The proteins I'm interested in create disulfide bonds using cysteine pairs. In different states of the proteins, those pairs are sometimes formed, other times they stay separated.

I would like to calculate a score for the conservation of specific amino acids in my multiple sequence alignment (MSA).

The thing is that some of these amino acids or structures are irreplaceable. For example disulfide bridges must be conserved, otherwise the correct structure will not be adopted.

So, in some way, the score is already very high. But is there a way to calculate the score between different disulfide bridges?

Also, is there a way to state that a MSA shows an amino acid or motif to be highly conserved (as opposed to "just" conserved)? Is there a quantitative way to differentiate the one from the other?

  • $\begingroup$ Are you looking at just the amino acid sequence, or do you have nucleotide sequences as well to compare? $\endgroup$ Oct 20, 2023 at 16:13
  • $\begingroup$ there are many ways and terms that have been proposed for very conserved chunks of protein or DNA, e.g. ancient conserved region or ultraconserved region. However I think that it is more useful to just state the degree of conservation (100% identity, 99% similarity, etc.). There aren't really meaningful "conserved" vs "highly conserved" categories, and conserved in hominids vs. conserved in mammals vs. conserved in metazoans all mean different things. $\endgroup$ Oct 20, 2023 at 17:51
  • $\begingroup$ nonetheless, your intuition is correct that disulfides are relatively conserved. $\endgroup$ Oct 20, 2023 at 17:53
  • 2
    $\begingroup$ I am still not clear precisely what you want to do. Look at my answer and then, if it is not on target, comment indicating precisely why. $\endgroup$
    – David
    Oct 26, 2023 at 16:04
  • 1
    $\begingroup$ I am somewhat disappointed after the effort I put in to trying to answer your question you have not responded to it in any way. As I believe that we are both professional scientists I find that somewhat uncollegiate. $\endgroup$
    – David
    Nov 7, 2023 at 21:12

1 Answer 1


I am not quite sure what the poster is trying to do, in particular the question “is there a way to calculate the score between different disulfide bridges?”. Clearly, if different bridges are completely conserved between species in two different proteins or within the same protein, that’s it. (Sort of the converse of ‘slightly pregnant’.)
Nevertheless this answer discusses ‘scoring’ multiple sequence alignments (MSA) and describing the results, so it can be updated in response to any clarification the poster makes.

The Clustal MSA program already scores positions in an alignment
I do not know what MSA program the poster is using, but the output from Clustal X (at least standalone v. 2.1 for the Mac) already presents scores of a sort. This can be seen in a sample output, below:

Example MSA from Clustal X

The height of the individual bars in the bar graph below the alignment ranges from a maximum (where all five sequences in this set align) to what I assume is the zero baseline (where one of the sequences introduces a gap).
Closer examination, however, shows that the scoring system used is not a simple percentage identity: 0, 20, 40, 60, 80, 100%.

Detail of MSA

In the detail it can be seen that (4xK + 1xN) scores higher than (4xV + 1xK), and also gets a ‘:’ comment (see below). This is presumably because it is using the same PAM or BLOSUM amino acid comparison matrix used in the alignment to compute the height of the bars. (I leave the interested reader to plough through the literature to see if this is documented.)

What can one say?
The question seems to wish to transform the numerical score into a descriptive statement. The following assumes that one is taking one sequence as reference — here 1qjo.

• The simplest possibility would be a statement of the type:

“P at position 8 in the reference sequence is completely (100%) conserved in the data set, whereas at position 49 the conservation is only 80%, and at position 51 only 60%”. This could be converted to descriptions using the percentage scale in a subjective and arbitrary manner which one would have to define, e.g. >75% but <100% = ‘highly conserved’, >50% but <75% = ‘moderately conserved’

• An alternative would be to use the descriptive categories — * : . — and convert them to descriptive terms ‘completely’, ‘highly’ and ‘moderately’. This result of this would be different from using percentages. What one would really be describing here is not the conservation of a particular amino acid, but the conservation of amino type at a particular position. So the statement would be something like (choosing a different example):

“As regards the eight positions in the reference sequence where E was found in the data set, only position 40 was highly conserved and position 58 moderately conserved.”

Context matters
My example statements above all refer to “the data set”, emphasizing that any numbers or descriptions cannot be absolute, and their significance depends on the extent of the data set and whether it is biased towards sequences of a particular origin. The most obvious cause of bias is species e.g. a data set with 20 mammalian species and two insects. Not only species range is important, but it can happen that a problem is complicated by different forms of a protein in the same species, and that these forms may be more similar between species. A way of approaching the latter problem is to consult SCOPe, and if necessary construct a data set containing a range of SCOPe subcategories.

Other approaches to the disulphide bond problem
I am not quite clear whether any of the members of the family of proteins of interest to the poster lack cysteine at the position of the disulphide-bond forming cysteines of the references. If they do I would be tempted to use alphaFold to obtain predicted structures of these exceptional members and see if the positions at which the cysteines in the reference sequence occur are predicted also to be in close proximity in the exception.

  • $\begingroup$ I really appreciate your answer and I think it is a good start toward what I'm looking for. The two possibilities you offered sounds good. I'm starting to understand that it is not really easy (if possible at all) to get a number stating the degree of "conservativity". I know about the ClustalO/Mafft/T-coffee/etc. results and the way they show how is conserve a position. But if I understand it correctly, this states only how conserved the position is by itself, and not compared to others. $\endgroup$ Nov 8, 2023 at 9:19
  • $\begingroup$ I am not sure, if this is possible or even make sense at all. I'm interested though in trying to put a number to a specific position in the MSA compared to other positions. Or as @David put it. Is this one slightly more pregnant than the other one? After thinking and searching the problem, I also see, that a structural comparison might give better (or at least more conclusive) results, as to whether or not a postion is more conserved. $\endgroup$ Nov 8, 2023 at 9:19
  • $\begingroup$ The disulphide bridges I mentioned are a bit of a problem, because those Cysteines ** have to be there** in roder for the bride to exists. No cysteines, no bridge! Therefore they are always very conserved. This is why, even though the answer is a good one to decide how to regard the question, your two suggestions will not really be helpful here, as they Cysteine-pairs will have most of the times 100%. So by themselves they will always be highly conservative. But how is it comparing them to one another? $\endgroup$ Nov 8, 2023 at 9:19
  • $\begingroup$ So, maybe the question would be to see what around the bridges, maybe what is inside the loop the bridges create. If i can show that there are other, more unique AA there (e.g. W or Y), which might be needed for some kind a binding event, this might make this bridge "stronger conserved" than a bridge, where this AA is not available anymore, as the binding activity can't take palce anymore. I'll take a look into AlphaFold and see if I can make use of it. thanks $\endgroup$ Nov 8, 2023 at 9:20

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .