Editorial board editor-in-chief

KERAm: A NOVEl STAND-AlONE APPlICATION
FOR CORRElATED muTATIONS IDENTIFICATION AND ANAlYSIS
J
acek
k
uśka
1,4
, j
Acek
l
eluk
2,4
, b
ogdAn
l
esyng
3,4
1
 Faculty of Biology, University of Warsaw, Poland 
2
 Faculty of Biological Sciences, University of Zielona Góra, Poland 
3
 Faculty of Physics, University of Warsaw, Poland 
4
 CoE BioExploratorium, University of Warsaw, Poland
Abstract:
 
Keram is a stand-alone Windows 2000/XP/Vista application designed for the detection and analysis of the corre-
lated mutations. Study of this phenomenon provides important information about protein structure stability factors as well as 
the formation of protein complexes. It is generally assumed that the mechanism of compensation explains the mutations that 
occur simultaneously. Keram is designed to detect the mutational correlations by comparative analysis of multiple sequence 
alignments. Additionally a three dimensional structure can be applied to calculate the distance between correlated positions in 
the protein molecule.
Keram has been succesfully applied for the analysis of kinase subfamilies. The obtained data suggest that the mechanism of 
compensation does not explain utterly this phenomenon which seems to be much more complex and diverse. The residues 
that are detected as correlated are often placed at very distant positions of the protein structure, therefore the direct mutual 
interaction between them is impossible. We have detected not only correlated pairs, but also clusters of positions (even more 
than 10) that reveal correlated changeability. 
Keywords:
correlated mutations, multiple sequence alignment, protein-protein interaction
Background
Correlated mutations are the mutations occuring simultane-
ously at two or more positions and dependent on each other 
[Oliveira 
et. al., 
2002]. This phenomenon is presumed to be 
always concerned with the direct mutual interaction of the 
correlated positions [Neher, 1993]. It is assumed that cor
-
related mutations give important information about protein-
protein interaction [Valencia and Pazos, 2002] and/or by the 
compensation mechanism (Fig. 1). This theory involves local 
‘complementarity’ of amino acids in a protein structure. The 
current approaches of protein folding prediction on the basis 
of known amino acid sequence are aided by the detection and 
analysis of correlated mutations.
So far the correlated mutations has been succesfully detected 
and analysed in the study of kinase families, serine proteinase 
families and protein inhibitor families with the aid of another ap-
plication (CORM) developed by our group [Górecki 
et al.
, 2005]. 
At present the CORM results are veriied and conirmed by the 
Keram as an independent approach designed for the same 
purpose of theoretical protein analysis. 
Implementation
At present the multiple sequence alignment in raw text format 
is accepted by Keram, fasta format is not supported. Keram 
searches for pairs that reveal mutational correlations. 
Prior to analysis, the parameter of minimal variability must be 
declared for two positions suspected to be correlated. variability 
is deined as the number of different amino acids that occupy 
the corresponding positions in the multiple sequence alignment. 
Obviously the minimal possible value is 2. There is a possibility to 
exclude about 5 upstream and downstream positions in primary 
structure to avoid trivial results for the pairs that interact directly 
e.g. in alpha helices. It is also possible to apply three more ad
-
ditional ilters. The irst one is the ratio, which enables to search 
for pairs that show correlation phenomenon. The second ilter 
refers to the threshold value which is the minimal percentage of 
residues in analyzed positions that reveal correlation. The third 
ilter that should be applied is the maximum variability difference 
value. This ilter delimits the analysis to the positions in which
number of different amino acids are mutually different. After 
setting all the required parameters, the analysis is executed. 

---

Bioinf
or
ma
tics
72
Keram: a novel stand-alone application for correlated mutations identiication and analysis
Keram generates three types of output data. The irst one and 
the most essential is the list of correlations (Fig. 3). The second 
output ile is the correlation map (Fig. 4). This map is a graphic 
representation of the positions found to be correlated. The vertical 
and horizontal axes represent the sequence while the black dots 
mark the correlations. Keram prints also the data of distribution 
values of correlation for all 400 possible amino acid pairs (Fig. 5). 
These data contain average threshold and variability values for 
detected correlations. Keram is a user friendly application and 
does not require any advanced course to be applied.
Fig. 1.
A theoretical model showing the mechanism of compensation. Two residues that completely ill up a structural cavity must 
mutate simultaneously to preserve the structure. A) Two hydrophobic residues: leucine (large) and alanine (small) are simultane
-
ously replaced by two valines (medium size). The single mutation of the alanine into valine is ‘forbidden’ due to size disruption. 
B) Two charged residues mutate simultaneously to maintain the charge arrangement. The single replacement of aspartic acid by 
histidine is not favored. Simultaneous mutation of aspartic acid into histidine and arginine into glutamic acid maintains the charge 
arrangement and does not disrupt the protein structure.

---

73
Bioinf
or
ma
tics
Keram: a novel stand-alone application for correlated mutations identiication and analysis
Fig. 2.
Keram’s GUI
Correlated Mutations detected by Keram
Start of the computation: 2009-06-13 12:01
Exclude +/- 0 residues, Threshold: 80%
Minimal variability of the 1st residue: 3
Minimal variability of the 2nd residue: 3
1 | 21 | 6 | L | 35 | 4 | G | 80%
2 | 21 | 6 | L | 36 | 5 | L | 82%
3 | 21 | 6 | L | 48 | 4 | M | 80%
4 | 21 | 6 | L | 57 | 3 | P | 82%
5 | 21 | 6 | L | 59 | 3 | G | 82%
6 | 21 | 6 | L | 61 | 3 | E | 82%
7 | 21 | 6 | L | 71 | 4 | G | 80%
8 | 21 | 6 | L | 72 | 3 | G | 82%
9 | 21 | 6 | L | 76 | 3 | R | 82%
10 | 21 | 6 | L | 122 | 6 | F | 80%