Simulated NGS read datasets for novel human virus prediction

<p>This repository contains simulated Illumina&nbsp;read datasets for novel human virus prediction and associated metadata extracted from the Virus Host Database (<a href="https://www.genome.jp/virushostdb/">https://www.genome.jp/virushostdb/</a>). The reads are 250bp long and were simulated with Mason (<a href="https://www.seqan.de/apps/mason/">https://www.seqan.de/apps/mason/</a>) from genomes downloaded from NCBI. The training-validation-test split was done on whole viral sequences to ensure &quot;novelty&quot; of validation and test viruses. The training sets contain 10 million reads per class, validation sets - 1.25 million reads per class, and test sets - 1.25 million paired reads per class. The negative class sets contain reads simulated from chordate-infecting (&quot;cho&quot;), metazoan-infecting (&quot;met&quot;), eukariote-infecting (&quot;euk&quot;) and all-nonhuman viruses. The positive class contains human-infecting viruses. The stratified dataset (&quot;strat&quot;) contains an equal number of reads from &quot;cho&quot;, &quot;met but not cho&quot;, &quot;euk but not met&quot; and &quot;all but not euk&quot;.&nbsp;</p> <p>Species-level datasets (&quot;humspec&quot;, &quot;allspec&quot; and &quot;chospec&quot;, with&nbsp;the corresponding fasta and *_species.rds files) are constructed analogously, but ensuring that all viruses of a given species were assigned to either training, val or test set. This is a stricter setting modelling a &quot;novel viral species&quot; scenario while reflecting within-species phenotype diversity.</p> <p>blast_hits.gz contains blast hits of human virome reads form Moustafa et al., 2017 (https://doi.org/10.1371/journal.ppat.1006292) blasted against our training database (see paper for details). In the second column you can find the matched label and the accession number of the matched reference. blast_labels_complete.gz contains extracted labels for all virome reads, including those without any matches. Note: one of the read headers (&gt;3c8ac47039d32b11c8fe23f588e444e9) from Moustafa et al. is slightly corrupted with null characters. You can remove them with&nbsp;sed &#39;s/\x0//g&#39; or equivalent.</p> <p>&nbsp;</p>