<tool id="nanonet_1D" name="Nanonet 1D" version="2.0.0">

    <description>ONT development basecaller</description>

    <!-- ***************************************************************** -->
   
    <!--
    <requirements>
        <requirement type="package" version="2.0.0">nanonet</requirement>
    </requirements>
    -->

    <!-- ***************************************************************** -->

    <version_command>echo "2.0.0"</version_command>

    <!-- ***************************************************************** -->

    <command detect_errors="aggressive">
    <![CDATA[

    python3 $__tool_directory__/nanonet_1D.py $input $output \${GALAXY_SLOTS:-1}

    ]]>
    </command>

    <!-- ***************************************************************** -->

    <inputs>

        <param name="input" type="data" format="fast5.tar" label="Input reads" />

    </inputs>

    <!-- ***************************************************************** -->

    <outputs>

        <data name="output" format="fastq" label="${tool.name} on ${on_string} (called.fastq)" />

    </outputs>

    <!-- ***************************************************************** -->

    <tests>
        <!-- multithreaded output is non-deterministic, so simply compare file
        sizes -->
        <test>
            <param name="input" value="test_data.fast5.tar.gz" ftype="fast5.tar.gz" />
            <output name="output" file="test_data.fastq" compare="sim_size" delta="0" />
        </test>
    </tests>

    <!-- ***************************************************************** -->

    <help>
    <![CDATA[

**Description**

Nanonet provides recurrent neural network basecalling for Oxford Nanopore
MinION data. It represents the first generation of such a basecaller from
Oxford Nanopore Technologies, and is provided as a technology demonstrator.
Nanonet is provided unsupported by Oxford Nanopore Technologies, see
LICENSE.md for more information.

For training networks, Nanonet leverages currennt to run recurrent neural
networks. Currennt is generally run with GPUs to aid performance but can be
run in a CPU only environment. The basecaller does not require currennt, and
is written in pure python with minimal requirements.

The Galaxy wrapper has modified nanonet to take a gzip tarball of FAST5 reads
as input, such as can be produced by `poretools combine`, and always outputs a
single FASTQ file.

This is the 1D basecaller.

    ]]>
    </help>

    <!-- ***************************************************************** -->
    
    <citations>
    </citations>

</tool>