(I also asked this question on Biostars here.)

I'm stumped. I'm trying to plot a few transcripts at the same time, given transcript names and a TxDb. These are examples of approaches I've tried:

# ------------------------------------------------------------------------------
# Setup:
library(TxDb.Hsapiens.UCSC.hg19.knownGene)
library(Gviz)
txdb <- TxDb.Hsapiens.UCSC.hg19.knownGene

# ------------------------------------------------------------------------------
# Try 1:
gr <- GenomicFeatures::exons(
  txdb,
  vals = list(tx_name = c("uc001aaa.3", "uc010nxq.1")),
  columns = list("EXONNAME", "TXNAME", "GENEID"))
track <- Gviz::GeneRegionTrack(gr)
Gviz::plotTracks(track)
# Creates a plot, but doesn't show transcript grouping
​

# ------------------------------------------------------------------------------
# Try 2
gr <- GenomicFeatures::transcripts(
  txdb,
  vals = list(tx_name = c("uc001aaa.3", "uc010nxq.1")),
  columns = list("EXONNAME", "TXNAME", "GENEID"))
track <- Gviz::GeneRegionTrack(gr)
Gviz::plotTracks(track)
# Creates a plot, but has no exon/intron information

# ------------------------------------------------------------------------------
# Try 3
gr <- exonsBy(txdb, by = "tx", use.names=TRUE)[c("uc001aaa.3", "uc010nxq.1")]
track <- Gviz::GeneRegionTrack(gr)
# Error in .fillWithDefaults(DataFrame(chromosome = as.character(seqnames(range)),  :
# Number of elements in argument 'feature' is invalid

None of these work for me. I want to display exon/intron-structures and have arrows between exons. Does anyone have suggestions?

I found a solution that worked! Typical to find it shortly after asking the question, but here it is:

# ------------------------------------------------------------------------------
# Try 4
gr <- exonsBy(txdb, by = "tx", use.names=TRUE)[c("uc001aaa.3", "uc010nxq.1")]
gr <- unlist(gr)
elementMetadata(gr)$transcript <- names(gr)
track <- Gviz::GeneRegionTrack(gr)
Gviz::plotTracks(track)

Using exonsBy gives a GrangesList with one GRanges object for each transcript. Unlisting the GrangesList creates a named (important!) GRanges object with all exons. Using those names to set the $transcript metadata column solved the problem.