Vorpal: A novel RNA virus feature-extraction algorithm demonstrated through interpretable genotype-to-phenotype linear models

Phillip Davis; John Bagnoli; David Yarmosh; Alan Shteyman; Lance Presser; Sharon Altmann; Shelton Bradrick; Joseph A. Russell

AKRATGKTGTTAATMAA motif appears in all five of the civet SARS assemblies in the 450 dataset. In the case of the camel isolates, the motif KGATGTTGTTARWCAAY, which is also 451 related to the one mentioned above, is another high coefficient predictor for human pathogenicity 452 and it appears in 231 of the 232 Camel-MERS instances in the training set. This motif also 453 appears in the emerging 2019-nCoV as noted in Table 2. 454 455 As for the obstacles for predictive efforts, there are many opportunities for improvements in the 456 collection and annotation of viral genomic data. In Table 1 Genbank records for the Influenza A training set members for "passage" annotations revealed 464 that 42.3% of the instances in that set contained such annotations for cell passage. In contrast, the 465 Influenza A test set members, which represents more recently generated data, only contained 466 "passage" annotations in 29.0% of those records. 467 468