Jensen-Shannon Divergence (Capra and Singh, 2007) scored with substitution matrix for conservation score

Question

[Jonathan Lefebre]

Dear all,

I have a question regarding conservation scores. In a well-known paper from 2007, Capra and Sing (https://doi.org/10.1093/bioinformatics/btm270) proposed the Jensen-Shannon divergence as a score for sequence conservation. They also compare different conservation scores that include substitution matrices and property based scores.

Now my question is, does it make sense to score the Jensen-Shannon divergence additionally by a substitution matrix or a property-based grouping of amino acids? I know that there is a score that does this with the relative entropy (Kullback-Leibler entropy, Williamson, 1995).

If no what would speak against it?

Best,

Jonathan

Answer 1

[Mensur Dlakic]

It depends on what exactly you are trying to achieve. For highly conserved residues, it won't make much of a difference whether one uses matrix-based scoring or not. For a paper and a program with various matrix-based conservation options, see below.

• https://pubmed.ncbi.nlm.nih.gov/11524371/
• http://prodata.swmed.edu/al2co/al2co.php

[Jonathan Lefebre]

Thanks Mensur!

I would like to find how conserved calcium binding motifs are. So I thought it would be great to have a score that incorporates properties like acidity etc. From the literature it seems that entropy-based scores perform very well and they are easy to implement, e.g. in a python script.

But my question was rather in the direction, whether it makes sense to additionally score a relative entropy-based score against a substitution matrix. I was wondering why no one has ever done this. Only thing I have read so far is the property grouping for those entropy-based scores, so I was wondering whether there is a statistical reasoning to why not to do this.