Exploiting single-molecule transcript sequencing for eukaryotic gene prediction

Table 5 Sensitivity and precision of predicted genes after applying different settings

Setting	Sensitivity in %^a	Precision in %^b
Default A. thaliana parameter set	71.7	42.4
B. vulgaris trained parameter set^c	82.3	58.9
+ Hint masking	82.9	62.3
+ Hint masking enforcement	83.6	71.8
+ Additional mRNA-seq hints	76.4	44.6
+ mRNA-seq noise reduction	84.7	73.5
+ Higher weighting of introns	85.0	73.9
+ SMRT reads as anchors^d	91.1	77.9

Settings marked by ‘+’ were added to the setting of the previous line
^aPercent of correctly predicted transcripts in the set of SMRT derived test genes not overlapping the training gene set
^bPercent of wrongly predicted genes of all correctly and wrongly predicted gene models in genomic regions of SMRT derived test genes
^cTraining based on SMRT and manually validated genes
^d‘SMRT reads as anchors’ only affected genes covered by SMRT sequences

ISSN: 1474-760X