of structure determination.

5. Compared with low expressed proteins, highly expressed proteins are more likely to have both specific and collective functions. Thus, there are two sources of negative selective pressure on genes with highly expressed protein products, and only one source of negative selection pressure on genes with low expressed protein products. So the former genes tend to evolve more slowly, and are less likely to be affected by deletion mutations since their contributions to collective functions are shared with other highly expressed proteins, which can compensate. Since they evolve slowly, they are better able to accommodate structure at the nucleic acid level. I consider this more extensively elsewhere.

6. Genes under positive Darwinian selection pressure may be of various expression levels, so "amino acid substitution rate is negatively correlated with mRNA folding strength, with or without the control of expression level." Furthermore, there is "a significant [negative] correlation between mRNA folding strength and dN/dS, even when gene expression level is controlled."

All this comes with no guarantee. I may be wrong. If you have any problem with these comments please get back to me. Having crossed this territory in the 1990s I feel some obligation to warn later explorers of possible pitfalls!'

The ELS and BioEssays papers referred to in the above email were:

Nature Encyclopedia of Life Sciences 7, 396-403 (2002). Functional constraint and molecular evolution. This was updated for Wiley Online Library in 2005 and 2012.

BioEssays (2012) 34, 930-933 Forsdyke DR, 2012.

On 2015 Jun 26, Donald Forsdyke commented:

In the light of new reviews (Zhang J, 2015 and Forsdyke DR, 2015), the following email to the senior author (Feb 6 2013) may be of interest:

'I was delighted to see a new paper from your laboratory in this week’s PNAS Early Edition Park C, 2013 and, as expected, it did not disappoint me. The convergence of our thinking can be seen in the earlier emails and the attached articles on "Functional Constraint" (ELS where I squeezed in a last-minute reference to your work) and on "Collective Gene Functions" (BioEssays). Since I differ from you and your coauthors in some current interpretations, I hope you will not mind my again offering comments, which I hope will be helpful.

1. I think you would agree that your statement "Amino acid substitutions are slower as the mRNA folding strength increases" could be restated as "Amino acid substitutions are faster as the mRNA folding strength decreases". The latter was the main observation of my paper in Molecular Biology and Evolution (1995 12, 1157-1165, http://post.queensu.ca/~forsdyke/introns.htm). In other words, under strong Darwinian positive selection (high dN), amino acid substitutions (codon changes) primarily serve the needs of protein function, rather than of nucleic acid structure. When the protein is conserved (evolving slowly) then the needs of nucleic acid structure are more readily accommodated. This approach can even be used to assay positive Darwinian selection (see Immunogenetics 1996 43, 182-189, http://post.queensu.ca/~forsdyke/introns.htm). The point was recently reiterated (see J. Biol. Systems 2007 15, 95-108, http://post.queensu.ca/~forsdyke/speciat3.htm).

2. While your Introduction implies the possibility of "a selective pressure at any level (DNA, mRNA, or protein)," you assume that there is "a major role of natural selection at the mRNA level in constraining protein evolution." You do not mention the possibility that mRNA structure is by default, because the encoding DNA needs structure (i.e. DNA has the potential to extrude single strands as stem-loops). Since such structure is pervasive in DNA (in exons and introns and in extragenic locations), it appears that DNA needs structure, whether it encodes proteins or not. I have considered the rationale for this elsewhere.

3. You note that "most of the correlations reported in this study are strong" when using enzymic (Fig. 1a) rather than computational (Fig. 1b) RNA structure determination. And you point out that mRNA folding strength may be impacted by "different amino acid, nucleotide, and synonymous codon frequencies." Here it might have been better first to look at the nucleotide component by comparing folding energies with randomly shuffled sequences. Your use of "pseudo-mRNAs" as a basis for comparison was useful for eliminating the possibilities of roles for "specific protein sequences or synonymous codon usages," but would have greatly reduced the statistical significance of your results.

4. You call for "further improvement of the computational prediction." Base composition tends to be a genome-wide, or segment-wide, characteristic; it tends not to be a local characteristic. On the other hand, base order is a local characteristic. Nucleic acid structure depends on both base composition and order, but for many purposes, base order provides a more sensitive measure. If you follow my method for dissecting out the base order-dependent component of the folding energy, I believe you will obtain more satisfactory results with the computational method of structure determination.

5. Compared with low expressed proteins, highly expressed proteins are more likely to have both specific and collective functions. Thus, there are two sources of negative selective pressure on genes with highly expressed protein products, and only one source of negative selection pressure on genes with low expressed protein products. So the former genes tend to evolve more slowly, and are less likely to be affected by deletion mutations since their contributions to collective functions are shared with other highly expressed proteins, which can compensate. Since they evolve slowly, they are better able to accommodate structure at the nucleic acid level. I consider this more extensively elsewhere.

6. Genes under positive Darwinian selection pressure may be of various expression levels, so "amino acid substitution rate is negatively correlated with mRNA folding strength, with or without the control of expression level." Furthermore, there is "a significant [negative] correlation between mRNA folding strength and dN/dS, even when gene expression level is controlled."

All this comes with no guarantee. I may be wrong. If you have any problem with these comments please get back to me. Having crossed this territory in the 1990s I feel some obligation to warn later explorers of possible pitfalls!'

The ELS and BioEssays papers referred to in the above email were:

Nature Encyclopedia of Life Sciences 7, 396-403 (2002). Functional constraint and molecular evolution. This was updated for Wiley Online Library in 2005 and 2012.

BioEssays (2012) 34, 930-933 Forsdyke DR, 2012.