<tool id="model_prediction" name="Model Prediction" version="@VERSION@">
    <description>predicts on new data using a preffited model</description>
    <macros>
        <import>main_macros.xml</import>
        <import>keras_macros.xml</import>
    </macros>
    <expand macro="python_requirements"/>
    <expand macro="macro_stdio"/>
    <version_command>echo "@VERSION@"</version_command>
    <command>
        <![CDATA[
        export HDF5_USE_FILE_LOCKING='FALSE';
        python '$__tool_directory__/model_prediction.py'
            --inputs '$inputs'
            --infile_estimator '$infile_estimator'
            --outfile_predict '$outfile_predict'
            --infile_weights '$infile_weights'
            #if $input_options.selected_input == 'seq_fasta'
            --fasta_path '$input_options.fasta_path'
            #elif $input_options.selected_input == 'variant_effect'
            --ref_seq '$input_options.ref_genome_file'
            --vcf_path '$input_options.vcf_file'
            #else
            --infile1 '$input_options.infile1'
            #end if
        ]]>
    </command>
    <configfiles>
        <inputs name="inputs" />
    </configfiles>
    <inputs>
        <param name="infile_estimator" type="data" format="zip" label="Choose the dataset containing pipeline/estimator object"/>
        <param name="infile_weights" type="data" format="h5" optional="true" label="Choose the dataset containing weights for the estimator above" help="Optional. For deep learning only."/>
        <param argument="method" type="select" label="Select invocation method">
            <option value="predict" selected="true">predict</option>
            <option value="predict_proba">predict_proba</option>
        </param>
        <conditional name="input_options">
            <param name="selected_input" type="select" label="Select input data type for prediction">
                <option value="tabular" selected="true">tabular data</option>
                <option value="sparse">sparse matrix</option>
                <option value="seq_fasta">sequnences in a fasta file</option>
                <option value="variant_effect">reference genome and variant call file</option>
            </param>
            <when value="tabular">
                <param name="infile1" type="data" format="tabular" label="Training samples dataset:"/>
                <param name="header1" type="boolean" optional="true" truevalue="booltrue" falsevalue="boolfalse" checked="False" label="Does the dataset contain header:" />
                <conditional name="column_selector_options_1">
                    <expand macro="samples_column_selector_options" multiple="true"/>
                </conditional>
            </when>
            <when value="sparse">
                <param name="infile1" type="data" format="txt" label="Select a sparse matrix" help=""/>
            </when>
            <when value="seq_fasta">
                <param name="fasta_path" type="data" format="fasta" label="Dataset containing fasta genomic/protein sequences" help="Sequences will be one-hot encoded to arrays."/>
                <param name="seq_type" type="select" label="Sequence type">
                    <option value="FastaDNABatchGenerator">DNA</option>
                    <option value="FastaRNABatchGenerator">RNA</option>
                    <option value="FastaProteinBatchGenerator">Protein</option>
                </param>
            </when>
            <when value="variant_effect">
                <param name="ref_genome_file" type="data" format="fasta" label="Dataset containing reference genomic sequence" help="fasta"/>
                <param name="blacklist_regions" type="select" label="blacklist regioins" help="A pre-loaded list of blacklisted intervals.Refer to `selene` for details.">
                    <option value="none" selected="true">None</option>
                    <option value="hg38">hg38</option>
                    <option value="hg19">hg19</option>
                </param>
                <param name="vcf_file" type="data" format="vcf" label="Dataset containing sequence variations" help="vcf"/>
                <param name="seq_length" type="integer" value="1000" label="Encoding seqence length" help="A stretch of sequence surrounding the variation position on the reference genome."/>
                <param name="output_reference" type="boolean" truevalue="booltrue" falsevalue="boolfalse" checked="false" label="Predict the reference sequence?" help="If False, predict on the variant sequence."/>
            </when>
        </conditional>
    </inputs>
    <outputs>
        <data format="tabular" name="outfile_predict"/>
    </outputs>
    <tests>
        <test>
            <param name="infile_estimator" value="best_estimator_.zip" ftype="zip"/>
            <param name="method" value="predict"/>
            <param name="infile1" value="regression_X.tabular" ftype="tabular"/>
            <param name="header1" value="true" />
            <param name="col1" value="1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17"/>
            <output name="outfile_predict" file="model_pred01.tabular"/>
        </test>
        <test>
            <param name="infile_estimator" value="keras_model04" ftype="zip"/>
            <param name="infile_weights" value="train_test_eval_weights02.h5" ftype="h5"/>
            <param name="method" value="predict"/>
            <param name="infile1" value="regression_X.tabular" ftype="tabular"/>
            <param name="header1" value="true" />
            <param name="col1" value="1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17"/>
            <output name="outfile_predict" >
                <assert_contents>
                    <has_n_columns n="1"/>
                    <has_text text="70.2"/>
                    <has_text text="61.2"/>
                    <has_text text="74.2"/>
                    <has_text text="65.9"/>
                    <has_text text="52.9"/>
                </assert_contents>
            </output>
        </test>
    </tests>
    <help>
        <![CDATA[
**What it does**

Given a fitted estimator and new data sets, this tool outpus the prediction results on the data sets via invoking the estimator's `predict` or `predict_proba` method.

For estimator, this tool supports fitted sklearn estimators (pickled) and trained deep learning models (model skeleton + weights). It predicts on three different dataset inputs,

- tabular

- sparse

- bio-sequences in a fasta file

- reference genome and variant call file

        ]]>
    </help>
    <expand macro="sklearn_citation">
        <expand macro="keras_citation"/>
        <expand macro="selene_citation"/>
    </expand>
</tool>