updating the README info for 1.1

Author: holtjma

README.md

@@ -1,11 +1,24 @@
 # VarSight
 
-VarSight is a collection of scripts used to test the application of classification algorithms for identifying variants that will be returned on a clinical report.  VarSight relies on [Codicem](http://envisiongenomics.com/codicem-analysis-platform/) to perform common clinical pipeline pre-processing, namely variant annotation and filtering.  In addition to data available in Codicem, it also uses gene rankings that are based on Human Phenotype Ontology (HPO) terms using both a built in calculation (cosine score) and the external tool [PyxisMap](https://github.com/HudsonAlpha/LayeredGraph).  The classifiers in VarSight ingest a total of 50 features that are used to test both classification of reported variants and prioritization of reported variants in a larger filtered list.  Results from v1 of VarSight tests are available in the folder labeled "paper".
+VarSight is a collection of scripts used to test the application of classification algorithms for identifying variants that will be returned on a clinical report.  VarSight relies on [Codicem](http://envisiongenomics.com/codicem-analysis-platform/) to perform common clinical pipeline pre-processing, namely variant annotation and filtering.  In addition to data available in Codicem, it also uses gene rankings that are based on Human Phenotype Ontology (HPO) terms using both a built in calculation (cosine score) and the external tool [PyxisMap](https://github.com/HudsonAlpha/LayeredGraph).  The classifiers in VarSight ingest a total of 95 features that are run through a feature selection process to get the top 20 features.  Those 20 features are used to test both classification of reported variants and prioritization of reported variants in a larger filtered list.  Results from the latest version of VarSight tests are available in the folder labeled "paper".
+
+## Running VarSight (Beta)
+To run the trained models, you need a Codicem JSON file containing pre-annotated and pre-filtered variants.  The filter file used can be found in the `CODI_metadata` subdirectory.  Additionally, a file with one HPO term per line describing the patient is also required.  Finally, the trained classifiers (pickled in the `models` subdirectory) are required.  Given that information, the following command will print the VarSight predicted ranked ordering of variants:
+```
+python3 scripts/VarSight.py codicemJson hpoFN modelsFN
+```
+
+For more information on the VarSight command-line interface, run the help command:
+```
+python3 scripts/VarSight.py -h
+```
 
 ## Sub-folders
 1. CODI_metadata - Contains metadata files that describe how Codicem data was filtered and how annotations were used by the classifiers.
-2. paper - Contains the main LaTeX file for the paper, multiple Jinja2 template files to use for rendering our results tables automatically, and the results JSON file that contains all output from the classifier results.
-3. scripts - Contains all Python3 scripts used to test the classifiers and report results.
+2. ExomiserTemplates - Contains the templated parameter files used to run Exomiser as a comparison.
+3. models - Contains the pickled models after hyperparameter tuning and fitting to the training set.  The models can be used by the VarSight.py script to rank variants from a Codicem filtered variant file.
+4. paper - Contains the main LaTeX file for the paper, multiple Jinja2 template files to use for rendering our results tables automatically, and the results JSON file that contains all output from the classifier results.
+5. scripts - Contains all Python3 scripts used to test the classifiers and report results.
 
 ## Publications
 

scripts/README.md

@@ -1,15 +1,20 @@
 # VarSight Scripts
 This folder contains a wide variety of scripts that were used to gather data, train the classifiers, test the classifiers, and report the results.
 
+## VarSight main
+`VarSight.py` contains the main command-line interface for running VarSight after models have been trained.  Please refer to the main README for more instructions.
+
 ## Data Parsing Scripts
 The following files are primarily used to pre-process data into easy-to-manage formats in Python.
 
 1. CodiDumpUtil.py - This script contains helper functions for parsing a filtered Codicem JSON. It's main purpose is to load a Codicem file and full out the fields requested for use by the classifiers.
-2. ExomiserUtil.py - This script contains helper functions for parsing prioritized variants from Exomiser.  It's main purpose is to search for reported variants in the Exomiser output and return the ranks of those variants if found.  If not found, it conservatively assumes the next best rank is where the variant is found.
+2. ExomiserUtil.py - This script contains helper functions for parsing prioritized variants from Exomiser.
 3. HPOUtil.py - This script contains helper functions for calculating the gene rankings based on the cosine score from the Human Phenotype Ontology (HPO) terms.
 4. OntologyUtil.py - This script contains helper functions for ontologies.  It is primarily support code for HPOUtil.py.
-5. PyxisMapUtil.py - This script contains helper functions for retrieving gene rankings based on the PyxisMap ranks using the HPO terms.
-6. SummaryDBUtil.py - This script contains helper functions for parsing the database dump files containing metadata for the Undiagnosed Diseases Network including samples IDs, HPO terms, and reported primary variants.  Note: these files are not available on GitHub due to Personal Health Information (PHI).
+5. PVPUtil.py - This script contains helper functions for parsing prioritized variants from DeepPVP.
+6. PhenGenUtil.py - This script contains helper function for parsing prioritized variants from Phen-Gen.
+7. PyxisMapUtil.py - This script contains helper functions for retrieving gene rankings based on the PyxisMap ranks using the HPO terms.
+8. SummaryDBUtil.py - This script contains helper functions for parsing the database dump files containing metadata for the Undiagnosed Diseases Network including samples IDs, HPO terms, and reported primary variants.  Note: these files are not available on GitHub due to Personal Health Information (PHI).
 
 ## Training/Testing Scripts
 The following files perform the core workhorse training, testing, and reporting of results from VarSight: