2.3. Transcriptome assembly

Illumina sequencing generated 117,478,076 paired-end reads from the pooled cDNA libraries (Table 1). Sequence filtering was performed in CLC Genomics Workbench 12 software (Qiagen, Denmark). Adapter sequences, low quality reads (Phred score ≤ 30) and reads shorter than 50 bp were removed. The resulting filtered reads were assembled using the de Bruijn graph-based de novo assembler of CLC Genomics [2]. Assembly parameters: k-mer and bubble size were varied to optimize the assembled contigs. The final assembly (minimum contig length = 200 bp) was done with k-mer = 35, and bubble size = 300, which was based on the output parameters: high N50, low total number of contigs, high average contig length and high percentage of reads mapped back to transcripts. The cluster tool cd-hit-est with a sequence identity threshold of 0.95 was used for redundancy filtration of the assembly [3]. Numbers of reads mapping back to the contigs were converted to transcripts per million (TPM) expression values [4] to estimate the transcript abundance. In the initial data investigation, we also performed a principal component analysis and global Pearson correlation analysis to test the significance of the clusters and correlation between samples.