Etchevers:Notebook/STRA6 in eye development/2009/07/30
|Genetics of human eye development||Main project page|
Previous entry Next entry
FindPeaks update and questions
Anthony has sent the following messages for each update to FP4 from which I've extracted what I think relevant to us:
* Bugfix: only write out filtered peaks if not (index == -1). (e.g. there is a peak there.) * Bugfix: Retrieving control threshold fails when dataset is empty. Now returns a threshold of zero. (Bug identified by Deniz Koellhofer) * ApplyCompare can now write out a wig for just filtered peaks.
* Nothing special for us.
Version 4.0.3 provides bug fixes for several advanced features including a more appropriate window size for * peak pairing in compare/control mode, * fixes incorrect pairing of chromosomes under certain conditions for control/compare, * fixing correct minimum height when writing out files.
This version also now adds the ability to provide jpg/png graphics of the compare/control results being processed.
* Bugfix: Incorrect variable while writing out mapfiles. * A simple library for producing graphs within the context of FindPeaks. Supports PNG and JPG. * Replacing the overly specific variable pair object with a more generic model: Tuple<Object,Object> * Upgrade all FindPeaks Modules to use Tuple<> instead of IntPair, FloatPair, FloatInt objects. * Update Build.xml to include Tuple.class * First Pass build of HistogramImage creating object. Will take any histogram and generate a histogram image. * minor improvements to image libraries. * Adding SVN:Keywords to new objects * Bugfix: writing out files was dropping the minimum height requirement due to the incorrect use of a gt or eq, instead of just gt * Bugfix: applycontrol parameters should be sent the whole dataset, which will require normalizing for proper statistics. * LinearRegressionPerpendicular now exposes slope for use in higher up normalizations. * Apply Control now requires a normalization for significance calculation. This replaces the previous method of * calculating p-values for the peaks in the peak files (when available.) * FindPeaks now makes appropriate calls to get Normalization factor and utilize it (when available) for writing out peaks files. * PeakDataSetParent uses normalization and passes it along to write out peaks files. * Change default window_size to 400. 100 was WAY too small. [We used first the default 100 then tried 50] * Significantly upgraded the graphics produced by graphimage. Applycompare now uses them. * Update build.xml to include graphic generating files used for compare scatterplots. * Added ability to generate lines to graph image (based on slope and intercept) * ApplyCompare now also writes lines to image * Producing two jpgs: before and after filtering, for greater clarity. * Flip height and width in basic image init, to conform with the rest of the graphics package standard notation. * New Error checking on AlignedRead direction. Must be "+" or "-"
So it looks like we should install version 4.0.3 (possibly here?)
Will want to ask why .wig files are so bloody enormous that we have to upload them to UCSC from a server to be efficient, and why UCSC only takes five? (Unpacked, they were 3.6-3.7 Gb each, just for one chromosome.) Also, once the .wig files are up, what is the unit on the y axis? We are getting between 0 and 1, which is odd.