cbcrg/psldoc2-nf repository overview

⁠PSLDoc2

PSLDoc2 is the extended version of PSLDoc and it is implemented based on the nextflow framework. https://github.com/paoloditommaso/nextflow⁠

web site: http://tcoffee.crg.cat/psldoc2⁠

⁠Dependencies

• PSI-BLAST (Position-Specific Initiated BLAST 2.2.28+)
• FactoMineR R package (it requires R packages: car, ellipse, lattice, cluster, scatterplot3d, leaps) http://cran.r-project.org/web/packages/FactoMineR/index.html⁠

⁠Configuration

Extra R libraries can be specified using the R_LIBS environment variables. Add this variable to the nextflow.config file in the project root folder, as shown in the following example:

env {
  R_LIBS = "$PWD/r_libs/"
}

⁠Stand alone version

⁠Prediction

1. localizaiton
   • prokaryotic : PSORTb v3 data
     • gram-negative
     • gram-positve
     • archaeal
   • eukaryotic
     • plant protein, http://cal.tongji.edu.cn/PlantLoc/⁠
     • human
     • animal
     • fungi
     • yeast
     • Viral protein
2. nuclear
   • uclear裡的子胞器(sub-localization)是針對細胞核這位置再繼續細分下去，也有sub-mitochondrian等子胞器的分類
3. protein function
   • Kinase

⁠Build your own model

1. training phase

   • psldoc2-train.nf -> CA_train+nFoldValidation.R

   • input: model.fasta

   • output: model.tfpssm, plot_model.json, accuracy-nfold.csv, predict.csv (CA dims)

      $ nextflow psldoc2_train.nf --model=../../../data/PSL/PSORTb3.0/Archaeal.fasta --fold_num=5
     

     Copy

2. predicting phase

   • psldoc2-pred.nf -> CA_pred.R (query.fasta, model.tfpssm, CA dims)

   • input: query.fasta, model.tfpssm

   • output: plot_query.json, query.pred

      $ nextflow psldoc2_pred.nf --query small.fa --model data/prokaryotic/Archaeal.tfpssm --CA_dim 36
     

     Copy