The droplet reader software counts positive and negative droplets by using a threshold of fluorescence between the well-defined populations of high and low fluorescence amplitude droplets. One particular TaqMan probe displayed a fluorescence amplitude for droplets containing amplicons larger than 660 bp that is too low to reliably discriminate between positive and negative droplets when templates are amplified with a 1 min elongation time. When this is the case, the average fluorescence amplitude for these amplicons cannot be calculated. Increasing the elongation time to two minutes increases the fluorescence amplitude of all droplets containing amplifiable template (Figure 2). This enables the acquisition of accurate concentration and fluorescence amplitude data for longer templates, but the slope of the relationship between amplicon size and fluorescence amplitude is decreased (from m = -11.66 to m = -9.12), which decreases the ability to resolve small differences in amplicon size (Figure 2). Decreasing the elongation time to 30 s increases the resolution of the relationship between amplicon size and fluorescence amplitude, but prevents targets longer than 460 bp from amplifying to the point that they fluoresce detectably above the background fluorescence (Figure 2). This is likely due to the fact that longer products require more time for complete polymerization of nascent strands to occur. Thus, there is a tradeoff between the resolution and range of QuantiSize, though the assay can be easily adjusted to fit particular experimental needs.
To validate the use of the QuantiSize assay for the sizing and quantification steps in library preparation for NGS, we compared the quantity and size distribution of library DNA predicted by the QuantiSize assay to the quantity and size distribution of reads generated by the Illumina MiSeq platform. Eight uniquely indexed test libraries were generated by ligating DNA sheared to an average size of 150 bp onto MiSeq compatible adapter sequences similar to those used to create the aforementioned size standards. The libraries were run in individual wells of a ddPCR experiment alongside the set of size standards. Using the concentrations measured by ddPCR, the 8 uniquely indexed libraries were diluted and combined in a molar ratio of 100:50:10:1 with 2 libraries at each concentration. The observed number of MiSeq reads containing each index was compared with the expected number of copies of each uniquely indexed library loaded onto the MiSeq. The observed ratio of the number of reads containing each index very closely matched the expected ratio of 100:50:10:1 that was measured by ddPCR and the correlation between the expected and actual number of library molecules gave an R2 value of 0.9693 (Figure 3A).
The Nextera XT DNA Sample Preparation Kit was used to prepare a sequencing library from genomic DNA extracted from the human colon cancer cell line HCT 116. In lieu of the optional bead normalization step in the Nextera XT protocol, the concentration and size distribution of the library were measured with QuantiSize and the library concentration was adjusted to the proper concentration based on this measurement. The process of quantifying, normalizing, denaturing, and loading the library onto the MiSeq was repeated three times to demonstrate the precision of QuantiSize for predicting cluster density. The target cluster density for each MiSeq run was 1.0×106 clusters/mm2 (1.028×106 clusters/mm2 including the phi X control DNA). The mean cluster density ± SEM obtained was 1.039 ± 0.053×106 clusters/mm2 (Figure 3B). The mean number of reads ± SEM obtained was 1.813 ± 0.070×107 (Figure 3C). There are several potential sources of error in the MiSeq sample loading process including the pipetting of small volumes, variability in the efficiency of the denaturation reaction, variability of flow cell surface area, and user error. These factors may account for some of the observed variance in cluster density. However, even with the potential error caused by these factors, the target cluster density was achieved with high precision using the QuantiSize method.