TODO:
- predictive models (logistic regression)

# SIMMS 1.3.2
==========================

- resubmission following archival due to broken dependency randomForestSRC. Did not require
 any changes in SIMMS, just resubmission with a new version

# SIMMS 1.3.1
==========================

- changed default params of random survival forest

# SIMMS 1.3.0
==========================

- implemented ensemble learning using random forest (survival) to create multivariate
  models
- [bug fix] saving subnet selected features for each dataset separately

# SIMMS 1.2.0
==========================

- support data scaling beyond z-scores including median and custom centering for methylation data.
- general improvement to documentation
- improved vignettes explaining how to structure custom network DB and annotation data
- added censorship tick marks for KM curves

# SIMMS 1.1.2
==========================

- Updates to handling of betas for discrete/nominal variables akin to predict()
- Formalised terminology for copy-number aberrations as 'cna' instead of 'cnv'
- Improved interaction model (abs(beta)) such that higher scores are given to 
  same signed (two) genes. Also fixed the extraction of beta/P from correct interaction row
  when >2 levels discrete/nominal variables are specified
- Implemented inline documentation using Roxygen

# SIMMS 1.1.1
==========================

- Converted elastic-net mixing parameter alpha [0,1] to search for optimal alpha
- Implemented parallel processing for glm alpha search space

# SIMMS 1.1.0
==========================

- Added elastic-net mixing parameter to support alpha [0,1] i.e ridge to LASSO

# SIMMS 1.0.3
==========================

- Added rmarkdown in Suggests section

# SIMMS 1.0.2
==========================

- Exception handling for interaction model
- Save selected features for every subnet that pass p.threshold and used in estimating subnet MDS
- Fixed NAMESPACE to comply with latest CRAN policies of explicit exports/imports

# SIMMS 1.0.1
==========================

- Fixed vignettes to comply with CRAN policy of not writing to user's home 
  directory

# SIMMS 1.0.0
==========================

- Added Generalised Linear Models (L1-penalty) based Cox models for fitting 
  multi-subnetwork models

- Added vignettes

- Added full support for discrete data types e.g copy-number aberrations

- General run-time optimisations and exception handling

- Added additional network databases including TCGA MEMo subnetworks for 
  breast, colon, glioblastoma, kidney and ovarian cancers

# SIMMS 0.0.1
==========================

- Added SIMMS first implementation