Figure 4
(a) Schematic of the PCR simulator software used in this study. The software includes adding barcodes to molecules (‘labelling’), PCR amplification with a specified number of cycles, efficiency model and error rate and sampling and sequencing from the amplified pool. (b) Time taken to perform a full simulation, which includes initialisation, labelling initial molecules, PCR cycles (using a standard branching process model), sampling from the amplified pool and sequencing. Simulations are performed with the indicated PCR error rate (per base per cycle) and the given number of initial template molecules. Simulations consist of 15 cycles of PCR with efficiency 0.8, a sample size equal to the number of initial molecules being chosen from the amplified pool and sequencing with error rate 10−4. Data shown is the mean of 5 repeated simulations at each set of conditions, as measured on a 2.8 GHz Intel Core i7 MacBook Pro. (c) The distribution of the number of copies of each of 100,000 initial target molecules after 25 cycles of PCR at efficiencies of 0.85 (red), 0.9 (blue) or 0.95 (green). (d) The distribution of observed barcode family sizes (coloured lines) after simulating PCR cycles (25 cycles at 0.9 efficiency) on 100,000 initial molecules and then sampling from the amplified pool to select those molecules that are observed in the sequencer output. The number of molecules sequenced is expressed as a proportion (the ‘sample ratio’) of the number of initial molecules (100,000). The solid coloured lines are the mean of 5 repeated simulations and the dashed coloured lines are the expected distribution (a zero truncated Poisson with parameter equal to the sample ratio) if the sample was drawn from a uniformly distributed pool (which would occur if every initial molecule was uniquely barcoded and amplified identically). The black solid line is a representative example of the barcode family size distribution observed in TCR sequencing data from healthy volunteer PBMC.
