160 
Ganl. Bull. Singapore 70 (Suppl. 1) 2018 
Using tagged primers similar to the protocol used for COI barcoding, data was 
generated on a MiSeq sequencer. To cost-efficiently carry out genome skimming for 
~240 species, multiplexing was done by ligating a 20bp species-specific tag to the 
DNA of different species using a modified version of the Meyer & Kircher (2010) 
protocol. Three such “plant pools” libraries comprising of 75-100 species each with 
insert sizes of 400-900bp were prepared and sequenced on a HiSeq 2500 (250PE) 
platform. The sequence data were demultiplexed based on species-specific tags using 
SABRE (https://github.com/najoshi/sabre) and quality checked using custom scripts. 
Trimming of reads were carried out in CLC Genomics Workbench (Limit=0.001, 
https://www.qiagenbioinformatics.com) and assembled into chloroplast contigs using 
default parameters in MITOBIM (Hahn et al., 2013), by iterative mapping onto a 
closely-related reference chloroplast genome. Species reads were mapped to the 
MITOBIM-assembled contigs in CLC Genomics Workbench to calculate the average 
coverage of each chloroplast genome. 
Tree identifications via sapwood samples 
The suitability of all four plant barcodes for tree identification via sapwood material 
was assessed in preliminary experiments, which showed the ~400bp fragment of the 
ITS2 marker to be most effective at identification. However, PCR amplification and 
Sanger sequencing successes with this marker were low due to length variants (success 
rates of only -30%). Hence, we switched to using the trnL markers (short fragment of 
10-50bp) for sapwood-based identifications of the remaining samples unidentifiable 
with the previous marker. Between one to five PCR amplifications using tagged 
primers for the trnL marker were performed for each sample, and sequenced on an 
Illumina™ MiSeq Nano run. Sequence data obtained from the run were demultiplexed 
and binned into unique read clusters using PEAR (Zhang et al., 2014) and OBITOOLS 
(Boyer et al., 2016) respectively. Consensus sequences of each unique cluster were 
then matched against both the global and local plant trnL databases for identifications 
via blastn (BLAST 2.2.28+, Camacho et al ., 2009). 
Specimen imaging and online database 
Photography and image preparation. The specimens are kept at the Lee Kong Chian 
Natural History museum (specimens in main collection and DNA in cryo-collection). 
One specimen per species was imaged using a high-resolution photomacrography 
system (Visionary Digital™ Lab Plus System). Specimens were imaged under high 
magnification at different focal depths and exported via Adobe Lightroom. These 
images were then digitally stacked into a completely focused composite image using 
Helicon Locus Pro. The composite images were then digitally optimised in Photoshop 
CS5 Extended by white-balancing, image sharpening, light/shadow adjustments, and 
digitally removing impurities from background and specimens. Depending on the 
taxon, specimens were imaged in different orientations and magnifications to illustrate 
key diagnostic features. These separate images were then digitally stitched into an 
image plate. 
