hzi-bifo / hcv-mds

Neutralization maps for HCV data

Neutralization plots for HCV

Requirements

• Clone this repository:

  git clone https://github.com/hzi-bifo/hcv-mds 
  

• Python 2 with the following packages:

  • numpy
  • scipy
  • matplotlib
  • pandas
  • sklearn
  • gapkmean

  If these are not installed, you can install them with pip.

   cd ./scripts
   pip2 install -r requirements.txt 
  

• R version 3.4 or higher with the following packages:

  • plotrix
  • smacof
  • rgl
  • polynom
  • Hmisc

  If these are not installed, you can install them by running:

  R
  install.packages(c("plotrix", "smacof", "rgl", "polynom", "Hmisc"))
  

  Please make sure that the smacof package version is 1.8.13. You can explicitly install this version by running:

  install.packages("https://cran.r-project.org/src/contrib/Archive/smacof/smacof_1.8-13.tar.gz", repos=NULL, type='source')
  

Usage

• Put the file containing neutralization matrix (residual infectivity at maximum concentration) in the input folder. The file should be a comma separated csv file with the viruses as columns and sera as rows. Use NA for missing values. Here is an example:

  Sera / Viruses	Gt1b(Con1)	Gt2a(Jc1)	Gt2a(2a-3)	Gt2b(J8)
  37	80.2895218854	57.5271175607	74.9637533368	80.3419066704
  39	65.5662622736	72.1845541645	70.7020142884	71.0087697487
  41	84.0995706752	50.7927096018	41.204264273	46.658115028
  46	83.2643944991	76.0598932927	74.5105488987	NA
  47	70.9418542308	69.7642736513	75.3173168277	49.7798284265

• Go in the scripts folder and run the run_r.py file.

  cd scripts
  python run_r.py data.csv 
  

  There are different flags for selecting the weighting scheme, normalization method and number of dimensions for the multi-dimensional scaling.

  python run_r.py inputfile.csv [-h ] [-dim ] [-w ] [-norm ] [-loo ]
  
  -h, --help          :  show this help message and exit
  
  -norm, --normalize  :  Normalization method 
                         0 - No Normalization (Default) 
                         1 - take column minimum 
                         2 - take global minimum 
                       >=3 - minimum column basis 
  
  -w, --weight WEIGHT :  Weighing method 
                         0 - No weights (Default)
                         1 - 1/dij as weights 
  
  -dim, --dim DIM     :  Number of dimensions 
                         2 - Two dimensions (Default) 
                         3 - Three dimensions
                         4 or higher
  
  -loo, --loo LOO     :  Leave-one-out tests 
                         0 - No loo test (Default)) 
                         1 - Compute loo error
  

• After running this, the Neutralization map will be displayed and you'll be prompted to enter the value of K for K-means clustering. You can select the K by looking at the gap-statistics plot which will also be displayed along with the neutralization map. Once you enter the value of K, the final map with clustered viruses will also be displayed. When you close all the plots, program will automatically exit.

Results

• The log file for the entire process can be found in log folder with the name log.txt.
• The final plots and the summary of the result will be in the results folder. inputfile_nmds_plot.eps : eps file containing the neutralization plot of the input data inputfile_nmds_plot.png : png file containing the neutralization plot of the input data inputfile_summary.txt : txt file containing the stress values, loo error, cluster information etc. inputfile_cluser.png : plot of final clusters of the viruses
• The normalization files, LOO-error files, clustering analysis files etc. will be in the processing folder.

hcv-mds's Issues

Structure the repository

• data folder
• scripts
• results

Variable argument for results directory
give full path and name of results directory as an argument to the main script; If that dir exists, that one is used, else create it.

Readme extension

• add best plot to the main readme, instead of arbitrary one, and specify full legend (e.g. how scaled).
• create a link to a page where multiple relevant plots are shown and explained (the argument we follow in slelecting the settings).

add installation instructions to README

walk people through the installation.

Extend readme

Add descriptions of in- and outputs to readme. You can link to the respective files for an example data then, if these are plots that are produced, they can directly be visualized in the readme.