TMAinspiration: Decode interdependencies in multi-factorial tissue microarray data

Supplementary information

Boecker F, Buerger H, Mallela N, Korsching E.
TMAinspiration: Decode Interdependencies in Multifactorial Tissue Microarray Data.
Cancer Informatics. 2016;15:143–149.
doi: 10.4137/CIN.S39112

Note: The binaries are developed for the Debian/Ubuntu Linux (64) platform. They are compiled with the option 'static', so almost all of the dependencies should be fulfilled. In the strange case that you get a 'missing dependency' error try to use 'ldd FILE' (where FILE is the executable you try). The output might point to missing libraries/binaries which have to be installed on the computer system. Additionally to that the Intel runtime libraries for the Fortran binaries are provided. We welcome feedback.


Best practice

Read carefully the step by step tutorial of the software and try to understand the algorithm. Otherwise it is very likely that you might draw wrong conclusions.
Create a work folder in your home folder. Download the required parts from the download sections below and place these items into the work folder. Create again a subfolder in this folder called 'results'. So the folder structure should now conform with the Figure 1 on page two of the description. If not already done, open a terminal and 'cd' into the created work folder. Execute:

./tins_s_omp ./data.txt ./data.mapping.txt ./results/tins_s.result ./results/tins_s.log 10 b 6 1

If the basic requirements are fine you will get something like that:

Finished job on 1 thread after 0.025 seconds

indicating a successful run. Additionally you will find two new files in the subfolder 'results'.

Note: It is always advised to perform an appropriate number of resampling runs to prove the stability of the result. In this context the combinatorial space of the raw data has to be considered.



1) Download the command line binaries, the data and a tool description.
Step by step tutorial including an example of use.
Test data and essential mapping scheme for the test data.
A bash script file with several examples how to use the command line executables.

2) Find an optimal interdependency solution for the selected partitions and test the quality of the solution:
tins_s_mpi (3.9 MB) : selection of a specific partition and resampling test (MPI version).
tins_s_omp (3.2 MB) : selection of a specific partition and resampling test (OpenMP version).
Explore all partitions of a certain size:
tins_mpi (3.9 MB) : selection of a specific partition and all combinations thereof (MPI version).
tins_omp (3.2 MB) : selection of a specific partition and all combinations thereof (OpenMP version).

3) Reengineering
Source code.

R code for further analysis

1) R scripts - including a function to import raw data (tins.import.R), a function to create the dependency graph (tins.plot.dependencies.R), two functions to create the quality plots (tins.ref.size.impact.R, tins.sample.size.R) and a very basic example of use (tins.examples.R).

2) Fortran library for use with R to calculate the Pearson correlation table like the Fortran programs do. In this case you will need additionally the Fortran run time libraries - unzip and put all into the R work folder.

3) A .RData workspace file containing the test data and results for testing purpose.

Links to the R Project for Statistical Computing and MPICH

1) R - information and download.
2) MPICH - information and download - one implementation of the Message Passing Interface (MPI) standard.