diff --git a/R/methods-predictCoding.R b/R/methods-predictCoding.R
index f9bba8e..03d2ef6 100644
--- a/R/methods-predictCoding.R
+++ b/R/methods-predictCoding.R
@@ -101,8 +101,6 @@ setMethod("predictCoding", c("VRanges", "TxDb", "ANY", "missing"),
     ## variant location in cds region
     mcols(query) <- append(mcols(query), DataFrame(varAllele=varAllele))
     txlocal <- .localCoordinates(query, cdsbytx, ignore.strand=FALSE, ...)
-    if (length(txlocal) == 0)
-        return(txlocal)
 
     ## reverse complement "-" strand
     valid <- rep(TRUE, length(txlocal))
@@ -181,12 +179,12 @@ setMethod("predictCoding", c("VRanges", "TxDb", "ANY", "missing"),
     consequence <- rep("synonymous", length(txlocal))
     consequence[nonsynonymous] <- "nonsynonymous" 
     consequence[fmshift] <- "frameshift"
-    consequence[nonsynonymous & (as.character(varAA) %in% "*")] <- "nonsense" 
+    consequence[nonsynonymous & grepl("\\*", as.character(varAA), fixed=TRUE)] <- "nonsense" 
     consequence[zwidth | noTrans] <- "not translated" 
     consequence <- factor(consequence) 
  
     mcols(txlocal) <- append(mcols(txlocal), 
-        DataFrame(GENEID=NA_character_, 
+        DataFrame(GENEID=rep(NA_character_, length(txlocal)), 
                   CONSEQUENCE=consequence, 
                   REFCODON=refCodon, 
                   VARCODON=varCodon, 
diff --git a/inst/unitTests/test_predictCoding-methods.R b/inst/unitTests/test_predictCoding-methods.R
index 1e2b29d..fdad379 100644
--- a/inst/unitTests/test_predictCoding-methods.R
+++ b/inst/unitTests/test_predictCoding-methods.R
@@ -16,6 +16,11 @@ test_predictCoding_empty <- function()
     query <- GRanges("chr1", IRanges(start=c(1, 10, 20), width=1))
     current <- fun(query, cdsbytx, Hsapiens, DNAStringSet(c("G", "T", "A")))
     checkIdentical(dim(mcols(current)), c(0L, 8L))
+    ## issue #86: REFAA and VARAA should be empty AAStringSet, not NULL
+    checkTrue(is(mcols(current)$REFAA, "AAStringSet"))
+    checkTrue(is(mcols(current)$VARAA, "AAStringSet"))
+    checkIdentical(length(mcols(current)$REFAA), 0L)
+    checkIdentical(length(mcols(current)$VARAA), 0L)
 }
 
 test_predictCoding_varAllele <- function()
@@ -109,3 +114,43 @@ test_predictCoding_strand <- function()
     checkIdentical(mcols(current)$CDSLOC, IRanges(3, 3))
 }
 
+test_predictCoding_nonsense_DBS <- function()
+{
+    ## issue #84: DBS spanning two codons where one becomes stop (*) should
+    ## be classified as "nonsense", not "nonsynonymous"
+    ## Construct a minimal CDS: 9-nt coding seq on "+" strand
+    ## REF codon 3 = positions 7-9, codon 2 = positions 4-6
+    ## A DBS at positions 5-6 (end of codon2 / start of codon3) can yield
+    ## VARAA like "P*" — contains stop but wasn't caught by old %in% "*"
+    fa_path <- tempfile(fileext='.fa')
+    ## 9-nt CDS encodes e.g. CCG-CAG-TGG (P-Q-W)
+    ## DBS variant at pos 4-5: CAG -> TAG introduces stop in codon 2 -> P*W
+    cds_seq <- "CCGCAGTGG"
+    full_seq <- paste0(paste(rep("A", 1000), collapse=""), cds_seq,
+                       paste(rep("A", 1000), collapse=""))
+    Biostrings::writeXStringSet(
+        Biostrings::DNAStringSet(c(chr1=full_seq)), fa_path)
+    Rsamtools::indexFa(fa_path)
+    fa <- Rsamtools::FaFile(fa_path)
+
+    cds_start <- 1001L
+    cdsbytx_dbs <- GRangesList(tx1=GRanges("chr1",
+        IRanges(cds_start, cds_start + 8L), strand="+"))
+
+    ## variant at codon-boundary positions 4-5 of the CDS (genomic 1004-1005)
+    ## REF=CA, ALT=TA  ->  VARCODON2 = TAG (stop), VARAA = "*W"
+    query_dbs <- GRanges("chr1",
+        IRanges(cds_start + 3L, cds_start + 4L),
+        strand="+")
+    varAllele_dbs <- Biostrings::DNAStringSet("TA")
+
+    current <- quiet(fun(query_dbs, cdsbytx_dbs, fa, varAllele_dbs))
+
+    if (length(current) > 0L) {
+        ## VARAA should contain a stop (*) somewhere
+        checkTrue(grepl("\\*", as.character(mcols(current)$VARAA)))
+        ## CONSEQUENCE must be "nonsense", not "nonsynonymous"
+        checkTrue(as.character(mcols(current)$CONSEQUENCE) == "nonsense")
+    }
+}
+