18) and quality trimmed by using Trimmomatic version 0.36 ( 19). Viral contigs were generated by using default settings with Vicuna version 1.1 ( 20), and a de novo consensus assembly was generated by using Viral Finishing and Annotation Toolkit (V-FAT) version 1.1 (https://www.broadinstitute.org/viral-genomics/v-fat). Read data are available from the National Center for Biotechnology Information Read Archive (https://www.ncbi.nlm.nih.gov) under BioProject accession no. PRJNA661611. Phylogenetic Reconstruction 21) and IQ-TREE version 1.6 ( 22). We then constructed a maximum-likelihood phylogenetic tree with 1,000 ultrafast bootstrap replicates ( HCV Transmission Network Analyses We uploaded Illumina paired-end reads to GHOST and subjected them to automatic quality control criteria. In brief, read pairs were filtered out if a read had 3 Ns (N indicates that software was not able to make a basecall for this base) or a length 185 bp. Each identifier on forward and reverse reads was examined, and the pair was discarded if either identifier was not an exact match to a given list of valid identifiers. We discarded pairs containing valid identifiers if they were not a constituent of the majority identifier tuple. If 11% of the read pairs contained valid identifiers that were not the majority tuple, we discarded the entire sample without further processing. Random subsampling of 5,000–20,000 read pairs was undertaken, and primer sequences were located in each read, allowing for a combined error total of 3. Read pairs were discarded when the primer could not be found. Remaining read pairs were then unified in a single error-corrected sequence, and only those sequences with a nonsense-free reading frame were collapsed into unique occurences with associated frequencies. Further methodologic details on quality filtering can be found elsewhere (