The article introduces a novel method called Reference Annotation-Based Transcript Assembly (RABT) for identifying novel transcripts in annotated genomes using RNA-Seq data. RABT leverages existing annotations to improve the assembly of transcripts, particularly those with low expression levels that are poorly covered by sequencing reads. The method involves generating faux-reads from reference transcripts to fill in gaps in coverage and then using a spliced read mapper to align these faux-reads and sequencing reads to the genome. The Cufflinks assembler is modified to incorporate these faux-reads, resulting in a more comprehensive and accurate assembly of transcripts. The authors demonstrate the effectiveness of RABT by comparing it to the original Cufflinks assembler on human and Drosophila melanogaster datasets, showing that RABT can identify more novel isoforms and improve transcript lengths. The method is implemented in the Cufflinks software suite, which is freely available under the BOOST license.The article introduces a novel method called Reference Annotation-Based Transcript Assembly (RABT) for identifying novel transcripts in annotated genomes using RNA-Seq data. RABT leverages existing annotations to improve the assembly of transcripts, particularly those with low expression levels that are poorly covered by sequencing reads. The method involves generating faux-reads from reference transcripts to fill in gaps in coverage and then using a spliced read mapper to align these faux-reads and sequencing reads to the genome. The Cufflinks assembler is modified to incorporate these faux-reads, resulting in a more comprehensive and accurate assembly of transcripts. The authors demonstrate the effectiveness of RABT by comparing it to the original Cufflinks assembler on human and Drosophila melanogaster datasets, showing that RABT can identify more novel isoforms and improve transcript lengths. The method is implemented in the Cufflinks software suite, which is freely available under the BOOST license.