Computing residues per turn in an alpha helix

Single mutation to be modeled

I recently had an interesting challenge, that of creating a molecular model of a protein expressed from a gene with a newly identified single amino acid mutation. I (wrongly) assumed that it would be a “simple” 3D homology modeling problem, solved by common methods. It turns out that the protein in question is in the Keratin family that forms a hetero-duplex coiled-coil structure.

Coiled coils

It was great to learn about the supercoiling of neighboring alpha helices, predicted independently cerca 1953 by Linus Pauling at Caltech  naming these “supercoiled structures compound helices,” and Francis Crick at Cambridge who called them coiled coils, the name that stuck.   With the foresight of supercoiling twistings, Crick’s key insight was the prediction of side-chains interlocking along the core of the superhelical structure, repeating the same interactions every seven residues (or two turns of the α-helix). His parameterization predicted a sequence periodicity of 7/2 with interlocking of sidechains as ‘knobs’ into ‘holes’. One of the fascinating features of coiled coils is the heptad periodicity predicted by Crick. Amazingly this  was confirmed more than 20 years later when Stone et al. in 1975 published the sequence of tropomyosin. The knobs-into-holes packing was observed with the publication of the structure of influenza hemagglutinin by Wilson et al. in 1981. (Summarized from Review by Lupas et al., 2017.)

Predicting software

The review of Lupas et al. lists all software that was created for the prediction of coiled coils from sequence as both structure prediction probabilities and heptad position. Most of the software can be found as web servers. But there are also software that can predict or analyze  existing 3D structures or models (CCBuilder, CCCP fit, Socket.) A server link list is available below.

The web server provide a detailed analysis of 3D models with a vast amount of various measurements, some of them rather abstract and difficult to visualize. The “one thing missing” while using these web server there was to find a measurement of the number of residues per turn (rpt.) I wanted to check if the CCbuilder models had an rpt similar to those in PDB structures. The book description seemed to imply that an alpha helix would have an rpt=3.6 as described in text books, while coiled coils could have a slightly more relaxed rpt of perhaps 3.5 which is 7/2 from the standard Crick description.

After some search I found just one web site that could provide this answer from the early CCBuilder paper: the HELANAL web server. This unique server is located in India and sometimes does not work well. In particular it is required to provide the “HELIX” records within a submitted 3D PDB-format structure file, or provide such data in difficult to understand format by people not familiar with FORTRAN, the programming language used to create the software.

The visualization software UCSF Chimera can provide the HELIX records. I also discovered that the executable binaries for HELANAL were available for both Windows and Linux within the web servers. These work in the same way but being a local software avoids web server time-out, perhaps due to geographical distance. In order to use the Linux executable I created a Docker image which is available on Docker hub as: helanal-ub20.

Implementation

Creating the HELIX records: assuming that this is the only model within Chimera the model number would be #0. The command to create the records ksdssp #0 and the file can be resaved with the records included with a command in the file pdbrun nouser all cat > ~/Downloads/saved.pdb.

The file can then be used by HELANAL providing that some other specific formats such as the length of the file name, or the avoidance of the _ symbol within the file name.

Results

Values for coiled coil models created with CCBuilder 2.o were typically rpt=3.59 (not shown.)

The table below shows HELANAL measurements from PDB structures. Most are in the ~3.62 range and the standard deviation means that there is ample room to accommodate variations that would include the coiled coil structures.

Indeed Woods et al. (2014) analyzed thousands of high resolution structures with HELANAL . They confirmed the text-book value of rpt=3.60  for an unstrained alpha-helix.  For coiled coil they found rpt=3.62. But the 2 distributions were indistinguishable by Student t-test. They conclude that “models returned with n = 3.65 ± 0.07 should be considered as models with good backbone geometry for coiled coils and helical assemblies in general.”

Docker image

Docker image is available on Docker hub as: helanal-ub20. It is based on Ubuntu 20:04. The compressed image is only 55.6 Mb.

It was constructed so that the default user is a “regular” user, but with the ability to sudo commands as administrator without the need of a password in order to simplify the addition of other software (with apt-get install.) A few utilities have already been added to the default Docker minimal version: nano text editor, wget web downloader, and mkdssp (man pages: mkdssp or mkdssp).

The mkdssp program was originally designed by Wolfgang Kabsch and Chris Sander to standardize secondary structure assignment. DSSP is a database of secondary structure assignments (and much more) for all protein entries in the Protein Data Bank (PDB)

The HELANAL manual in PDF form as well as HTML and text format derived from the server web site are included in the default user named helanaluser.

To use this image sharing a directory from the host computer: cd into the directory you want to share, then:

docker run -it --rm -v ${PWD}:/data jysgro/helanal-ub20

• Helanal will not work properly if the file name is too complicated or too long. Avoid using - or _.
• One easy way is to create a symlink with ln -s to create a short nam without the need to rename to file.

Useful web servers

Structural building and/or analysis servers

Sequence-based analysis server

This server has an implementation of multiple sequence-base coiled coil prediction software: Marcoil, Multicoil, Multicoil2, Ncoils, Paircoil, Paircoil2, and oligomerization prediction: Scorer 2.0, PrOcoil, LOGICOIL.

A helical wheel as well as a helical net are also provided by the web outptut. A command-line version is available but without these graphical extras.

A web search could provide alternate server for most of the these predictors, but this site offers them “all in one” place.

References

Historical references

Server references

• HELANAL-Plus: a web server for analysis of helix geometry in protein structures

Simm D., Hatje K. and Kollmar M. (2015) Waggawagga: comparative visualization of coiled-coil predictions and detection of stable single α-helices (SAH domains). Bioinformatics. 31(5):767-769. – 10.1093/bioinformatics/btu700