Merge pull request #527 from GoekeLab/devel_pre

Fix xgboost related issues and update devel branch

Author: cying111

.github/workflows/check-bioc.yml

@@ -22,9 +22,9 @@
 
 on:
   push:
-    branches: [main, devel]
+    branches: [devel, devel_pre, devel_pre_v4]
   pull_request:
-    branches: [main, devel]
+    branches: [devel, devel_pre, devel_pre_v4]
 
 name: R-CMD-check-bioc
 
@@ -54,8 +54,8 @@ jobs:
       fail-fast: false
       matrix:
         config:
-          - { os: ubuntu-latest, r: '4.4.2', bioc: '3.20', cont: "bioconductor/bioconductor_docker:RELEASE_3_20", rspm: "https://packagemanager.rstudio.com/cran/__linux__/focal/latest" }
-          - { os: macOS-latest, r: '4.4.2', bioc: '3.20'}
+          - { os: ubuntu-latest, r: '4.5.2', bioc: '3.22', cont: "bioconductor/bioconductor_docker:RELEASE_3_22", rspm: "https://packagemanager.rstudio.com/cran/__linux__/focal/latest" }
+          - { os: macOS-latest, r: '4.5.2', bioc: '3.22'}
             ##- { os: windows-latest, r: '4.3', bioc: '3.18'}
           ## Check https://github.com/r-lib/actions/tree/master/examples
           ## for examples using the http-user-agent
@@ -107,16 +107,16 @@ jobs:
         uses: actions/cache@v3
         with:
           path: ${{ env.R_LIBS_USER }}
-          key: ${{ env.cache-version }}-${{ runner.os }}-biocversion-RELEASE-r-4.4.2-${{ hashFiles('.github/depends.Rds') }}
-          restore-keys: ${{ env.cache-version }}-${{ runner.os }}-biocversion-RELEASE-r-4.4.2-
+          key: ${{ env.cache-version }}-${{ runner.os }}-biocversion-RELEASE-r-4.5.2-${{ hashFiles('.github/depends.Rds') }}
+          restore-keys: ${{ env.cache-version }}-${{ runner.os }}-biocversion-RELEASE-r-4.5.2-
 
       - name: Cache R packages on Linux
         if: "!contains(github.event.head_commit.message, '/nocache') && runner.os == 'Linux' "
         uses: actions/cache@v3
         with:
           path: /home/runner/work/_temp/Library
-          key: ${{ env.cache-version }}-${{ runner.os }}-biocversion-devel-r-4.4.2-${{ hashFiles('.github/depends.Rds') }}
-          restore-keys: ${{ env.cache-version }}-${{ runner.os }}-biocversion-devel-r-4.4.2-
+          key: ${{ env.cache-version }}-${{ runner.os }}-biocversion-devel-r-4.5.2-${{ hashFiles('.github/depends.Rds') }}
+          restore-keys: ${{ env.cache-version }}-${{ runner.os }}-biocversion-devel-r-4.5.2-
 
       - name: Install Linux system dependencies
         if: runner.os == 'Linux'
@@ -339,7 +339,7 @@ jobs:
         if: failure()
         uses: actions/upload-artifact@v4
         with:
-          name: ${{ runner.os }}-biocversion-RELEASE-r-4.4.2-results
+          name: ${{ runner.os }}-biocversion-RELEASE-r-4.5.2-results
           path: check
 
       - uses: docker/build-push-action@v1

.github/workflows/lint.yaml

@@ -2,10 +2,9 @@
 # Need help debugging build failures? Start at https://github.com/r-lib/actions#where-to-find-help
 on:
   push:
-    branches: 
-      - devel
+    branches: [devel, devel_pre, devel_pre_v4]
   pull_request:
-    branches: [main, devel]
+    branches: [devel, devel_pre, devel_pre_v4]
 
 name: lint
 

.github/workflows/test-coverage.yaml

@@ -2,10 +2,9 @@
 # Need help debugging build failures? Start at https://github.com/r-lib/actions#where-to-find-help
 on:
   push:
-    branches: 
-      - devel
+    branches: [devel, devel_pre, devel_pre_v4]
   pull_request:
-    branches: [main, devel]
+    branches: [devel, devel_pre, devel_pre_v4]
 
 name: test-coverage
 

R/bambu-extendAnnotations-utilityCombine.R

@@ -300,23 +300,15 @@ makeUnsplicedTibble <- function(combinedNewUnsplicedSe,newUnsplicedSeList,
         } , BPPARAM = bpParameters))
     newUnsplicedTibble <- newUnsplicedTibble %>% 
         left_join(rowDataCombined, by =  "row_id") %>%
-        separate(row_id, c("sample","rcName"), sep = "\\-") %>%
-        mutate(sample_id = as.integer(gsub("s","",sample))) %>%
-        mutate(sample_name = colDataNames[sample_id]) %>%
-        select(-sample, -sample_id) %>%
         mutate(readCount_tmp = readCount) %>%
-        group_by(chr,strand, start, end, sample_name) %>%
+        group_by(chr,strand, start, end) %>%
         summarise(readCount = sum(readCount),
-                    geneReadProp = sum(geneReadProp),
-                    txScore = weighted.mean(txScore, readCount_tmp),
-                    txScore.noFit = weighted.mean(txScore.noFit, readCount_tmp)) %>%
-        group_by(chr, strand, start, end) %>% 
-        summarise(readCount = sum(readCount),
-                    maxTxScore = txScore,
-                    maxTxScore.noFit = txScore.noFit,
-                    NSampleReadCount = sum(readCount >= min.readCount), 
-                    NSampleReadProp = sum(geneReadProp >= 
-                                            min.readFractionByGene),
-                    NSampleTxScore = sum(maxTxScore > min.txScore.singleExon))
+                  maxTxScore = weighted.mean(txScore, readCount_tmp),
+                  maxTxScore.noFit = weighted.mean(txScore.noFit, readCount_tmp),
+                  NSampleReadCount = sum(readCount_tmp >= min.readCount), 
+                  NSampleReadProp = sum(geneReadProp >= 
+                                          min.readFractionByGene),
+                  NSampleTxScore = sum(txScore > min.txScore.singleExon))
+    
     return(newUnsplicedTibble)
 }

R/bambu-extendAnnotations-utilityExtend.R

@@ -130,7 +130,12 @@ recommendNDR <- function(combinedTranscripts, baselineFDR = 0.1, NDR = NULL, def
     equal[is.na(equal)] = FALSE
 
     #add envirnment so poly() works
-    attr(defaultModels$lmNDR[["terms"]], ".Environment") <- new.env(parent = parent.env(globalenv()))
+	if(is.null(defaultModels$lmNDR[["terms"]])){
+		frm <- defaultModels$lmNDR$call$formula
+        defaultModels$lmNDR[["terms"]] <- terms(as.formula(frm))
+        attr(defaultModels$lmNDR[["terms"]], ".Environment") <- new.env(parent = parent.env(globalenv()))
+	}
+    #attr(defaultModels$lmNDR[["terms"]], ".Environment") <- new.env(parent = parent.env(globalenv()))
     baseline = predict(defaultModels$lmNDR, newdata=data.frame(NDR=baselineFDR))
     attr(defaultModels$lmNDR[["terms"]], ".Environment") = c()
 

R/bambu-processReads_scoreReadClasses.R

@@ -158,9 +158,9 @@ getTranscriptScore = function(rowData, model = NULL, defaultModels){
         } else txScoreSE = NULL
     } else {
         if (!is.null(defaultModels)){
-            txScore = predict(defaultModels$transcriptModelME, 
+			txScore = predict(defaultModels$transcriptModelME, 
                 as.matrix(features))
-            txScoreSE = predict(defaultModels$transcriptModelSE, 
+			txScoreSE = predict(defaultModels$transcriptModelSE, 
                 as.matrix(features))
         } else {
             warning("Transcript model not trained. ",
@@ -292,6 +292,7 @@ prepareTranscriptModelFeatures = function(rowData){
                       tx_strand_bias = readCount.posStrand, labels = equal) %>%
         mutate(
             tx_strand_bias=(1-abs(0.5-(tx_strand_bias/numReads))),
+			labels = as.logical(labels),
             numReads = log2(pmax(1,1+(numReads/scalingFactor)))
         )
     return(outData)

R/bambu-processReads_utilityJunctionErrorCorrection.R

@@ -91,7 +91,7 @@ testSpliceSites <- function(data, splice = "Start", prime = "start",
         predSplice.prime <- NULL
         if (is.null(junctionModel)) { 
             model = fitXGBoostModel(labels.train = 
-                as.integer(annotatedSplice)[mySet.all][mySet.training], 
+                as.logical(annotatedSplice)[mySet.all][mySet.training], 
                 data.train = modelmatrix[mySet.training,],
                 show.cv = verbose, maxSize.cv = 10000)
 
@@ -210,11 +210,10 @@ fitXGBoostModel <- function(labels.train, data.train, nrounds = 50,
         data.train.cv.test <- data.train[-mySample,]
         labels.train.cv.test <- labels.train[-mySample]
 
-        cv.fit <- xgboost(data = data.train.cv, 
-            label = labels.train.cv, nthread = 1, nrounds = nrounds, 
+        cv.fit <- xgboost(x = data.train.cv, 
+            y = labels.train.cv, nthread = 1, nrounds = nrounds, 
             objective = "binary:logistic", 
-            eval_metric = 'error',
-            verbose = 0)
+            eval_metric = 'error')
         predictions <- predict(cv.fit, data.train.cv.test)
         message('prediction accuracy (CV) (higher for splice ',
                 'donor than splice acceptor)')
@@ -228,11 +227,10 @@ fitXGBoostModel <- function(labels.train, data.train, nrounds = 50,
         message("AUC: ", evaluatePerformance(labels.train.cv.test == 1,predictions)$AUC)
     }
 
-    cv.fit <- xgboost(data = data.train, 
-                      label = labels.train, nthread=1, nrounds=nrounds, 
+    cv.fit <- xgboost(x = data.train, 
+                      y = labels.train, nthread=1, nrounds=nrounds, 
                       objective = "binary:logistic", 
-                      eval_metric='error',
-                      verbose = 0)
+                      eval_metric='error')
 
     return(cv.fit)
 }

R/bambu.R

@@ -149,7 +149,7 @@ bambu <- function(reads, annotations = NULL, genome = NULL, NDR = NULL,
 
     emParameters <- setEmParameters(emParameters = opt.em)
     bpParameters <- setBiocParallelParameters(reads, ncore, verbose)
-
+    xgb.set.config(nthread = 1)
     rm.readClassSe <- FALSE
     readClassList = reads
     isRDSs = all(sapply(reads, class)=="RangedSummarizedExperiment")

R/bambu_utilityFunctions.R

@@ -4,9 +4,6 @@
 #' @importFrom BiocParallel bpparam
 #' @noRd
 setBiocParallelParameters <- function(reads, ncore, verbose){
-    if(ncore >= 2) message("WARNING - If you change the number of cores (ncore) ",
-    "between Bambu runs and there is no progress please restart your R session ",
-    "to resolve the issue that originates from the XGboost package.")
     bpParameters <- bpparam()
     #===# set parallel options: otherwise use parallel to distribute samples
     bpParameters$workers <- ifelse(length(reads) == 1, 1, ncore)