HDP-Align

This repository primarily contains the implementation for HDP-Align, a hierarchical Bayesian non-parametric model that performs peak matching for liquid-chromatography mass-spectrometry data using a Hierarchical Dirichlet Process mixture model. Detailed description of the paper can be found in HDP-Align: Hierarchical Dirichlet Process Clustering for Multiple Peak Alignment of LC-MS Data by Wandy, et al. (2015).

Model prototypes in Matlab can be found in the AlignmentModel folder. The working Java implementation of HDP-Align is inside HDP-Align (click here for the primary class that performs the clustering), alongside a few other alignment methods used in the evaluation experiments. MultiAlignPipeline contains the pipeline used for loading feature data in FeatureXML or SIMA format, performs alignment on them and computes the performance measures (Precision/Recall as defined in the paper). Local copies of MzMine2 and SIMA used for the purpose of evaluation in the paper is also provided in this repository.

Experiments

The folder experiments contains all the experiments done to produce the results in the paper. Further details can be found inside.

Development Set-up

Only if you're a developer!!

Importing projects

1. Install the Maven and eGit plug-ins in Eclipse
2. Clone the repository https://github.com/joewandy/HDP-Align.git using either the git clone command or do it from within Eclipse. If you're doing it from the command line via git clone, be sure to add the repository to Eclipse as well (via Git Repositories view -> Add an existing local Git repository).
3. Go to the Git Repositories view in Eclipse and expand the Working Directories in the added repository. We want to add each project in the repository into Eclipse. Start with PeakML. Right click PeakML, select Import Project > Use the New Project wizard > Java Project > enter PeakML for the project name, change the default location to the git working directory (e.g. /home/joewandy/git/HDP-Align/PeakML), click Finish.
4. Repeat for the remaining projects too: mzMatch and cmdline are imported as existing Java projects. MZmine2, MUltiAlignPipeline and HDP-Align should be imported as existing Maven projects (use File > Import > Existing Maven Projects from the menu).
5. There will be compilation errors in most of the projects due to unresolved dependencies. See next section.

Setting up dependencies

To keep it simple, we will manage dependencies between projects using Eclipse.

1. PeakML and MZmine2 do not depend on anything.
2. mzMatch depends on PeakML and cmdline. Right-click on the mzMatch project from Package Explorer, select Properties, then go to Java Build Path, go to Projects tab, add PeakML and cmdline there.
3. MultiAlignPipeline depends on cmdline, HDP-Align and PeakML, so follow step (2) to add these dependencies.
4. HDP-Align depends on mzMatch, MZmine2 and PeakML, so add these project dependencies as well. Additionally, go to the Libraries tab and also add all the jars inside HDP-Align/lib.

Extra stuff

git config http.postBuffer 524288000

git remote set-url origin https://<user>:<password>@github.com/joewandy/HDP-Align.git</password></user>