Expose options for supervised learning models selection

Major

• Add option for 'fast' run (-f / --fast) with 4 ML models (instead of the default 6 models)

• Add option (-m) to explicitly specify the supervised models to be trained by mantis-ml.
  Available model options are:

  • et: Extra Trees
  • rf: Random Forest
  • gb: Gradient Boosting
  • xgb: XGBoost
  • svc: Support Vector Classifier
  • dnn: Deep Neural Net
  • stack: Stacking classifier

  Multiple models may be specified using a ',' separator, e.g. -m et,rf,stack

• Stacking classifier may now be run using the -m option with the stack arg, i.e.: -m stack

Minor

• Fix bug in hypergeom_enrichment module when reading external ranked files (only tab-delimiter is currently allowed in case of an external file with two columns)

Streamlined deployment - Simplified input parameters

• Streamlined installation with "python setup.py install"
• Tool now available at PyPI and can be installed through pip
• The tool now offers three executable scripts to run from the command line (mantisml, mantisml-profiler and mantisml-overlap)
• The user can define their own output folder via the -o option
• Config input parameters have been drastically simplified: a lot of the underlying complexity has been hidden to the end user and only required input is disease/phenotype-associated terms in free text
• The mantisml-overlap script is a new addition to the tool as it packages the enrichment test functionality between mantis-ml predictions and any external ranked gene list that is provided by the user