<tool id="Find peaks" name="PeakPicker" version="1.0.0">
  <description>Finding Peaks in a GFF Nimblegen File</description>
  <command interpreter="perl">PeakPeaker2.pl --in $input --out $output --t $type --dist_peaks $dist_peaks --col3 $col3 --log $log --perc $perc --num $num --dist $dist --w $window --f_pv $output2</command>
  <inputs>
    <param format="tabular" name="input" type="data" label="Source file"/>
    <param name="col3" size="20" type="text" value="Analisys" label="Analisys name"/>
    <param name="type" type="select" label="Analysis type">
      <option value="p">p-value</option>
      <option value="s">score</option>
     </param>
    <param name="perc" size="4" type="text" value="0.95" label="percentile value"/>
    <param name="log" size="2" type="text" value="7" label="-log p-value cutoff"/>
    <param name="num" size="2" type="text" value="3" label="minimal number of probes"/>
    <param name="dist" size="4" type="text" value="100" label="max distance between two probes"/>
    <param name="dist_peaks" size="4" type="text" value="200" label="min distance between two peaks"/>
    <param name="window" size="4" type="text" value="500" label="window length"/>
  </inputs>
  <outputs>
    <data format="bed" name="output" />
    <data format="gff" name="output2" />
  </outputs>


  <help>
 .. class:: infomark

**What it does**

This tool utilizes NimbleGen ratio files in gff format as INPUT FILE and provides a table of the computed peaks in the same gff format.

--------

**Parameters:**

- **Analysis type:** 
                     - **p-value** analysis performs peaks determination based on p-value inference
                     - **score** analysis performs peaks determination based on a scoring system
- **Percentile value:** it is used to calculate the threshold rate based on dataset distribution to filter out background
- **-log p-value cutoff:** (required only for p-value based analysis) cutoff integer to be used to identify a significant peak
- **minimal # of probes:** minimal number of consecutive probes used to define a peak
- **max distance 2 probes:** greatest nucleotide distance (bp) between two probes that allow to consider two probes as adjacent
- **min distance 2 peaks:** minimum nucleotide distance (bp) required to consider two peaks as separate entities
- **window length:** length in bp of the window used for statistical analysis

--------


**INPUT FILE**

Nimblegen gives you back a GFF file with the coordinates of each probe and the normalized signal value --> log2(Cy5/Cy3).

Click here_ to download a GFF file example.

.. _here: /static/example_file/GFF_file_norm.txt.zip

Example of Nimblegen GFF format::

    chr19  Nimblegen  tiling_array  100000  1000051  -1.2	 +   .  probe_name
    chr19  Nimblegen  tiling_array  100100  1000151   2.9	 +   .  probe_name

.. class:: warningmark

The sixth column **must** contain the normalized log2(cy5/cy3) that Nimblegen gives you back after the experiment


---------

.. class:: infomark

**How does it work?**

**Two assumptions:**


- data are enriched for signal in the positive direction ("one-tailed")
- a peak (or enriched region) is represented by multiple probes that are genomically located close to each other


**Statistical approach: sliding window**


A window centered at each probe of the array moves probe by probe. In each window Chi squared is calculated


.. image:: static/images/CARPET/chi_squared.png


by building a contingency table for each probe, and a p-value is assigned


.. image:: static/images/CARPET/centered.png


**"-log2(p-value)"** is associated to each probe. This value takes in account the neighbouring probes effect.
This approach dramatically decreases the background signal.


.. image:: static/images/CARPET/background.png


New values are considered to defined an enriched locus


.. image:: static/images/CARPET/pvalue.png


Moreover a score is calculated taking into account the length and the raw signal of the peak


.. image:: static/images/CARPET/pvalue_score.png


Output is a gff file


.. image:: static/images/CARPET/table_pv.png


**NON Statistical approach: score**


Only the raw signal of each probe is considered. Only the regions with a number of consecutive probes above the defined threshold are selected


.. image:: static/images/CARPET/score.png


Output is a GFF file


.. image:: static/images/CARPET/table_score.png


and a GFF file with the p-values associate to each probe

 </help>
  
</tool>