<tool id="celltypist_predict" name="CellTypist: Predict Cell Types" version="1.6.3+galaxy1" profile="20.05" license="MIT">
    <description>Predict cell types using an existing CellTypist model.</description>
    <requirements>
        <requirement type="package" version="1.6.3">celltypist</requirement>
    </requirements>
    <command detect_errors="exit_code">
    <![CDATA[
    #if $inp.format == 'h5ad':
        ln -s "$inp.data" input.h5ad &&
    #else:
        ln -s "$inp.data" input.mtx &&
    #end if
    ln -s "$model_file" model.pkl &&
    python $__tool_directory__/celltypist_CLI.py --action predict
      --output_path "$output_predictions"
      --model model.pkl
      #if $inp.format == 'h5ad':
        --input_path input.h5ad
        #if $inp.raw_counts_layer:
          --raw_counts_layer "$inp.raw_counts_layer"
        #end if
        #if $inp.gene_symbols_field:
          --gene_symbols_field "$inp.gene_symbols_field"
        #end if
      #else:
        --input_path input.mtx
        --gene_file "$inp.gene_file"
        --cell_file "$inp.cell_file"
      #end if
      
      ${normalize}
      ${transpose_input}
      --mode "$mode"
      --p_thres "$p_thres"
      ${majority_voting}
      #if $over_clustering:
        --over_clustering "$over_clustering"
      #end if
    ]]>
  </command>
    <inputs>
        <conditional name="inp" label="Input format">
            <param name="format" type="select" label="Input format" help="Format for the input files, either AnnData or Matrix market format (10x)">
                <option value="h5ad">AnnData</option>
                <option value="mtx">Matrix market</option>
            </param>
            <when value="h5ad">
                <param format="h5ad" type="data" name="data" label="Input data (AnnData)" help="The input data for prediction."/>
                <param type="text" name="raw_counts_layer" label="Raw counts layer" help="The name of the layer that stores the raw counts. Uses default matrix if not present."/>
                <param type="text" name="gene_symbols_field" label="Gene symbols field" help="The field in AnnData where the gene symbols are stored, if not in index."/>
            </when>
            <when value="mtx">
                <param format="mtx,txt" type="data" name="data" label="Input matrix (mtx file)" help="The input data for prediction."/>
                <param format="txt,tabular" type="data" name="gene_file" label="Gene file" help="The file containing one gene per line corresponding to the genes in the input data (required for .mtx data)."/>
                <param format="txt,tabular" type="data" name="cell_file" label="Cell file" help="The file containing one cell per line corresponding to the cells in the input data (required for .mtx data)."/>
            </when>
        </conditional>
        <param format="data" type="data" name="model_file" optional="false" label="Model file" help="A CellTypist model file in .pkl format."/>
        <param type="boolean" argument="--normalize" label="Normalize" truevalue="--normalize" falsevalue="" help="If raw counts are provided in the AnnData object, they need to be normalized."/>
        <param type="boolean" argument="--transpose_input" label="Transpose input" truevalue="--transpose_input" falsevalue="" help="If the provided matrix is in the gene-by-cell format, please transpose the input to cell-by-gene format."/>
        <param type="select" name="mode" label="Mode" help="Mode for the prediction (e.g., best match, majority vote).">
            <option value="best_match" selected="true">Best match</option>
            <option value="prob_match">Probability match</option>
        </param>
        <param type="float" name="p_thres" value="0.5" label="Probability threshold" help="Probability threshold for assigning a cell type in a multiclass problem. Ignored if Mode set to Best match."/>
        <param type="boolean" argument="--majority_voting" label="Majority voting" truevalue="--majority_voting" falsevalue="" help="Refine the predicted labels by running the majority voting classifier after over-clustering."/>
        <param type="text" name="over_clustering" label="Over-clustering" help="If majority voting is set to True, specify the type of over-clustering that is to be performed. This can be specified in the AnnData or an input file specifying the over-clustering per cell. If not present, then the default heuristic over-clustering based on input data will be used."/>
    </inputs>
    <outputs>
        <data format="tabular" name="output_predictions" label="${tool.name} on ${on_string}: cell typing table"/>
    </outputs>
    <tests>
        <test expect_num_outputs="1">
            <param name="model_file" value="Healthy_COVID19_PBMC.pkl"/>
            <param name="normalize" value="true"/>
            <param name="mode" value="best_match"/>
            <param name="p_thres" value="0.5"/>
            <conditional name="inp">
                <param name="format" value="h5ad"/>
                <param name="data" value="raw_counts_pbmc3k.h5ad"/>
                <param name="gene_symbols_field" value="SYMBOL"/>
            </conditional>
            <output name="output_predictions">
                <assert_contents>
                    <has_text text="MAIT"/>
                    <has_n_lines n="2701"/>
                    <has_n_columns n="2"/>
                </assert_contents>
            </output>
        </test>
    </tests>
    <help><![CDATA[
.. _celltypist_predict_tool:

CellTypist: Predict Cell Types
------------------------------

CellTypist is a tool for automatic cell type annotation using single-cell RNA sequencing data. This Galaxy tool wrapper executes the CellTypist script in "predict" mode, allowing you to predict cell types using an existing CellTypist model.

Input
^^^^^

- **Input format**: Choose the format for the input files, either AnnData or Matrix market format (10x).

  - If AnnData is selected, provide the input data (AnnData file).
  - If Matrix market is selected, provide the input matrix (mtx file), gene file, and cell file.

- **Input data**: The input data for prediction.

- **Gene file**: The file containing one gene per line corresponding to the genes in the input data (required for .mtx data).

- **Cell file**: The file containing one cell per line corresponding to the cells in the input data (required for .mtx data).

- **Model file**: The existing CellTypist model file in .pkl format.

- **Raw counts layer**: The name of the layer that stores the raw counts. Uses the default matrix if not provided.

- **Gene symbols field**: The field in AnnData where the gene symbols are stored, if not in index.

- **Normalize**: If raw counts are provided in the AnnData object, they need to be normalized.

- **Transpose input**: If the provided matrix is in the gene-by-cell format, transpose the input to cell-by-gene format.

- **Mode**: Mode for the prediction (e.g., best match, majority vote).

- **Probability threshold**: Probability threshold for assigning a cell type in a multiclass problem. Ignored if Mode set to Best match.

- **Majority voting**: Refine the predicted labels by running the majority voting classifier after over-clustering.

- **Over-clustering**: If majority voting is set to True, specify the type of over-clustering that is to be performed. This can be specified in the AnnData or an input file specifying the over-clustering per cell. If not provided, the default heuristic over-clustering based on input data will be used.

Output
^^^^^^

- **Output predictions**: The predicted cell types for the input data.

Citation
^^^^^^^^

- CellTypist: a tool for automatic cell type annotation using single-cell RNA sequencing data doi/10.1126/science.abl5197

History
^^^^^^^

Version 1.6.3+galaxy0: Initial wrapper, Pablo Moreno. CLI written by Alex Proutski.
        ]]>
    </help>
    <citations>
        <citation type="bibtex">@article{dominguez_conde2022,
            title={Cross-tissue immune cell analysis reveals tissue-specific features in humans},
            author={Domínguez Conde, C and Xu, C and Jarvis, LB and Rainbow, DB and Wells, SB and Gomes, T and Howlett, SK and Suchanek, O and Polanski, K and King, HW and Mamanova, L and Huang, N and Szabo, PA and Richardson, L and Bolt, L and Fasouli, ES and Mahbubani, KT and Prete, M and Tuck, L and Richoz, N and Tuong, ZK and Campos, L and Mousa, HS and Needham, EJ and Pritchard, S and Li, T and Elmentaite, R and Park, J and Rahmani, E and Chen, D and Menon, DK and Bayraktar, OA and James, LK and Meyer, KB and Yosef, N and Clatworthy, MR and Sims, PA and Farber, DL and Saeb-Parsy, K and Jones, JL and Teichmann, SA},
            journal={Science},
            volume={376},
            number={6594},
            pages={eabl5197},
            year={2022},
            publisher={American Association for the Advancement of Science},
            pmid={35549406},
            pmcid={PMC7612735}
          }</citation>
    </citations>
</tool>