Fig. 3

Post-sequencing optimisation of multiplex bisulphite PCR assay. Boxplots show the range of sequencing coverage for individual amplicons in the a prostate cancer and b breast cancer panels. Light grey and blue boxes are used to depict the sequencing coverage before and after post-sequencing optimisation respectively. The purple background in a highlights the amplicons that originally failed sequencing (i.e. coverage < 100). The primers for these amplicons were pooled and re-amplified in an individual third multiplex panel. b For the breast cancer panels—the green background indicates the amplicons that originally failed sequencing and the red background indicates the amplicons that were originally amplified more than needed. For these primer pairs, primer concentrations were doubled (20 to 40 μM) or halved (20 to 10 μM) respectively to achieve more uniform amplicon coverage. The corresponding barplot shows the change in sequencing coverage before and after post-sequencing optimisation. c (i) PCR bias is introduced by PCR amplification of 2 example prostate amplicons (amplicon 46 and 35) using methylated-control DNAs (0%, 10%, 25%, 50%, 75%, 90%, 100%). Observed methylation after amplification (y-axis) is plotted against expected methylation levels (x-axis). Regression analysis was used to calculate a value of bias (b) as described by Warnecke et al. [28]. Red line = line of best fit from the regression; dotted line = line of best fit if data was unbiased (i.e. b = 1). (ii) The result of PCR-bias correction by regression on the methylated control DNA. The corrected methylation level (y-axis) is plotted against the expected methylation level (x-axis) showing that PCR bias has been effectively corrected. (iii) Multiplex bisulphite PCR methylation values for four biological samples are corrected for PCR bias based on the calculated bias value from (i) (light pink = LNCaP, light green = PrEC, violet = CAF, light brown = NPF). The corrected values are more similar than the uncorrected values to the same samples profiled by WGBS (iv)