Overview
Transcriptome assembly is the computational reconstruction of expressed transcript sequences from RNA-seq reads, performed either with or without a reference genome. For organisms lacking a sequenced genome, de novo transcriptome assembly is the only option, providing the first view of an organism’s coding potential. Even when a reference genome is available, transcriptome assembly can capture novel isoforms, fusion transcripts, and sequences from poorly assembled genomic regions. The assembled transcriptome serves as the foundation for downstream analyses including expression quantification, functional annotation, and comparative studies.
Methods
De novo transcriptome assembly uses assembly algorithms designed for uneven coverage and alternative splicing. Popular tools include Trinity, which uses a de Bruijn graph approach with multiple k-mer sizes; rnaSPAdes, adapted from genome assembly; and SOAPdenovo-Trans. These tools assemble reads into contigs representing transcript fragments, then cluster related contigs into isoform groups and resolve full-length transcripts. Reference-guided assemblers (StringTie, Cufflinks) leverage splice-aware alignments to the genome and assemble overlapping reads into transcript models. Key quality metrics include assembly completeness (BUSCO scores against conserved orthologs), N50 length, and the number of full-length transcripts recovered. Redundancy reduction using CD-HIT or Corset clusters highly similar transcripts.
Applications
Transcriptome assembly enables gene discovery in non-model organisms, from agricultural crops to underexplored marine species. It identifies differentially expressed genes, tissue-specific isoforms, and fusion transcripts in cancer. The technique is essential when RNA sequencing data come from organisms without a reference, and it integrates deeply with next-generation sequencing workflows. Assembled transcriptomes also contribute to evolutionary studies by enabling cross-species comparisons of RNA structure and types. As long-read sequencing (Iso-Seq, Oxford Nanopore) improves, hybrid assembly strategies combining short and long reads are producing more complete and accurate transcriptomes than ever before.
Practical Protocol
For de novo transcriptome assembly with Trinity, first ensure RNA-seq reads have undergone quality trimming with Trimmomatic. Run Trinity with the command: Trinity --seqType fq --max_memory 50G --CPU 8 --left reads_R1.fastq --right reads_R2.fastq --output trinity_out_dir. Trinity operates in three modules: Inchworm assembles k-mer graphs into contigs, Chrysalis clusters contigs into de Bruijn graphs representing full transcript variants, and Butterfly traces paths through these graphs to reconstruct isoform sequences. The output is Trinity.fasta containing assembled transcripts. Assess assembly quality with BUSCO against the appropriate lineage dataset: busco -i Trinity.fasta -l eukaryota_odb10 -o busco_out -m transcriptome. A complete BUSCO score above 80% indicates a high-quality assembly. Additional metrics include contig N50 (calculated with TrinityStats.pl Trinity.fasta) and the number of assembled transcripts matching known proteins via BLASTX against UniProt. Reduce redundancy with CD-HIT-EST -c 0.95 -i Trinity.fasta -o Trinity_nr.fasta to cluster isoforms at 95% identity. For reference-guided assembly with StringTie, first align reads with STAR: stringtie -p 8 -G annotation.gtf -o assembly.gtf aligned.bam. StringTie assembles transcripts using a network flow algorithm and can be run without a reference GTF for novel transcript discovery. Merge assemblies across samples with stringtie --merge to create a unified transcriptome. Quantify expression with stringtie -e -B and use prepDE.py to generate a count matrix for differential expression analysis.
