From 091880d20567beb732b5d23e0b3b9a4397bb29a0 Mon Sep 17 00:00:00 2001 From: hyginn Date: Fri, 25 Sep 2020 21:50:42 +1000 Subject: [PATCH] Update script location --- BIN-ALI-BLAST.R | 35 +++++++++++++++++++---------------- 1 file changed, 19 insertions(+), 16 deletions(-) diff --git a/BIN-ALI-BLAST.R b/BIN-ALI-BLAST.R index 55ac089..191ca5f 100644 --- a/BIN-ALI-BLAST.R +++ b/BIN-ALI-BLAST.R @@ -1,20 +1,15 @@ # tocID <- "BIN-ALI-BLAST.R" # -# ---------------------------------------------------------------------------- # -# PATIENCE ... # -# Do not yet work wih this code. Updates in progress. Thank you. # -# boris.steipe@utoronto.ca # -# ---------------------------------------------------------------------------- # -# # Purpose: A Bioinformatics Course: # R code accompanying the BIN-ALI-BLAST unit. # -# Version: 1.2 +# Version: 1.3 # -# Date: 2017 10 - 2019 01 +# Date: 2017-10 - 2020-09 # Author: Boris Steipe (boris.steipe@utoronto.ca) # # Versions: +# 1.3 2020 Maintenance # 1.2 Change from require() to requireNamespace(), # use ::() idiom throughout # 1.1 Fixed parsing logic. @@ -35,13 +30,13 @@ #TOC> ========================================================================== -#TOC> +#TOC> #TOC> Section Title Line #TOC> --------------------------------------------------- -#TOC> 1 Defining the APSES domain 42 -#TOC> 2 Executing the BLAST search 64 -#TOC> 3 Analysing results 86 -#TOC> +#TOC> 1 Defining the APSES domain 43 +#TOC> 2 Executing the BLAST search 73 +#TOC> 3 Analysing results 95 +#TOC> #TOC> ========================================================================== @@ -50,10 +45,18 @@ # Load your protein database source("makeProteinDB.R") -# Get the APSES domain sequence for MBP1_MYSPE feature annotation. (You have -# entered this data in the BIN-ALI-Optimal_sequence_alignment unit.) +# Get the APSES domain sequence via your MBP1_MYSPE feature annotation. (You +# have entered this data into your database in the +# BIN-ALI-Optimal_sequence_alignment unit.) -(proID <- myDB$protein$ID[myDB$protein$name == "MBP1_"]) # <<< EDIT +( myOrth <- sprintf("MBP1_%s", biCode(MYSPE)) ) # If this is not the correct + # name of the Mbp1 orthologue + # of Mbp1 in your protein + # database, DON'T continue. We + # need to fix this problem. + # Get in touch. + +(proID <- myDB$protein$ID[myDB$protein$name == myOrth]) (ftrID <- myDB$feature$ID[myDB$feature$name == "APSES fold"]) (fanID <- myDB$annotation$ID[myDB$annotation$proteinID == proID & myDB$annotation$featureID == ftrID])