I am a high school student and am currently learning about evolutionary relationship study in biology.

My teacher said that a comparative study of amino acid sequences is more useful than a comparative study of nucleotide sequences, because the genetic code is degenerate in nature — several codons may code for the same amino acid.

However, I just do not understand the logic.

Since several codons may code for the same amino acid, I (as a math person) consider the conversion of a nucleotide sequence to an amino acid sequence as a non-injective function, and thus is information-losing.

(Analogy: consider the function $f(x)=x^2$. Imagine that you have a number, and you plug it into $f(x)$ to get $1$ as the output. You would never know if the original number is $1$ or $-1$.)

Therefore I arrive at the exact opposite conclusion. Is my conclusion correct or not, and why?

  • $\begingroup$ I have corrected your misuse of Genetic Code leaving only the single correct instance of use (and tidied your question a little). I have explained this in the first part of my answer and given some more references. $\endgroup$ – David Aug 31 '19 at 16:58
  • $\begingroup$ I have provided you with a well documented answer. (Your teacher is right in most circumstances.) Please read it and let me know if there is anything you don't follow in the biology. We (or at lest I) cannot help you with the maths, so consider my suggestion that you — as a mathematician — find the flaw in your logic, and perhaps let us know when you have done so. This is a real-world example of the dangers of simplistic application of mathematical logic, so it worth your attention. $\endgroup$ – David Sep 4 '19 at 22:23
  • $\begingroup$ @David I think I need more time to understand as it is a bit complicated to me. Please be patient. $\endgroup$ – Szeto Sep 4 '19 at 23:27
  • $\begingroup$ No hurry. In fact it might be better to ask the question on a Maths site. One difficulty for you will be lack of familiarity with the sequence comparison algorithms. Because these algorithms compare sequences on a letter-by-letter basis, when they compare DNA they are not using the information of the genetic code, which would involve comparison in groups of three. There must be other examples where mathematical transformations driven by external considerations produce a more useful form of data. $\endgroup$ – David Sep 5 '19 at 12:49

Each has its own utility, depending on the time-frame you are looking at. For evolutionary studies, you need variation, but not so much variation that one substitution at the same position overwrites a previous substitution. So if you are looking at deep splits, over hundreds of millions of years, it may be that amino acids are more reliable. But you are correct in that since they are functionally more important than silent substitutions (nucleotide changes that don't change the amino acid), it is possible to have amino acids converge on the same state, independently. Nucleotides do this too. Both require statistical models (maximum likelihood) that accommodate the possibility of multiple changes at the same site. If you are looking at recent evolutionary splits, there may not be enough (or any) amino acid changes to compare, so in this case, nucleotides would be better. You would not measure continental drift with a stopwatch, or a 100-meter dash with radiometric dating.


The Answer

It is correct that the product of the conceptual translation of a nucleotide sequence into an amino acid sequence results in the loss of certain information present in the former. An obvious example is that the amino acid sequences of the same protein in two individuals may be identical, but there may be silent mutations in the DNA, and these can be useful in tracing ancestry. The one part of Crick’s Central Dogma about which there can be no argument is that you cannot go from protein to DNA because the information for the nucleotide sequence is not present in the protein, with or without the genetic code.


An amino acid sequence contains information that is not present in the gene from which it originates, if we just consider the sequence as mathematical sequence of symbols. And, with 20 letters instead of 4, this new information has a different (and greater) complexity. The mistake is an unspoken assumption that the information of the genetic code is inherent in the nucleotide sequence. It is not. Yes, if we have the information of the genetic code, then the nucleotide sequence also has the information of the amino acid sequence, but that is not the practical question at issue.

So (addressing the poster) in the majority of practical instances your school teacher is correct. I am not a mathematician, so I cannot be sure what is the flaw of your argument. Perhaps it is the fact that only a subsection of the information can be used in the sequence comparison, perhaps the fact that you are talking three symbols from a set of four in your non-injective function to produce one symbol from a set of 20, or perhaps it is the biology. That’s something for you to work out. But if your conclusions are wrong (which they are) there must be a flaw in your logic)

The Question at Issue

The practical question at issue is:

Which is more suitable for determining the evolutionary relatedness of two organisms — a pairwise comparison of the amino acid sequences of a functionally similar protein (e.g. cytochrome c) or of the nucleotide sequences of the corresponding gene?

The general answer is:

It depends on the relatedness of the organisms, but, except for very close kinship (e.g. humans and neanderthals) or certain specialized problems, the answer is likely to be the amino acid sequences.

How can this be?

In relationship to the evolutionary distance between organisms, it is necessary to consider the different rates at which nucleotides and amino acids mutate, and the constraints on what mutations are likely to occur. If the rate of mutation is too fast there will be a time difference after which it will be difficult or impossible to calculate their evolutionary divergence accurately and ultimately even to detect any relationship between them.

Nucleotides mutate more rapidly than amino acids, and in practice comparison of nucleotide sequences is less useful than comparison of amino acid sequences for longer timespans.

  1. Because of the degeneracy of the genetic code (the fact that an amino acid can be encoded by more than one triplet of nucleotides) it is possible for one or even two nucleotides to mutate without affecting the amino acid sequence. (And the similarity between sequences is computed from a letter-by-letter comparison.)

  2. Statistics is not my forté, but in a general sense, because there are only four bases, 25% identity between two nucleotide sequences would be expected to occur by chance, whereas two amino acid sequences that are 25% identical would be statistically significantly similar because there are 20 amino acids. (Only 5% identity would arise by chance.)

There is a further aspect of divergence of amino acid sequence that is useful for evolutionary comparison, and this is that the nature of the mutation of amino acids is far more constrained than that of nucleotides. Admittedly purine-to-purine or pyrimidine-to-pyrimidine mutations are more frequent than purine/pyrimidine mutations, but amino acid mutations are often constrained by the role the amino acid plays in a protein. However one can construct empirical matrices of the likelihood of different amino acid mutations to obtain a more subtle and accurate estimate of relatedness.

What this means in practice is that instead of having to use a scoring system for comparisons of amino acid sequences that is either 1 for identity or 0 for non-identity, one can use a scoring system that gives ‘half marks’ (as it were) for structural/functional similarity. Thus, two amino acid sequences having 5% identity in pairwise comparison could be shown to be related because of an overall higher ‘similarity’ score.

Appendix 1: Sequence Comparison

It is important to realize that however much information resides in nucleotide or amino acid sequences, only the information that is actually used in the practical methods of determining evolutionary differences is relevant. These methods involve computer programs that compare sequences according to mathematical algorithms to answer the question of how similar two (or more) sequences are. So, regardless of the fact that the amino acid sequence is generally computed from the gene sequence, the question is “should I imput nucleotide or amino acid sequences into the program to get the best comparison?”. It is in this context that the remarks above about rate of change and likelihood of interconversions should be taken.

To quote from an article by one of the pioneers in sequence comparison, W. R. Pearson:

“Protein (and translated-DNA) similarity searches are much more sensitive than DNA:DNA searches. DNA:DNA alignments have between 5–10-fold shorter evolutionary look-back time than protein:protein or translated DNA:protein alignments. DNA:DNA alignments rarely detect homology after more than 200–400 million years of divergence; protein:protein alignments routinely detect homology in sequences that last shared a common ancestor more than 2.5 billion years ago (e.g. humans to bacteria). Moreover, DNA:DNA alignment statistics are less accurate than protein:protein statistics; while protein:protein alignments with expectation values < 0.001 can reliably be used to infer homology, DNA:DNA expecation values < 10−6 often occur by chance, and 10−10 is a more widely accepted threshold for homology based on DNA:DNA searches.”

There are Wikipedia articles about sequence alignment, and about the use of BLOSUM and PAM matrices. The section on sequence alignment in Berg et al. online — which involves amino acid, rather than nucleotide sequences — may also be of interest.

Appendix 2: Terminology and Definitions

As the term, Genetic Code, was misused in the unedited version of the question — and is widely misused in the press — I thought that a glossary of terms might be helpful

DNA (from which the genome and its constituent genes are constructed) are linear polymers of 4 nucleotides. The order of these is called the nucleotide sequence, or, because the only the purine or pyrimidine base varies between nucleotides, the base sequence.

Proteins are linear polymers of 20* amino acids. The order of these is called the amino acid sequence.

The Genetic Code is a cipher — and can be represented as a table showing the correspondence between 64 triplets of three nucleotides and the 20 amino acids and three stop signals when these nucleotides are part of the translatable part of a gene. The genetic code is highly — but not absolutely — conserved between organisms (and differs for proteins encoded by mitochondrial DNA).

In NO circumstances can the word Genetic Code be used as a synonym of Genome, although this is abused by even the scientific press, and is difficult for computer programmers to come to terms with, working as they do in a field where the noun ‘code’ is used for the product of encoding instructions.

*The genetic code has a certain plasticity and two additional amino acids can be encoded by termination codons in certain circumstances.

  • $\begingroup$ In your referenced answer you say, "An amino acid sequence contains information that is not present in the gene from which it originates, if we just consider the sequence as mathematical sequence of symbols. And, with 20 letters instead of 4, this new information has a different (and greater) complexity. The mistake is an unspoken assumption that the information of the genetic code is inherent in the nucleotide sequence." Please explain this in more detail. Having a larger alphabet does not imply greater complexity nor that proteins contain information not written into them by DNA. $\endgroup$ – S. McGrew May 6 '20 at 21:06
  • $\begingroup$ @S.McGrew — I think it speaks for itself. The poster is concerned with a theoretical informational problem (of his own making). I explain how he has forgotten the additional information required if one performs the conceptual translation of a nucleotide sequence into amino acids. The practical explanation of why protein sequences are used for evolutionary comparisons over periods longer than a particular time is different. It is just that the same number of mutations produce smaller changes in proteins than DNA. For comparison within a species or between closely-related species DNA is better. $\endgroup$ – David May 7 '20 at 7:54
  • $\begingroup$ You seem to be stating that a protein (encoded by DNA) contains more information than the DNA that encodes it. Ts certainly not self-evident and, according to my understanding, incorrect. $\endgroup$ – S. McGrew May 7 '20 at 10:32
  • $\begingroup$ @S.McGrew — No. I am saying that a string (in computer terminology) composed of the four characters ATGC has less mathematical complexity than a string containing the 20 different letters of the alphabet minus B, J, O and Z. Although it is possible to derive the latter from the former this does not violate the second law of thermodynamics (as applied to information) because 1. The knowledge of which One of the six reading frames to select has to be invoked (I don’t think I mentioned this) 2. A hash table (a description of the Genetic Code in programming terms) has to be used. $\endgroup$ – David May 7 '20 at 12:55
  • $\begingroup$ The "structure" of a protein encoded by DNA is of course different from the "structure" of the DNA encoding it, but the Kolmogorov complexity pf the two is identical. There is no new information in the protein, unless you are asserting that mutations occur spontaneously in proteins and are then inheritable. $\endgroup$ – S. McGrew May 7 '20 at 14:15

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.