PdbTool: An object-oriented Julia tool to parse PDB files and work with them
Author christophfeinauer
7 Stars
Updated Last
2 Years Ago
Started In
January 2015

PdbTool: An object-oriented Julia tool to parse PDB files and work with them


This is an object-oriented Julia tool to parse PDB files (http://www.rcsb.org/) and work with them.

In this moment it can parse coordinate information of amino acid chains, RNA chains and secondary structure information. All information is stored in hierarchical Julia types and easily accessible. Some functions are implemented to work with the data, e.g. calculate residue distances with different distance measures.

A backmapper is implemented that allows to identify residues in the PDB with residues in a multiple sequence alignment if the Hidden-Markov model is available. This identification is stored in the PDB object and can be used to directly work in the alignment representation, e.g. to calculate residue distances between alignment positions.

If you're only interested in a mapping HMM <-> PDB and residue distances

If your only goal is to get a mapping between an HMM and PDB residues (and optionally distances between the mapped residues), there is no need to use Julia or the module directly. For this, a simple-to-use script called backmap.jl is included in the repository. After installation of Julia and the module, call

julia backmap.jl -h

from the command line to get usage instructions. The typical use case for this is if you want to get distances between positions in a Multiple Sequence Alignment made by the HMM.


The parser should not have any dependencies, but the backmapper expects to find hmmalign and (for the function mapChainToHmmLegacy if you want to use it) hmmsearch on the path. Furthermore, if you want to backmap RNA chains you need cmsearch as well.

Everything has been tested with HMMER 3.1b1, and 3.1b2 (see http://hmmer.janelia.org/ to get the newest version).


v0.1.0 can be run with julia 0.4 and julia 0.5 v0.2.0 now supports julia version 0.6 v07 supports julia version 0.7


To install the package under version < 0.7, use the command


To install the package under version >= 0.7 please use the new package manager (it can be activated from the REPL using the key ])

(v0.7) pkg> add https://github.com/christophfeinauer/PdbTool.jl #v07

Alternatively, clone/download the repository and do a


whith REPO_DIR replaced with the direcrory you download the repository to.

Test the installation

To test some major functions like the parser, the backmapper and the accessability of external functions, first run

julia>using PdbTool

and then



A real documentation is not available yet, but here are some usage examples to get you started:

Load the module:

julia>using PdbTool

Now let us parse the PDB file 5PTI.pdb. You can find it in the ./test directory of the repository. Exchange REPO_DIR with the directory of the repository in the following command and run it:


The object pdb contains collections of Chain objects, which contain collections of Residue objects, which contain collections of Atom objects.

To get the Atom object labeled with "N" in residue "10" of chain "A" type


Objects of type Atom have two values. The first, here 154, is the numerical identifier of the atom in the PDB file. The second are its coordinates. To get them as a tuple write


Notice that the key for the residue is a String while the identifier for the atom is a Int64. This is because in PDB files consecutive residues are often labeled e.g. with 10 and 10A.

To calculate the distance between the atom "N" and the atom "CA" in the same residue, type


To get the distance between residue "10" and "20", type


where as default the distance between the two closest heavy atoms in the residues is taken.

To map the chain to the Hidden-Markov model Kunitz_BPTI.hmm you find in the test directory type


The residue "10" in chain "A" now has been identified with a position in the Hidden-Markov model:


Residue "1" on the other hand could not be mapped and has retained the default alignmentPos value of -1


After the mapping, chain "A" has a non-empty collection align of length 53, containing Residue objects. This is the inverse information to the alignmentPos values of the residues: The residue corresponding to position 7 in the Hidden-Markov model (or the 7th column of the corresponding multiple sequence alignment) should be the residue labeled "10".

The identifier field of a Residue object gives the label:


Notice that residues with alignmentPos=-1 do not show up in the align collection.

To get a distance matrix for the columns of a multiple sequence alignment corresponding to the Hidden-Markov model you now just have to type

julia>ind=sort([k for k in keys(pdb.chain["A"].align)]) # Collect indices
julia>align=pdb.chain["A"].align; # Alias "align" 
julia>[PdbTool.residueDist(align[k1],align[k2]) for k1 in ind, k2 in ind] 
53x53 Array{Float64,2}:
 0.0      1.33243  …   3.68234   8.16824
 1.33243  0.0          2.04305   5.68643
 2.60381  1.29087      5.97459   7.68815
 1.99134  2.83832      6.94746  10.9973 
 5.73631  6.12322     10.1668   13.4135 
 ⋮                 ⋱                    
 7.65084  7.58533      3.44798   3.88685
 2.14929  3.52895  …   1.33761   3.46812
 3.68234  2.04305      0.0       1.27819
 8.16824  5.68643      1.27819   0.0