TiMAT User Manual
28 February 2007
Table of Contents
TiMAT is an open-source, Java based set of scripts used for processing chip-chip tiling array experiments. Its four main functionalities are (1) smooth noisy signals, (2) optionally calculate false discovery rates, (3) identify enriched intervals and (4) identify peaks within these enriched intervals. While parts of TiMAT might be used in analyzing expression arrays, other parts, such as the symmetric null-p calculations are not appropriate for such experiments-- TiMAT is designed for chip-chip experiments.
The computing idiom is mostly that of command line application. Generally, each step of the processing is encompassed by a separate application which will usually take several command line arguments. Almost always one to three of these arguments are simply paths to necessary data files. Sometimes additional flags allow setting of computational parameters (ex. changing smoothing window sizes) or optional algorithmic steps (ex. switching off quantile normalization). In addition to this user manual, almost every TiMAT application has significant on-line documentation that can be accessed by running the application without any command line arguments. TiMAT has been run, at least, on Linux, OSX and MS Windows. While any platform that supports Java and the R statistical programming environment should work fine, TiMAT has been developed on Linux and OSX and a platform that has a sane and well supported command line environment (such as Linux or OSX) is recommended. A number of bash shell scripts are included with the TiMAT bundles that should be useful with minor modifications on, of course, bash shells.
TiMAT processing starts with a chip vendor's aligned probe intensity matrix file, such as an Affymetrix .CEL file or Nimblegen .pair. TiMAT does not read the tiff or other image files produced directly by chip scanners. Currently, the software supports Affymetrix and Nimblegen chips with a framework that should be easily extensible to other chip vendors such as Agilent. Between this entry point and the final .gff, .sgr and .Report files, TiMAT uses a number of its own file formats. The important ones will be discussed further while considering the TiMAT processing pipeline.
BackgroundTiMAT has its development roots as a software prototype that started with strategies used with expression arrays, such as quantile normalization, and building from there. As such the code base evolved quickly with many ad hoc features in competition to produce the most useful and clean results. Now a few years and several iterations old, the program is beginning to stabilize and some refactoring has simplified the code. For the hacker, this prototyping will be evident in the underlying design as redundancies or vestigial features. To the end-user, one will find a number of applications bundled with TiMAT that are better or not-so-well supported depending on how much they've been used. This user manual mostly focuses on the applications used most frequently (and hence best supported) by the BDTNP. However, there may very well be useful code in the source tree not discussed here... users are encouraged to dig around. On that note, some features have disappeared during the evolution, though the TiMAT website usually maintains an archive of older releases. While the software is released as is, in general the goal is that the releases differ in features, not number of bugs! Older versions should work adequately and may have features not in new versions or visa versa. Be aware that versions are not always backward compatible.