Bug Fix: Infer effect allele

DESCRIPTION

@@ -1,7 +1,7 @@
 Package: MungeSumstats
 Type: Package
 Title: Standardise summary statistics from GWAS
-Version: 1.9.17
+Version: 1.9.18
 Authors@R:
     c(person(given = "Alan",
            family = "Murphy",

NAMESPACE

@@ -8,6 +8,7 @@ export(formatted_example)
 export(get_genome_builds)
 export(import_sumstats)
 export(index_tabular)
+export(infer_effect_column)
 export(liftover)
 export(list_sumstats)
 export(load_snp_loc_data)

NEWS.md

@@ -1,3 +1,44 @@
+## CHANGES IN VERSION 1.9.18
+
+### Bug fix
+* Fixed column header mappings
+  * Made all uncorrected header names uppercase and removed duplicates
+  * "TOTALSAMPLESIZE" now maps to "N" instead of "NSTUDY"
+  * "MAJORALLELE", "MAJOR_ALLELE", "MAJOR-ALLELE", and "MAJOR ALLELE" now map to 
+    "A1" instead of "A2"
+  * Removed the mappings for "OR-A1", "OR.A1", "OR_A1", and "BETA1" because MSS 
+    assumes that A2 is the effect allele
+  * Removed mappings for "A1FREQ", "A1FRQ", "AF1", "FREQ.A1.1000G.EUR", 
+    "FREQ.A1.ESP.EUR", "FREQ.ALLELE1.HAPMAPCEU", "FREQ1", "FREQ1.HAPMAP", and 
+    "FRQ_A1" because MSS defines "FRQ" to be the allele frequency of A2
+  * Removed mappings for "CHR36", "BASE_GRCH36", "POSITION36", "POSGRCH36", 
+    "BASEGRCH36", "POS36", "POS GRCH36", "POS.GRCH36", "POS-GRCH36", and 
+    "POS_GRCH36" 
+    because MSS does not support the GRCh36 genome build
+  * Removed the ambiguous mapping "NMISS" -> "N" because "NMISS" can refer to 
+    the number of samples with missing data
+  * Removed the ambiguous mapping "WEIGHT" -> "N" because "WEIGHT" can refer to 
+    coefficient weights
+* Fixed inference of allele where ambiguous (A1, A2) naming used (see 
+  infer_effect_column.R for code) but in short:
+  * Three checks now made to infer which allele the effect/frequency information
+    relates to. See infer_effect_column.R for further details.
+  * See get_eff_frq_allele_combns.R for how effect/frequency columns that infer 
+    the allele are captured in the mapping file
+
+### New features
+* New column header mappings:
+  * "VARIANT_ID" and "RSIDS" --> "SNP"
+  * "P_BOLT_LMM" --> "P"
+  * "NCASES" --> "N_CAS"
+  * "N_EFFECTIVE", "N_INFORMATIVE", and "TOTAL_N" --> "N"
+  * "HET_P" --> "HETPVAL"
+  * "HET_ISQ" --> "HETISQT"
+  * "ALL_AF" --> "FRQ"
+  * "DIRECT" --> "DIRECTION"
+  * "ALT_EFFSIZE" --> "BETA"
+  * "INFORMATIVE_ALT_AC" --> "AC"
+
 ## CHANGES IN VERSION 1.9.17
 
 ### Bug fix

R/format_sumstats.R

@@ -445,6 +445,17 @@ format_sumstats <- function(path,
           mapping_file <- rbind(mapping_file,es_cols)
         }
 
+        #### Check 2:Check for effect direction ####
+        sumstats_return <-
+          infer_effect_column(
+            sumstats_dt = sumstats_return$sumstats_dt,
+            mapping_file = mapping_file,
+            dbSNP = dbSNP,
+            nThread = nThread,
+            ref_genome = ref_genome,
+            on_ref_genome = on_ref_genome
+          )
+
         #### Check 3:Standardise headers for all OS ####
         sumstats_return <-
             standardise_sumstats_column_headers_crossplatform(

R/get_eff_frq_allele_combns.R

@@ -0,0 +1,77 @@
+#' Get combinations of uncorrected allele and effect (and frq) columns
+#'
+#' @inheritParams format_sumstats 
+#' @inheritParams compute_nsize
+#' @param eff_frq_cols Corrected effect or frequency column names found in a 
+#' sumstats. Default of BETA, OR, LOG_ODDS, SIGNED_SUMSTAT, Z and FRQ.
+#' @return datatable containing uncorrected and corrected combinations
+#' @importFrom data.table setnames as.data.table := setkey rbindlist data.table
+get_eff_frq_allele_combns <-
+  function(mapping_file = sumstatsColHeaders, 
+           eff_frq_cols = c("BETA", "OR", "LOG_ODDS", "SIGNED_SUMSTAT","Z",
+                            "FRQ")) {
+    ### Add this to avoid confusing BiocCheck
+    CORRECTED <- UNCORRECTED <- Var1 <- Var2 <- NULL 
+    colnames(mapping_file) <- toupper(colnames(mapping_file))
+    #get allele associated effect/FRQ columns
+    #get all combinations with allele columns
+    eff_frq_cols_uncorrc <- 
+      mapping_file[mapping_file$CORRECTED %in% eff_frq_cols,]$UNCORRECTED
+    #join with all allele cols
+    allele_uncorrc <- 
+      mapping_file[mapping_file$CORRECTED %in% c('A1','A2'),]$UNCORRECTED
+    #get combinations
+    eff_frq_allele_dt <- 
+      data.table::as.data.table(expand.grid(eff_frq_cols_uncorrc, 
+                                            allele_uncorrc))
+    mapping_file_dt <- data.table::as.data.table(mapping_file)
+    data.table::setkey(mapping_file_dt,"UNCORRECTED")
+    data.table::setkey(eff_frq_allele_dt,"Var1")
+    #add corrected
+    eff_frq_allele_dt[mapping_file,CORRECTED:=CORRECTED,]
+    #now loop through every joining character and join with eff both before
+    #and after
+    joining_char <- c("","_",".","-"," ")
+    all_combns <- vector(mode="list",length = length(joining_char)*2)
+    counter <- 1
+    for(join_i in joining_char){
+      eff_frq_allele_dt_i <- copy(eff_frq_allele_dt)  
+      eff_frq_allele_dt_i[,UNCORRECTED:=paste0(Var1,join_i,Var2)]
+      all_combns[[counter]] <- 
+        eff_frq_allele_dt_i[,c("UNCORRECTED","CORRECTED")]
+      counter <- counter+1
+      #same for Var 2 in front
+      eff_frq_allele_dt_i <- copy(eff_frq_allele_dt)  
+      eff_frq_allele_dt_i[,UNCORRECTED:=paste0(Var2,join_i,Var1)]
+      all_combns[[counter]] <- 
+        eff_frq_allele_dt_i[,c("UNCORRECTED","CORRECTED")]
+      counter <- counter+1
+    }
+    #join all together
+    eff_frq_allele_matches <- data.table::rbindlist(all_combns)
+    #finally add some custom ones 
+    custom_adds <- data.table::data.table("UNCORRECTED" = 
+                                            c("BETA1", "BETA2","AF1","AF2",
+                                              "FREQ.A1.1000G.EUR",
+                                              "FREQ.A2.1000G.EUR",
+                                              "FREQ.A1.ESP.EUR",
+                                              "FREQ.A2.ESP.EUR",
+                                              "FREQ.ALLELE1.HAPMAPCEU",
+                                              "FREQ.ALLELE2.HAPMAPCEU",
+                                              "FREQ1","FREQ2", 
+                                              "FREQ1.HAPMAP","FREQ2.HAPMAP"),
+                                          "CORRECTED" = 
+                                            c("BETA", "BETA","FRQ","FRQ",
+                                              "FRQ",
+                                              "FRQ",
+                                              "FRQ",
+                                              "FRQ",
+                                              "FRQ",
+                                              "FRQ",
+                                              "FRQ","FRQ",
+                                              "FRQ","FRQ"))
+    eff_frq_allele_matches <- data.table::rbindlist(list(
+      eff_frq_allele_matches,custom_adds))
+
+    return(eff_frq_allele_matches)
+  }