A typical process runs FUGAsseM per dataset.

• For a list of command line options, run:

  $ fugassem --help

  This command yields:

  usage: fugassem_workflow.py [-h] [--version]
                      [--taxon-level {MSP,Terminal,Species,Genus,Family,
                      Order,Class,Phylum}]
          ...

• Run the canonical function prediction of FUGAsseM, which requires MTX abundance data and annotation of protein families for function prediction.

  • Input files:

    • Normalized MTX abundance table stratified by taxa, e.g. demo_proteinfamilies_rna_CPM.stratified_Species_mtx.tsv
    • GO (raw) annotations for protein families, e.g. demo_proteinfamilies.GO.simple.tsv

  • Demo run of FUGAsseM-MTX model

    $ fugassem --basename $BASENAME \ 
    --input $INPUT_MTX \ 
    --input-annotation $INPUT_annotation \
    --output $OUTPUT_DIR

    • $INPUT_MTX = the protein families MTX abundances file (TSV format)
    • $INPUT_annotation = raw GO annotations for some of these protein families (TSV format)
    • $OUTPUT_DIR = the output folder
    • Output files will be created named with $BASENAME:
      1. $OUTPUT_DIR/merged/$BASENAME.finalized_ML.prediction.tsv: this file combines the finalized predictions from all taxa by using machine learning approaches based on MTX coexpression patterns (TSV format).
      2. Predictions files of each taxon will also be created:
         • The finalized prediction results using MTX-coexpression evidence per taxon are in the file: $OUTPUT_DIR/main/$TAXON_NAME/prediction/finalized/$BASENAME.$TAXON_NAME.finalized_ML.prediction.tsv.

• Run the integrated function prediction workflow of FUGAsseM. When other community-wide data are available, FUGAsseM can predict functions by integrating multiple pieces of evidence. The additional steps in this workflow are 1) building individual machine learning classifiers for each type of evidence including coexpression as discussed above, 2) and integration to generate an ensemble classifier for final function prediction.

  • Input files

    • normalized MTX-based abundance table stratified by taxa (TSV format file), e.g. demo_proteinfamilies_rna_CPM.stratified_Species_mtx.tsv
    • raw annotations for protein families (TSV format file), e.g. demo_proteinfamilies.GO.simple.tsv
    • vector-based evidence file (TSV format file), e.g. demo_proteinfamilies.GO.homology.tsv
    • matrix-based evidence files (TSV format file), e.g.
      • demo_proteinfamilies.DDI.simple.tsv: Domain-Domain interactions for building DDI network for prediction
      • demo_proteinfamilies.contig.simple.tsv: TAXON_NAMEMGX-based contigs of protein families for building co-contig network for prediction

  • Demo run of FUGAsseM-full model

    $ fugassem --basename $BASENAME \
    --input $INPUT_MTX \ 
    --input-annotation $INPUT_annotation \ 
    --vector-list $VECTOR_list --matrix-list $METRIX_list \ 
    --output $OUTPUT_DIR

    • $INPUT_MTX = the protein families MTX abundances file (TSV format)
    • $INPUT_annotation = raw GO annotations for some of these protein families (TSV format)
    • $VECTOR_list = file names of vector-based evidence data, provided as a string of 'file1,file2', semi-colon delimited for multiple files.
    • $METRIX_list = file names of matrix-based evidence data, provided as a string of 'file1,file2', semi-colon delimited for multiple files.
    • $OUTPUT_DIR = the output folder
    • Output files will be created named with $BASENAME:
      1. $OUTPUT_DIR/merged/$BASENAME.finalized_ML.prediction.tsv: this file combines the finalized predictions from all taxa by using machine learning approaches based on MTX coexpression patterns (TSV format).
      2. $OUTPUT_DIR/merged/$BASENAME.$EVIDENCE_TYPE_ML.prediction.tsv (where $EVIDENCE_TYPE = the basename of each piece of evidence): this file includes combined predictions based on individual type of evidence (TSV format file).
      3. Predictions files of each taxon will also be created:
         • FUGAsseM predicts functions based on input evidence data.
         • The finalized prediction results using integrated evidence per taxon are in the file: $OUTPUT_DIR/main/$TAXON_NAME/prediction/finalized/$BASENAME.$TAXON_NAME.finalized_ML.prediction.tsv.
         • The prediction results by using individual evidence per taxon are in the file: $OUTPUT_DIR/$TAXON_NAME/prediction/$EVIDENCE_TYPE/$BASENAME.$TAXON_NAME.$EVIDENCE_TYPE_ML.prediction.tsv (where $EVIDENCE_TYPE = the basename of each piece of evidence).