Can Deep Learning annotate Pandoravirus hypothetical proteins?

Original title: Deep Learning-based structural and functional annotation of Pandoravirus hypothetical proteins

Authors: Joseph L Horder,Abbie J Connor,Amy L Duggan,Joshua J Hale,Frederick J McDermott,Luke E Norris,Sophie JD Whinney,Shahram Mesdaghi,David L Murphy,Adam J Simpkin,Luciane V Mello,Daniel Rigden

This article explores the presence of genomic dark matter, or unknown genes, in giant viruses such as Pandoraviruses. These viruses contain a large number of hypothetical proteins, whose functions are not yet understood. To try to decipher their functions, the researchers utilize deep learning-based protein structure modeling. However, the current database lacks models for most viral proteins. To overcome this limitation, the researchers use a variety of predictive methods to make protein structure predictions for four Pandoraviruses. From these predictions, they are able to make strong functional predictions for several hypothetical proteins, including a nucleotidyltransferase involved in viral tRNA maturation and membrane channel sequences that may induce host cell membrane depolarization. They also identify homologues of potassium channel subunits and their likely counterparts in the Acanthamoeba cell. However, many other clusters of hypothetical proteins remain puzzling, suggesting that there is still much to learn about the structure and function of giant virus proteomes.

Original article: https://www.biorxiv.org/content/10.1101/2023.12.02.569716v1