Table 2.

Performance of Fulcrum during assembly of Illumina human-transcriptome data^a

Assembly statisticb	Assembly of trimmed raw readsc	Assembly of trimmed, collapsed readsc,d
Percent collapse	N.A.e	21
Number of contigs	63 931	63 009
N50	1380	1391
Maximum contig length	15 055	15 025
Mean contig length	679	687
velveth runtime	18 min	13 min
Maximum velveth memory usage	8.5 GB	8.0 GB
velvetg runtime	31 min	19 min
Maximum velvetg memory usage	7.5 GB	4.9 GB

^a 27 531 230 paired-end 76-bp Illumina sequence reads from a single lane (lane 5) (see http://hgdownload-test.cse.ucsc.edu/goldenPath/hg19/encodeDCC/wgEncodeCaltechRnaSeq/wgEncodeCaltechRnaSeqGm12878R2x75Il200FastqRd1Rep1.fastq.tgz and http://hgdownload-test.cse.ucsc.edu/goldenPath/hg19/encodeDCC/wgEncodeCaltechRnaSeq/wgEncodeCaltechRnaSeqGm12878R2x75Il200FastqRd2Rep1.fastq.tgz) were used for these trials.

^b Reads were assembled using Velvet/Oases (Velvet version 1.1.04 and Oases version 0.1.21) while monitoring memory and CPU usage by recording the output of top every 60 s.

^c Reads were filtered for quality and length. Briefly, reads were trimmed such that no nucleotide had a quality score <10, and no ambiguous nucleotides (Ns) remained in the read. Reads shorter than 45 bp were discarded.

^d Read pairs were collapsed allowing one mismatch of quality score 7 or higher per read pair, and the metadata provided by Fulcrum were discarded prior to assembly.

^e N.A., not applicable.