- Master branch:
- Experimental branch:
I DON'T HAVE ANY IDEA WHAT TYPES OF PROTEINS THIS IMPLEMENTATION SHOULD BE USED FOR, THAT'S WHY I SOLELY RELY ON ARTICLES (which apply folding to globular proteins - for details see readme.md in scala_code subfolder). ALSO I HAVE NO IDEA ABOUT HOW MANY CHAINS THEY HAVE (currently there is 1 chain) OR AVERAGE CHAIN LENGTH (that's why SimplifiedAminoacid class is made as simple as possible). Other details are selected because of articles used.
In future everything might have changed - i.e. if I implement alascan, I'll probably handle proteins with many chains.
It would be good if you add your remarks/bugreports to Issues. Any type of code review will also be appreciated.
This repo originally was supposed to contain code for finding specificity determining regions (in Scala). But now it contains classical folding algorithm implementation+qhull implementation. Original idea was given by my scientific advisor and it is still unclear, the name of the repo left unchanged.
Currently this repo is not actively maintained.
There is still much to do - to move from folding to alascan, to refactor code, etc. Some TODOs can be found in scala_code/README.md in experimental branch. Some of them will never appear in master branch, they are pure experimental. It isn't my main pet project any more, that's why these changes won't appear tomorrow.
Some kind of current Scala classes diagram can be found here. It shows aggregation and composition relations between classes (that little scala project parser app is my experiment also, it's unfinished and was made to save this (I mean, THIS) project from code smells - it actually helped with local God class. I hope it'll help you to understand this project's structure better).
I decided to make something for my own, evil purposes - probably I will implement these in experimental branch only and won't merge them to master branch (not too fast, because nobody needs it except me):
-
in Julia: make scripts for parameters generation from PDB sources. Currently these parameters are taken from MCDP dataset, converted to JSON.
-
in Scala: make visualizer. There is rcsb viewer for proteins, I want to try to add optional visualization for folding process with that viewer.
-
in Scala: it would be nice to reimplement some moves or to make them optional.
-
in Scala: it would be nice to implement fragment-based moves (to understand them better).
-
in Scala: there is always option to make algo much more faster - by implementing kd-tree or using some sort of hashing. This article might be useful: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3576911/
MIT license for code (at first I thought I should use beerware-like license, but... dude, I prefer coffee to beer and most probably we won't meet => sad squirrel gets no free coffee, alas! Sir Roland to The Dark Tower came and said - "SOOOO WAT??". ok, I see, you don't get stupid jokes).
MCDP_dataset currently used in energy function parameters estimates might have different license. In future I plan to write scripts to recalculate these estimates from collection of .pdb files.
Some of todos can be found in scala_code/README.md (in russian). There is also option to add specificity determining residues guessing algorithm to old code in python (branch python_proto).
Branch bond-restore has slightly different implementation of backbone reconstruction: it restores atoms laying on bond, not belonging to aminoacid. Worths mention: bond atoms are counted for aminoacid type closer to C-terminus (because I looked at proline and thought that aminoacid would be reconstructed better in that case). Anyway, with that type of reconstruction appearance of blank intervals, closer atoms triangles is still possible. [Feig et al.,2000] states that this type of reconstruction errors is typical for lattice models. That's why I decided not to merge that reconstruction to master branch, but decided to keep it there.
Latest changes will appear in experimental branch, some of them will be merged to master, other won't.
All references used in algorithm implementation are given in scala_code/README.md.
It turned out that there is no information on pdb backbone vectors statistics, at least I couldn't find any. Method of obtaining such information is well-defined - that's why I added helper script for collecting such information from protein data bank files (in folder julia_scripts).
The latest version of code can be found at:
http://github.com/biocad/sdr_tools
Feel free to report about any problems and bugs via Issues page.