<tool id="plasmidtrakr" name="PlasmidTrakr" version="0.2.1">
    <description>Screens assemblies against a Mash database and predicts isolate source using a trained machine learning model</description>
    
    <requirements>
        <requirement type="package" version="2.3">mash</requirement>
        <requirement type="package" version="2.3.3">pandas</requirement>
        <requirement type="package" version="1.6.1">scikit-learn</requirement>
    </requirements>

    <command detect_errors="exit_code"><![CDATA[
        ## 1. Create a sanitized variable for the input name (removes spaces/special chars for the shell)
        #set $input_name = $assembly_input.element_identifier.replace(" ", "_")

        ## 2. Symlink the Mash database
        ln -s '$mash_database.fields.path' queries.msh &&

        ## 3. Run Mash Screen
        ## We redirect the output to a file named after the original input
        mash screen
            -w
            -i $threshold
            queries.msh
            '$assembly_input'
         > '${input_name}.tabular'
        &&

        ## 4. Run PlasmidTrakr prediction
        ## We pass the newly named file to the python script
        python $__tool_directory__/predict_source.py
            -i '${input_name}.tabular'
            -b '$model_selection.fields.path'
            -t $threshold
            -o '$prediction_output'
    ]]></command>


    <inputs>
        <param name="assembly_input" type="data" format="fasta,fasta.gz,fastq,fastq.gz" label="Genome Assembly / Reads" help="The FASTA/FASTQ file containing the isolate sequence."/>
        
        <param name="mash_database" type="select" label="Select Mash Database" help="Choose the pre-computed Mash sketch database to screen against.">
            <options from_data_table="mash_sketches">
                <validator type="no_options" message="No Mash databases are configured. Please contact your Galaxy administrator." />
            </options>
        </param>
        
        <param name="model_selection" type="select" label="Select Prediction Model" help="Choose which trained model to use for prediction.">
            <options from_data_table="plasmidtrakr_models">
                <validator type="no_options" message="No prediction models are configured. Please contact your Galaxy administrator." />
            </options>
        </param>
        
        <param name="threshold" type="float" value="0.95" min="0.0" max="1.0" label="Mash Identity Threshold" help="Filter plasmid hits below this identity. Must match the threshold used for model training."/>
    </inputs>

    <outputs>
        <data name="prediction_output" format="tabular" label="Prediction for ${on_string}" />
    </outputs>

    <help><![CDATA[
**What it does**

This tool performs a complete workflow in a single step: it screens a genome assembly or read set against a built-in plasmid database using **mash screen**, and then feeds those plasmid hits into a pre-trained **machine learning model** to predict the original source of the isolate.

**Workflow**
1. Provide your **genome assembly (FASTA)** or raw reads.
2. Select your **Mash database** from the server's configured list.
3. Select the desired prediction model.
4. Execute to screen and predict in one step.

**Output**
A tabular file containing the isolate ID, the predicted source, and a confidence score.
    ]]></help>

    <citations>
        <citation type="bibtex">
            @misc{strain_2026_plasmidtrakr,
                author = {Strain, Errol},
                title = {PlasmidTrakr: A tool for predicting isolate source from plasmid profiles},
                year = {2026},
                publisher = {GitHub},
                journal = {GitHub repository},
                howpublished = {\url{https://github.com/estrain/plasmidtrakr}}
            }
        </citation>
    </citations>
</tool>