About PEACOAT

This paper in the Journal of Integrative Bioinformatics presents a Prediction-Enhanced Algorithm for Clustered, Ordered Assessment of Targeting in miR-mRNA interactions called PEACOAT. Please read the paper for details about the algorithm.

Using PEACOAT

WARNING: This code is messy and WILL be difficult to use. I am a mathematician, not a software developer, and I have put significant time into making this code work as well as possible. However, I am sure that it won't work the first time for everyone who tries it. If you are having difficulties after reading this documentation as well as the referenced description files, feel free to contact me with details about the problems you are having. Likewise, if you would like to contribute to PEACOAT and help make it better, I would greatly appreciate it.

This Git repository contains code for PEACOAT that can reproduce the results in the above paper on R version 3.0. Many packages are required, but for now I won't list them here; they will be obvious upon running the code.

To reproduce the results on the Myeloma data referenced in the paper, use the R script

./myeloma/script-myeloma.R

To simulate results as in the paper, use the R script

./simulation/script-simulations.R

with the parameter geo.files set to the appropriate paths to the dowloaded files from GSE17498.

Note, however, that one should be careful using these scripts, because

• they can take quite a long time to run
• there are many data file dependencies

So, please be sure to read and understand the code before running the scripts. I don't recommend running the entire script at once until you have run the code in sections at least once.

User-defined parameters

The file

./myeloma/README_userDefinedParameters_myeloma.R

and its counterpart in the simulation/ folder contain all of the user-defined parameters contains the configuration settings that a typical user might need. This is a fairly complex file and attention is needed to select the proper parameter values. A lot of comments have been included in the file for this reason.

Parameters must be set according to the expression data type that will be used. The following data types have code that supports them:

• GEO
• expression matrix
• Agilent
• Affymetrix
• simulation

Each may require certain R packages.

Among the many user-defined parameters, there are two that warrant special mention:

• typefollows.generator
• typeadder.script

Each of these is a path to a filename that handles the types within the expression data. For example, a data set might have three stages:

• embryo
• youth
• adult

and we would like to consider the stages in their natural order. To accomplish this, we would either create the variable typefollows to reflect this order (see the parameter file for examples) or we would write a script that generates the variable. The parameter typefollows.generator should be set to the name of this script. Often in simple cases, a script is not necessary, but in more complex cases it can be helpful. Likewise, the parameter typeadder.script should be set to the name of a script that adds the types to the expression object (in the targets slot), and is necessary only if the types are not present already.

See my previous paper in PLoS ONE for more info about types and orders.

File dependencies

Currently, starting from the repo directory---i.e. where the .git/ folder is located---I am using the following directories outside of the repo:

../databasefiles
../geo-data
../generatedfiles

The databasefiles/ contains

predicted/
    miRanda/
        hg19_predictions_S_C_aug2010.txt
        mm9_predictions_S_C_aug2010.txt
    targetscan/
        miR_Family_Info.txt
        human/
            Conserved_Site_Context_Scores.txt
        mouse/
            Conserved_Site_Context_Scores.txt

validated/
     miRecords_version3.csv
     miRWalk_validatedTargets.txt
     TarBase_V5.0.csv

where the files of predicted miR targets can be downloaded from the miRanda downloads page and the TargetScan downloads page for the organism in question.

The locations of the predicted files can be modified in the file

./src/loadDatabasePredictions.R

while the paths to the validated files as well as miR_Family_Info.txt can be nodified in

./src/modelDataInitializations.R

Generated files

The folder ../generatedfiles/, whose location can be changed in ./src/modelDataInitializations.R, is where intermediate, code-generated files are to be stored. Mainly, this is because processing the database prediction files and other initializations take some time, so the intermediate results are stored to save time on future script runs.

Note that the script checks for the existence of certain intermediate files, uses them if they are present, and generates them if they are not. Thus, to re-calculate a certain intermediate file, either delete it or re-name it.

Questions, concerns, and collaborators

Certainly I've skipped many details about the code here, but please feel free to contact me with questions or suggestions. I would also appreciate contributions, pull requests, etc.