LAMP-Seq: Population-Scale COVID-19 Diagnostics Using a Compressed Barcode Space

Jonathan L Schmid-Burgk; David Li; David Feldman; Mikolaj Slabicki; Jacob Borrajo; Jonathan Strecker; Brian Cleary; Aviv Regev; Feng Zhang

A standard Illumina NextSeq run can generate 200 million sequencing reads in 14 hours and we predict that this is sufficient for 100,000 patient samples per run, even accounting for library skewing due to differences in viral loads, largely because the vast majority of samples will be negative (for modeling see Supplementary Note 1). To explore parameters for using a compressed barcoding space, we conservatively assume that 1% of synthesized barcode primers systematically fail to work (∆ synth = 0.01), while additionally 5% of all sample-specific barcodes are not detected due to varying sequencing depth (∆ stoch = 0.05; this is independent of dropout due to low viral load). For automated assembly of testing reactions with unique barcode combinations, we anticipate that up to m = 10,000 barcode primers can be handled by available pipetting robots. Under these assumptions, we investigated for 100,000 samples what number of barcodes per sample (k), number of barcode primers total (m), and number of pools per run (m2) would allow for minimal false-positive and false-negative rates of detection ( Fig. 2A-C) .