@@ -2,7 +2,7 @@ Package: MSnID

 Type: Package

 Title: Utilities for Exploration and Assessment of Confidence of

     LC-MSn Proteomics Identifications

-Version: 1.7.3

+Version: 1.15.2

 Author: Vlad Petyuk with contributions from Laurent Gatto

 Maintainer: Vlad Petyuk <petyuk@gmail.com>

 Description: Extracts MS/MS ID data from mzIdentML (leveraging mzID package) or

@@ -19,4 +19,4 @@ Imports: MSnbase (>= 1.12.1), mzID (>= 1.3.5), R.cache,

     data.table, Biobase, ProtGenerics, reshape2, dplyr, mzR

 Suggests: BiocStyle, msmsTests, ggplot2, RUnit, BiocGenerics

 LazyData: yes

-biocViews: Proteomics, MassSpectrometry

+biocViews: Proteomics, MassSpectrometry, ImmunoOncology

@@ -13,7 +13,8 @@ importFrom("ProtGenerics", "proteins")

 importFrom("mzID", "mzID")

 importFrom("mzR","openIDfile","score","modifications","fileName")

 importMethodsFrom("mzID", "flatten")

-importFrom("parallel", "mclapply", "detectCores","makeCluster","stopCluster")

+importFrom("parallel", "mclapply", "detectCores","makeCluster",

+           "stopCluster","parLapply","clusterExport","clusterEvalQ")

 importClassesFrom("MSnbase", "MSnSet")

 importClassesFrom("Biobase", "AnnotatedDataFrame")

 #-----------------------------------------------------------------------------

@@ -1,3 +1,21 @@

+Version 1.15.2

+-------------

+ o Broadened the scope of the tool to ImmunoOncology - that is finding rare

+   events that reflected in protein sequence.

+   

+Version 1.15.1

+-------------

+ o mzR now added scan number(s) into the table representation of

+   mzIdentML object. As a results it caused an error in my unit test

+   checking if the file reads properly. Fixed this check with updated hash.

+

+Version 1.11.1

+-------------

+ o Switch from mclapply to parLapply in filter optimizaition.

+   This fixes a bug 

+   "Assertion failure at kmp_runtime.cpp(6480): __kmp_thread_pool == __null."

+   Also allows to run parallel on Window OS

+

 Version 1.7.3

 -------------

  o Logical error fix in PSM FDR calculation. Prior it ignored redundancy

@@ -1,5 +1,4 @@

-

 .PROTON_MASS <- 1.007276466812

 setClass(Class="MSnID",

@@ -14,4 +13,3 @@ setClass(Class="MSnIDFilter",

                 filterList="list",

                 validParNames="character")

 )

-

@@ -155,13 +155,26 @@ setMethod(

 #----Filter---------------------------------------------------------------------

+# # Old implementation. Just as a backup.

+# # See https://github.com/vladpetyuk/MSnID/issues/5 for the issue.

+# setMethod(

+#     "apply_filter",

+#     signature(msnidObj="MSnID", filterObj="character"),

+#     definition=function(msnidObj, filterObj)

+#     {

+#         exprssn <- parse(text=filterObj, srcfile=NULL, keep.source=FALSE)

+#         msnidObj@psms <- msnidObj@psms[eval(exprssn),]

+#         return(msnidObj)

+#     }

+# )

 setMethod(

     "apply_filter",

     signature(msnidObj="MSnID", filterObj="character"),

     definition=function(msnidObj, filterObj)

     {

         exprssn <- parse(text=filterObj, srcfile=NULL, keep.source=FALSE)

-        msnidObj@psms <- msnidObj@psms[eval(exprssn),]

+        idx <- eval(exprssn, envir = msnidObj@psms, enclos = parent.frame())

+        msnidObj@psms <- msnidObj@psms[idx,]

         return(msnidObj)

     }

 )

@@ -567,15 +580,29 @@ setAs("MSnID", "MSnSet",

 setMethod("infer_parsimonious_accessions", "MSnID",

           definition=function(object)

           {

+              # # Old code for inferring accessions.

+              # # It is too slow.

+              # infer_acc <- function(x){

+              #     res <- list()

+              #     while(nrow(x) > 0){

+              #         top_prot <- names(which.max(table(x[['accession']])))

+              #         top_peps <- subset(x, accession == top_prot)

+              #         res <- c(res, list(top_peps))

+              #         x <- subset(x, !(pepSeq %in% top_peps[,"pepSeq"]))

+              #     }

+              #     return(Reduce(rbind,res))

+              # }

+

               infer_acc <- function(x){

                   res <- list()

+                  setDT(x)

                   while(nrow(x) > 0){

-                      top_prot <- names(which.max(table(x[['accession']])))

+                      top_prot <- x[, .N, by=accession][which.max(N),,]$accession

                       top_peps <- subset(x, accession == top_prot)

                       res <- c(res, list(top_peps))

-                      x <- subset(x, !(pepSeq %in% top_peps[,"pepSeq"]))

+                      x <- subset(x, !(pepSeq %in% top_peps[[1]]))

                   }

-                  return(Reduce(rbind,res))

+                  return(rbindlist(res, use.names=F, fill=FALSE, idcol=NULL))

               }

               x <- unique(psms(object)[,c("pepSeq","accession")])

@@ -1,6 +1,4 @@

-

-

 .is_filterList_valid <- function(filterList)

 {

     entries=c("comparison","threshold")

@@ -141,4 +139,3 @@ setMethod("$<-", "MSnIDFilter",

             }

 )

-

@@ -88,6 +88,34 @@

 }

+.optimize_filter.grid.parLapply <- function(filterObj, msnidObj,

+                                           fdr.max, level, n.iter, mc.cores)

+{

+    par.grid <- .construct_optimization_grid(filterObj, msnidObj, n.iter)

+    cl <- makeCluster(mc.cores)

+    # clusterExport(cl, 

+    #                         list(".get_num_pep_for_fdr"),

+    #                         envir = "package:MSnID")

+    clusterEvalQ(cl, library("MSnID"))

+    evaluations <- parLapply(cl,

+                             seq_len(nrow(par.grid)),

+                             function(i){

+                                .get_num_pep_for_fdr(par.grid[i,],

+                                                     msnidObj,

+                                                     filterObj,

+                                                     fdr.max,

+                                                     level)})

+    stopCluster(cl)

+    evaluations <- round(unlist(evaluations))

+    if(all(evaluations == 0)){

+        warning(.msg.invalid.optimization.results.grid)

+        return(filterObj)

+    }

+    optim.pars <- par.grid[which.max(evaluations),]

+    newFilter <- update(filterObj, as.numeric(optim.pars))

+    return(newFilter)

+}

+

 setMethod("optimize_filter",

             signature(filterObj="MSnIDFilter", msnidObj="MSnID"),

             definition=function(filterObj, msnidObj, fdr.max,

@@ -120,19 +148,24 @@ setMethod("optimize_filter",

     }

     #

     if(method == "Grid"){

-        if(.Platform$OS.type == "unix"){

-            if(is.null(mc.cores))

-                mc.cores <- getOption("mc.cores", 2L)

-            optimFilter <-

-                .optimize_filter.grid.mclapply(filterObj, msnidObj,

-                                               fdr.max, level, 

-                                               n.iter, mc.cores)

-        }else{

-            # yet to implement effective parallelization on Windows

-            optimFilter <-

-                .optimize_filter.grid(filterObj, msnidObj,

-                                        fdr.max, level, n.iter)

-        }

+        # if(.Platform$OS.type == "unix"){

+        #     if(is.null(mc.cores))

+        #         mc.cores <- getOption("mc.cores", 2L)

+        #     optimFilter <-

+        #         .optimize_filter.grid.mclapply(filterObj, msnidObj,

+        #                                        fdr.max, level, 

+        #                                        n.iter, mc.cores)

+        # }else{

+        #     # yet to implement effective parallelization on Windows

+        #     optimFilter <-

+        #         .optimize_filter.grid(filterObj, msnidObj,

+        #                                 fdr.max, level, n.iter)

+        # }

+        mc.cores <- getOption("mc.cores", 2L)

+        optimFilter <-

+            .optimize_filter.grid.parLapply(filterObj, msnidObj,

+                                            fdr.max, level,

+                                            n.iter, mc.cores)

     }

     if(method %in% c("Nelder-Mead", "SANN")){

         optim.out <- optim(par=as.numeric(filterObj),

@@ -1,3 +1,5 @@

 MSnID

 ====

+[![Rdoc](http://www.rdocumentation.org/badges/version/MSnID)](http://www.rdocumentation.org/packages/MSnID)

+===

 `MSnID` is an R/Bioconductor package that provides convience tools for handling and filtering of MS/MS identification results. The official page is the Bioconductor landing page ([release](http://www.bioconductor.org/packages/release/bioc/html/MSnID.html) or [devel](http://www.bioconductor.org/packages/devel/bioc/html/MSnID.html) versions). The [github page](https://github.com/vladpetyuk/MSnID) page is for sharing, testing, issue tracking and forking/pulling purposes.

@@ -30,5 +30,6 @@ test_column_names <- function() {

 test_data_load_mzR <- function() {

     # now, check if it is what it supposed to be

-    checkIdentical(digest(psms(msnid3)),'e5c572c07878673f1165822969f81869')

+    # checkIdentical(digest(psms(msnid3)),'e5c572c07878673f1165822969f81869')

+    checkIdentical(digest(psms(msnid3)),'0b4e3b61e3fe007ed11651632fa3f1fb')

 }

@@ -1,9 +1,9 @@

 library("MSnID")

 data(c_elegans)

-test_infer_parsimonious_accessions <- function(){

-        # explicitely adding parameters that will be used for data filtering

+test_infer_parsimonious_accessions_old <- function(){

+    # explicitely adding parameters that will be used for data filtering

     msnidObj$msmsScore <- -log10(msnidObj$`MS-GF:SpecEValue`)

     msnidObj$absParentMassErrorPPM <- abs(mass_measurement_error(msnidObj))

@@ -16,3 +16,33 @@ test_infer_parsimonious_accessions <- function(){

 }

+

+# Above is the old function for testing protein inference.  I'll leave it for

+# now.  Below is the new way, where first all the inference will be done

+# outside of the test functions.

+

+

+# explicitely adding parameters that will be used for data filtering

+msnidObj$msmsScore <- -log10(msnidObj$`MS-GF:SpecEValue`)

+msnidObj$absParentMassErrorPPM <- abs(mass_measurement_error(msnidObj))

+# quick-and-dirty filter. The filter is too strong for the sake of saving time

+# at the minimal set of proteins inference step.

+msnidObj <- apply_filter(msnidObj, 'msmsScore > 12 & absParentMassErrorPPM < 2')

+msnidObj2 <- infer_parsimonious_accessions(msnidObj)

+

+test_infer_parsimonious_accessions_number <- function(){

+   checkEqualsNumeric(length(proteins(msnidObj2)), 551)

+}

+

+# test_infer_parsimonious_accessions_hash <- function(){

+#    checkIdentical(digest(psms(msnidObj2)),'42c967304603b17ef667fae5b8d5657f')

+# }

+

+test_infer_parsimonious_accessions_hash <- function(){

+    checkIdentical(digest(psms(msnidObj2)$accession),

+                   '6a3566a95b2e49a0f966d22ed897a752')

+}

+

+# Future challenges is to come up with tests that check inference that is

+# done outside of MSnID object

+

@@ -63,5 +63,3 @@ table(msnidObj$numMissCleavages)

 }

-

-

@@ -1,5 +1,3 @@

-% validCleavagePattern="[RK]\\.[^P]"

-

 \name{assess_termini}

 \alias{assess_termini}

@@ -60,5 +60,3 @@ massErr <- (msnidObj$experimentalMassToCharge -

 hist(massErr,xlim=c(-1,+1), breaks=seq(-1.5,+1.5,0.01))

 }

-

-

@@ -1,5 +1,3 @@

-% this is document of generic function

-

 \name{psms}

 \alias{psms}

 \alias{psms<-}

@@ -69,6 +69,3 @@ ppm <- mass_measurement_error(msnidObj)

 hist(ppm, 200, xlim=c(-20,+20))

 }

-

-

-

@@ -496,6 +496,3 @@ unlink(".Rcache", recursive=TRUE)

 \bibliographystyle{plainnat}

 \bibliography{msnid}

 \end{document}

-

-

-