4.5.2. De Novo Transcriptome Assembly and Annotation

In order to achieve high-quality and clean data, the raw data was filtered by removing the reads with adapter sequence and low-quality reads using the fastp software [50], then the clean reads were assembled into transcriptome by the Trinity v.2.0.6 software (Broad Institute, Cambridge, MA, USA). Unigene function was annotated based on these databases: NCBI non-redundant protein sequences (Nr, https://ftp.ncbi.nlm.nih.gov/blast/db/FASTA/), KEGG (https://www.genome.jp/kegg), Clusters of Orthologous Groups of proteins (COG, https://www.ncbi.nlm.nih.gov/COG/), GO (https://www.geneontology.org), the Swiss-Prot (http://www.ebi.ac.uk/uniprot/) and the translation of EMBL (TrEMBL).

### *4.6. Determination of the Flavonoid Contents During Di*ff*erent Growth Stages*
