Fig. 1

satmut_utils design and performance benchmarking. Solid circles represent single- or multiple-nucleotide polymorphisms (SNPs, MNPs), which may be either true or false positives (errors). A Variant simulation workflow. With “sim,” ultra-low-frequency variants in Variant Call Format (VCF) are edited into pre-existing sequencing read alignments (BAM). Edited reads (FASTQ) and true positive variants (Truth) are output with expected counts and frequencies. The “call” workflow B extracts quality features during variant calling, which may be used for assay design validation, software parameter tuning, and machine learning-based error correction. B Variant calling workflow. SNPs and MNPs are identified and quantified in paired-end reads following optional preprocessing to improve specificity. Transcript nucleotide and protein changes are annotated and a VCF and fragment coverage bedgraph file are output. C Performance of MAVE variant callers. Two hundred eighty-one variants were simulated in alignments for a single amplicon in CBS, and performance measures were evaluated after applying two simple count filters. nt: nucleotide/codon-level calls; aa: amino acid-level calls. D Accuracy of variant count estimates. The expected count is the simulated truth count. One outlier SNP was excluded for visualization (Enrich2 log ratio: 4.88; satmut_utils and dms_tools2 log ratio: 0.87). dedup = deduplication