The influence of transcript assembly on the proteogenomics discovery of microproteins

Proteogenomics methods have identified many non-annotated protein-coding genes in the human genome. Many of the newly discovered protein-coding genes encode peptides and small proteins, referred to collectively as microproteins. Microproteins are produced through ribosome translation of small open reading frames (smORFs). The discovery of many smORFs reveals a blind spot in traditional gene-finding algorithms for these genes. Biological studies have found roles for microproteins in cell biology and physiology, and the potential that there exists additional bioactive microproteins drives the interest in detection and discovery of these molecules. A key step in any proteogenomics workflow is the assembly of RNA-Seq data into likely mRNA transcripts that are then used to create a searchable protein database. Here we demonstrate that specific features of the assembled transcriptome impact microprotein detection by shotgun proteomics. By tailoring transcript assembly for downstream mass spectrometry searching, we show that we can detect more than double the number of high-quality microprotein candidates and introduce a novel open-source mRNA assembler for proteogenomics (MAPS) that incorporates all of these features. By integrating our specialized assembler, MAPS, and a popular generalized assembler into our proteogenomics pipeline, we detect 45 novel human microproteins from a high quality proteogenomics dataset of a human cell line. We then characterize the features of the novel microproteins, identifying two classes of microproteins. Our work highlights the importance of specialized transcriptome assembly upstream of proteomics validation when searching for short and potentially rare and poorly conserved proteins.