YeATSAM analysis of the walnut and chickpea transcriptome reveals key genes undetected by current annotation tools

Background: The transcriptome, a treasure trove of gene space information, remains severely under-used by current genome annotation methods.  Methods: Here, we present an annotation method in the YeATS suite (YeATSAM), based on information encoded by the transcriptome, that demonstrates artifacts of the assembler, which must be addressed to achieve proper annotation.  Results and Discussion: YeATSAM was applied to the transcriptome obtained from twenty walnut tissues and compared to MAKER-P annotation of the recently published walnut genome sequence (WGS). MAKER-P and YeATSAM both failed to annotate several hundred proteins found by the other. Although many of these unannotated proteins have repetitive sequences (possibly transposable elements), other crucial proteins were excluded by each method. An egg cell-secreted protein and a homer protein were undetected by YeATSAM, although these did not produce any transcripts. Importantly, MAKER-P failed to classify key photosynthesis-related proteins, which we show emanated from Trinity assembly artifacts potentially not handled by MAKER-P. Also, no proteins from the large berberine bridge enzyme (BBE) family were annotated by MAKER-P. BBE is implicated in biosynthesis of several alkaloids metabolites, like anti-microbial berberine. As further validation, YeATSAM identified ~1000 genes that are not annotated in the NCBI database by Gnomon. YeATSAM used a RNA-seq derived chickpea ( Cicer arietinum L.) transcriptome assembled using Newbler v2.3.  Conclusions: Since the current version of YeATSAM does not have an ab initio module, we suggest a combined annotation scheme using both MAKER-P and YeATSAM to comprehensively and accurately annotate the WGS.