The pluto problem in protein annotation

The human transcriptome undergoes extensive translation beyond the portion annotated as protein-coding. Especially, thousands of "non-canonical" translations have been identified using Ribo-seq, hundreds of which now have conventional proteomics or immunopeptidomics support. Should reference annotation projects such as Ensembl-GENCODE, UniProtKB and HGNC add these to the protein "canon"? Here, we see a fascinating parallel with astronomy. In 2006, following lively debate, it was considered that Pluto is unusual compared to the other eight planets, and one of many similarly small objects now observed in the solar system. Thus, after 75 years as a canonical planet, it was reclassified as a "dwarf planet". The vast majority of these newly detected non-canonical translations are also unorthodox; especially, the vast bulk are very small, and lack obvious signatures of protein evolution. Furthermore, many are thus far detected only in cancer samples or cell lines, suggesting that some might be aberrant molecules; others are transposon-derived. Pertinently, we observe that these characteristics are shared by reasonable numbers of canonical proteins, such as HMHB1 and MYEOV. How should annotation projects proceed into this space? We aim to start a similar discussion in our community, noting that decisions taken will have profound implications. In fact, just as the astronomists were led to reconsider the question of what actually is a planet, it may be necessary to rethink what we mean by a protein. Is it simply a molecule built from amino acids that can be detected in the cell? Or should the term carry a deeper meaning, invoking knowledge of actual physiology? Since most non-canonical translations remain functionally obscure, the question is then how to annotate when such knowledge is unavailable. At present, our projects are reluctant to annotate these non-canonical translations as proteins en masse; this would significantly inflate the catalog of protein-coding genes, while functional uncertainties would pass on as ambiguities to our users. However, it would be a mistake not to annotate them at all. We emphasise that numerous non-canonical translations have now turned out to encode bona fide functional microproteins, and we assume that more will be elucidated once studied in more detail. Meanwhile, a standardised annotation of aberrant translation products could be a boon in the study of disease; they are potential diagnostic markers or even therapeutic targets. Thus, we favour the dwarf planet option, whereby non-canonical translation products are classified as something else, while existing canonical proteins with such characteristics are reclassified. This would demand the creation of a new system of protein annotation and nomenclature, and a profound expansion and reorganisation of our catalogs. We will put proposals forward for discussion.