The calculation of this type of similarity score is built into all of the distance-based phylogenetic methods.  For example in protein-based neighbor-joining methods, all of the pairwise distances are calculated, based on an amino acid scoring matrix (Valine is more similar to Leucine than it is to Arginine for example).  From the resulting matrix of all pairwise distances, many analyses can be done, one of which is to build a phylogenetic tree.  The distances within the tree are then a good measure of diversity or conservation of the alignment used to build the tree.   For one example, the average distance from tips to root of the HIV-1 M group polymerase (protease, reverse transcriptase, RNAse, integrase) protein tree is far less than the average distance from tips to root of the HIV-1 M group envelope protein tree, because the envelope is driven to diversity by the human immune system, while Pol is driven to remain highly processive in its enzymatic functions, and is more hiidden from immune pressures (CTL attacks, but antibodies do not).

Your example of a 10 amino acid motif would not be good for accurate phylogenetic reconstructions of a history of much evolution.  But the scoring methods used for longer sequences would still be highly useful for scoring such short motifs.  The software and many "flavors" of the 20 by 20 (each amino acid distance to each other amino acid) scoring matrices (PAM250, BLOSUM, etc) are all available and validated on hundreds of data sets.

If I understand your question right, you are basically looking to quantify sequence motif conservation?

You could look at the positional Shannon entropy in your MSA and then use this as a measure of the degree of conservation of the motif? E.g. if at a given position, you have 'AAAAEEEEED', then the positional Shannon entropy would be [Reference: See link]:

H(p) = -SUM[ p_a*log2(p_a) ] = -[ 4/10*log2(4/10) + 5/10*log2(5/10) + 1/10*log2(1/10) ] = 1.36

Complete conservation of all 10 positions yield H(p) = 0. If you then sum the Shannon entropy over these 10 positions, this number would increase as the positional variability increases and thus quantify the degree of conservation. Maximum variability on all 10 positions would yield H(p) = 43.22*

This way you could quantify the degree of motif conservation as being a fraction of maximum variability, i.e. degree of conservation: DOG = [ H(p)_maximum - H(p)_observed ] / H(p)_maximum. If one position is then maximum variable (all 20 proteogenic amino acids are observed with equal probability) and the remaining 9 position completely conserved (only one amino acid residue present), then: DOG = (43.22 - 4.32) / 43.22 = 90%

Just an idea... Hope it helps,

Leon

*20 proteogenic amino acids, hence for ONE MSA position:

H(p)_max = -[ 20 * 1/20 * log2(1/20) ] = 4.32

http://www.cbs.dtu.dk/courses/27685.imm/reading/Ch4.pdf