BCF Home


Coiled Coil Domains

MARCOIL models and beyond

Mauro Delorenzi

The model for Coiled Coil Domains that of MARCOIL can be run also with MAMOT. The transitions probabilities in MARCOIL depend essetially on three parameters called i (initiation), t (termination) and r (ratio, between heptad and non-heptad transitions), as explained in the publication of 2002:

Delorenzi M. and Speed T., 2002.
An HMM model for coiled-coil domains and a comparison with PSSM-based predictions.

Bioinformatics, 18(4):617-625, 2002. Abstract

The two parametrizations that were used in the publication called H (high, 0.01) and L (low, 0.001).
In both models i=0.0001 and r=0.00001 (which is effectively almost like 0).
The model files for them are available here:
MAMOT model file for MARCOIL-H

MAMOT model file for MARCOIL-L

Model and Parameters
In the model file the architecture can be changed. Keeping the architecture of the model, the parameters can be retrained on new sequences, for example specialised for particular protein families. Transitions parameters can be retrained to vary the relative preference for short or long domains. New architectures can be built for Coiled-Coil Domains with non-heptad sequence patterns.

One feature that was added recently is the weighting of states during decoding. The MARCOIL models have no weights, respectively a weight of 1 for each state.

Discussion of coiled-coil prediction methods
Gruber M, Soeding J, Lupas AN., 2006.
Comparative analysis of coiled-coil prediction methods.

Trends Biochem Sci. 28(12):679-85, 2003.

In the image the coiled-coil Probabilities are coded by grey levels. Shown is region 140-162 of the protein Giardin-B from Giardia lamblia (Giardia intestinalis). The HMM model evidentiates well a very interesting progressive shift of heptad phase.

Discussion of coiled-coil structures
corresponding to sequence domains with non-heptad patterns can be found in the literature, for example

Gruber M, Lupas AN., 2003.
Historical review: another 50th anniversary--new periodicities in coiled coils.

Trends Biochem Sci. 28(12):679-85, 2003.

R code
A simple example R code script that shows how decoding results (posterior state probabilities) that are written by MAMOT to the file Pmatrix can be analyzed and plotted in R.

Generation of example CCDs
The parametes in MARCOIL were tuned for discrimination, not for generation of real-like CCDs.

For questions and comments:

Please send comments on web pages to bcf@isb-sib.ch