Transcripts were assembled using the program SOAPdenovo (v1.04) after the removal of the adapter sequences. Paired-end reads areused again for gap filling of scaffolds to get sequences with least Ns and cannot be extended on either end. Blastx alignment (e-value < 0.00001) between Unigenes and protein databases like nr, Swiss-Prot, KEGG and COG is performed, and the best aligning results are used to decide sequence direction of Unigenes.