Then, we trimmed the low quality bases with the PERL script of fastq_qualitytrim_window.pl from http://xyala.cap.ed.ac.uk/Gene_Pool/scripts.tar.gz.tar.gz
After trimming, the average quality score at each base position was greater than 30 (with the majority being closer to 40).
After trimming the low quality bases using in-house built script, the spliced mapping process was performed by TopHat for each sample against to both genome and transcriptome references.
We obtained high-quality reads after trimming the low-quality bases from the 3′ end and the adapter sequences from the 5′ end using the NGS QC toolkit [ 70].
Using a series of normalization, correction and quality-filtering algorithms, the 454 sequencing data were processed to screen and filter for weak signals and low-quality reads, and to trim the read ends for low-quality and 454 adaptor sequences.
Since low-quality nucleotides from the ends of reads may lead to incorrect assembly outputs [ 19], we trimmed the low-quality or ambiguous nucleotides at both ends of the reads.
After trimming the low-quality parts and removing reads of less than 50 bp, 436.9 million reads (42.9 Gb) remained.
After trimming the low-quality part at the 3' end of each read and filtering out singleton reads and reads less than 25 bp in length, 212.6 million PE reads (16.2 Gb) were used for downstream analyses (Table 1).
The reads were trimmed using the quality score limit of 0.08 and maximum limit of 2 ambiguous nucleotides.
The performance and accuracy of these tools was evaluated using Illumina NGS data [ 50], which revealed that trimming increases the quality of subsequent read alignment, assembly or single nucleotide polymorphism calling.
Further, reads were trimmed using 'Trim the reads by quality' (version 1.2.2) tool (Phred quality cutoff of 20 and minimum read length of 40 nucleotides).
