Under-the-radar dengue virus infections in natural populations of Aedes aegypti mosquitoes

Sean M. Boyles; Carla N. Mavian; Esteban Finol; Maria Ukhanova; Caroline J. Stephenson; Gabriela Hamerlinck; Seokyoung Kang; Caleb Baumgartner; Mary Geesey; Israel Stinton; Kate Williams; Derrick K. Mathias; Mattia Prosperi; Volker Mai; Marco Salemi; Eva A. Buckner; John A. Lednicky; Adam R. Rivers; Rhoel R. Dinglasan

The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.01.24.919282 doi: bioRxiv preprint 8 (https://sourceforge.net/projects/bbmap/) against the Ae. aegypti Liverpool genome 47 (AaegL5.1). Non-mosquito reads were assembled using Spades (3.11.1) in metagenomics 48 mode [8] . For each contig, local similarity search in protein space was run using Diamond 49 (0.9.17) [9] against the NCBI NR database. Reads were mapped against assemblies using 50 Bowtie (2.3.4.1) [10] , then sorted/indexed using Samtools (1.4.1) [11] . Megan 6 [12] was used 51 to assign contigs and read counts to the lowest common ancestor (LCA) and to view viral 52 contigs. To estimate microbial community abundance, Diamond (0.9.17) [9] was used to 53 search reads against the NCBI NR database, Megan 6 [12] was used to assign read counts to 54 the LCA, and R (3.6.0) package Compositions [13] (1.40-2) was used to create a sub-55 composition of RNA (Fig. 1b) . Compositional count data from the Megan [12] LCA 56 classification was assessed by ALDEx2 [14, 15] to estimate the statistical significance of the 57 change in DENV4 reads from 2016 to 2017. ALDEx2 [14, 15] uses a Dirichlet multinomial 58 Monte Carlo simulation to estimate the variance of the centered log ratio (CLR) values for taxa 59 amongst the reads. Using the variance of the CLR, ALDEx2 [14, 15] computes P-values 66 Two contigs covering most of the genome with a small gap were obtained. To create a closed 67 genome, a dataset of genomes for DENV1-4 (NC_001477.1, NC_001474.2, NC_001475.2, 68 NC_002640.1) and the two assembled contigs were used. We selected reads sharing a 31-69 mer with the dataset using BBduk (https://sourceforge.net/projects/bbmap/), followed by 10 assessed by calculating the Effective Sampling Size (ESS) of parameter estimates. The HKY 95 substitution model [28] was used with empirical base frequencies and gamma distribution of 96 site-specific rate heterogeneity. The fit of strict vs. relaxed uncorrelated molecular clock 97 models, and constant size vs. Bayesian Skyline Plot [29] demographic models were tested.