Valdar WSJ (2002). Scoring residue conservation. Proteins: Structure, Function, and Genetics. 43(2): 227-241.

To go and use The Scorecons Server, click here.
 

Contents

• Walkthrough examples
• Example 1
• Example 2
• Scoring methods
• summary of scores
• entropy scores
• entropy with 21 types
• entropy with 7 types
• mdm scores
• basicmdm
• valdar01
• trident score
• Score parameters
• Fractional ranks
• Substitution matrix
• Matrix transformation
• Gapphilia
• Trident score parameters
• Tips

Walkthrough example 1

1. First, we need to have our alignment in a format The Scorecons Server can read. It must be in either FASTA format (eg, ike this) or PIR format (eg, like this). Most multiple alignment programs can be made to write to at least one of these formats.
2. Tell The Scorecons Server where your alignment file is using the "Browse" button in the "Alignment input" box.
3. Skip to the "Action" box at the bottom of the page and press the "Score conservation" button.

1. The conservation score. This ranges from 0 for unconserved to 1 for highly conserved.
2. The focus residue identifier (see later).
3. The column of residues from which the score was calculated. So it can fit on one line, this column is shown on its side.

The first few columns are almost entirely gaps, and so have the lowest possible conservation. The last column shown contains only very similar types of amino acids and no gaps. Its score is correspondingly high.

This alignment contains lots of information. Almost every position score is different. This diversity of scores is recorded in the "Diversity assessment" part of the results. Diversity of position scores (or "Dops") considers the number of different scores in the alignment and the relative frequency of each score. The Dops for this alignment is 95.2%. If no two positions had the same score, the Dops would be 100%. If all positions had the same score, Dops would be 0%. For example, the alignment below of the PDB sequence 1af5 does not contain much information.

All the sequences are virtually identical, as are the conservation scores. The Dops for this alignment is 8.3%, reflecting this.
 

Walkthrough example 2

Suppose we are interested only in the cow alpha chain (Cow_alpha_Hb above) and how conserved its residues are. In this case, we can get The Scorecons Server to score only those columns in which the cow alpha chain has a residue.

1. As before make sure the alignment is in the right format and use the "Browse" button to locate the appropriate file.
2. In the "Output options" box, click the "only positions relating to sequence" option.
3. Cow alpha haemoglobin is the 2nd sequence in the alignment. Since The Scorecons Server treats the top sequence as sequence "0", the cow alpha haemoglobin sequence is therefore sequence "1", so type "1" in the associated box.
4. Click on "Score conservation" in the "Action" box.

0.320           M              MMMSV-VVMM--
0.278           V              VVVDHMHHGGG-
0.500           L              LLLYLLWWLLL-
0.603           S              SSSSTTTSSSS-
0.391           A              PAAAPAAEDDD-
0.424           A              AAAAEEEVGGQT
0.809           D              DDDDEEEEEEEE
0.453           K              KKKRKKKLWWWW
0.409           G              TGNASAQHQQQE

Notice information is given only positions that are non-gap for cow alpha haemoglobin. Also, notice the middle column. When we scored every position in the alignment indescriminantly, there were only "#"-signs in this column. Now this column contains the residue type of the target (or "focus") sequence. Reading the middle column from top to bottom gives the entire unbroken cow alpha chain sequence. The Dops is also slightly higher than in the previous example, at 95.9%. This is because it now measures the diversity of scores pertaining only to the cow alpha sequence.
 

Scoring methods

• Summary

Sum-of-Pairs scores sum all possible pairwise match scores between amino acids in an aligned column; entropic scores use Shannon's information theoretical entropy to measure the diversity of symbols (amino acids) in a column; matrix scores employ a substitution matrix to evaluate stereochemical diversity in a column; sequence weigted scores normalize against redundancy of sequences in the alignment.

• Entropy scores

Let

=number of residues in column

=number of residue types

=number of residues of type 

= 

Shannon's entropy is given by



Entropy scores used here normalize Shannon's entropy so that conserved (low entropy) columns score 1 and diverse (high entropy) columns score 0:



• with 21 types

Residues are classified into one of  types: 20 standard amino acid types + 1 gap type.
 
• with 7 types

Residues are classified into one of  types: aliphatic [AVLIMC], aromatic [FWYH], polar [STNQ], positive [KR], negative [DE], special conformations [GP] and gaps. This convention follows that of Mirny & Shakhnovich (1999, J Mol Biol 291:177-196).
• Mutation data matrix scores

Let

= normalized substitution matrix with range [0,1]

= the element  in the matrix 

= number of sequences in the alignment (equivalent to above definition of  )

= the  th sequence

= the amino acid at position  in the alignment that belongs to 
• Basicmdm score

Standard sum-of-pairs score.


 

• Valdar01 score

Score used in Valdar & Thornton (2001, Proteins 42:108-124):



where  is a scalar such that



where  is the weight of  such that



where  is the distance between sequences  and  , and is calculated as



where  is the set of all positions that manifest an amino acid in one or both of  and  , and  is the size of this set.
 

• Trident score

Trident scores on the basis of three parameters: symbol diversity, as measured by an entropy-related measurement (with sequence weighting), stereochemistry (ie, the diversity of physical and chemical properties in the column) and gaps. Read more about this score in William Valdar's PhD thesis.

Score parameters

• Fractional rank option

Click the fractional rank checkbox to show each conservation score as its fractional rank for the alignment. Eg, the most conserved position will score 1, the median will score 0.5 and so on. This artificially stretches the distribution of conservation scores evenly between 0 and 1. This involves a loss of information but may be useful for making vivid images of coloured-up proteins, since spreading conservation out in this way maximizes contrast.
 
• Substitution matrix choice

Some scores use substitution matrices to measure the diversity of residue types (see table of scores). The substitution matrix choice box lets you select which matrix is used. "Modified PET91" is the PET91 matrix of Jones & Thornton (1992) with diagonal (eg, A vs A) set constant.
 
• Matrix transformation choice

Before a substitution matrix can be used, it must first be transformed into a convenient range. A linear transformation is the simplest way of achieving this. The more sophisticated "Karlinlike" transformation ensures matches between identical amino acid types consistently score highest. When not using the modified PET91 matrix, prefer to use the Karlinlike transformation over the linear one.

Let
 

= original log odds mutation data matrix
= normalized matrix
• Linear:


 
 
• Karlinlike (after Karlin & Brocchieri, 1996, J Bacteriol 178:1881-1894):




where



• Gapphilia parameter

In sum-of-pairs type scores (see table of scores) gaps are considered outlying amino acid types in a substitution matrix. To penalize gaps, the match scores m(a,gap), where a is an amino acid, and m(gap,gap) are set to be low. This option allows the gap match score to set explicitly. Acceptable values are between 0 (gaps are strongly penalized) to 1 (gaps are favoured).
 
• Trident score parameters

Trident parameters range anywhere between 0 and infinity (although "infinity" will neither be accepted nor is it recommended).
1. Diversity: set the extent to which symbol diversity is penalized.
2. Chemistry: set the extent to which variability in amino acid physical and chemal properties is penalized.
3. Gaps: set the extent to which gaps are penalized.
By default, the values 1, 0.5 and 3 are chosen so that the trident score most resembles the valdar01 score.

Tips

• Although The Scorecons Server tolerates non-standard amino acids, such as B, Z and X, it does model them consistently. X is treated as if it were a gap. B and Z are present in substitution matrices and so are modeled in matrix scores; however, since they are artifacts of experimental uncertainty and have no biological meaning, they cannot be considered extra types in the entropy scores. Instead, entropy scores convert B and Z to gaps. Therefore, although a few non-standard amino acids shouldn't cause too many problems, for best results either avoid alignments containing them or convert them to standard amino acids or gaps.