Golang scripts for analyzing data at ProHits-viz.

• librsvg
• nestedcluster, only required when using biclustering option for clustering

go install ./...

We consider screen data to consist of a series of "conditions" that could be, for example, experimental timepoints, treatments or proteomic baits. For each condition there are "readouts" with associated data for the condition. Readouts will typically by genes or proteins. Readouts have an "abundance" that could be a gene expression value, spectral count or peptide intensity. The abundance is always assumed to be a non-negative number. Finally readouts have a "score" which is an indication of the confidence in the abundance.

All of the tools use this nomenclature. Input data files are expected to be in tabular format with a row for each data point, with each data point having a value for the condition, readout, abundance and score. Files can have additional columns and they will be ignored.

A sample file for testing can be found in sample-files/analysis-file.txt

Use a json file to specify all analysis settings. The type field is used to specify the type of analysis being performed.

pvanalyze --settings="settings.json"

Settings file format

{
  "fileList": ["file1.txt", "file2.txt"],
  "primaryFilter": 0.01,
  "type": "dotplot"
}

¹ Options: blue, green, grey, red, yellow

Data can be transformed prior to analysis:

1. Control values can be subtracted from readout abundances
2. Abundances can be adjusted to the "length" of a readout. For example if a readout is a protein, the abundance can be adjusted to the protein length, with abundances from smaller proteins increased and adundances from larger proteins reduced. Specifically the median readout length is divided by each readout's length to calculate the multiplier to use for adjustment.
3. Abundances can be normalized by condition, either by 1) total condition abundance or 2) a specific readout. For 1) the total readout abundance is summed for each condition and then each condition's readouts are normalized relative to the median of these sums. For 2) the median abundance of a specific readout is used to normalize all other readouts between conditions.
4. Abundances can be log transformed

To create images with conditions in a specific order, set conditionClustering to "none" and specify a list of conditions in the order you wish them to appear. If conditionClustering is set to "conditions", conditions will be hierarchically clustered.

You can control what conditions and readouts are shown on the image by setting both conditionClustering and readoutClustering to "none" and supplying lists for each. Alternatively, if you only what to specify a list of conditions, set readoutClustering to "readouts" and all readouts will be included in the analysis and they will be hierarchically clustered.

go test