to BioTechniques free email alert service to receive content updates.
Improving the limit of detection for Sanger sequencing: A comparison of methodologies for KRAS variant detection
Colin J. Davidson1, Emily Zeringer4, Kristen J. Champion2, Marie-Pierre Gauthier1, Fawn Wang1, Jerry Boonyaratanakornkit3, Julie R. Jones2, and Edgar Schreiber1
Full Text (PDF)
Supplementary Material

Figure 1.  Comparison of sequencing reaction electropherograms derived from the analysis of KRAS wild type and c.38G>A (p.G13D) mutant gDNA mixtures of 50%, 20%, 10%, 5% (mutant/wild type). (Click to enlarge)

Analysis of a sequencing reaction as a fragment file (.fsa) also benefits from the greater user control over peak detection that is possible with genotyping software compared with sequencing software. In addition to automated allele detection by fragment size and dye color, the advanced peak detection algorithm in GeneMapper software allows user control over peak amplitude thresholds that can be set individually for each dye color. This allows the user to determine the peak height at which the software detects peaks of interest but eliminates noise with the software reporting to the user only those peaks with heights that are at least the peak amplitude threshold for that dye. Additional parameters, such as polynomial degree and peak window size, can be used to adjust the sensitivity of peak detection. It is advisable that the above parameters be empirically determined to optimize the peak detection sensitivity for a particular sequencing reaction.

Using purified gDNA mixtures extracted from KRAS mutant and wild-type cell-lines (sample type 1), the limit of detection was improved to 5% for the detection of G>A transitions (c.34G>A, p.G12S; c.38G>A, p.G13D) as well as for G>T (c.34G>T, p.G12C; c.35G>T, p.G12V) and G>C transversions (c.35G>C, p.G12A), when the fragment approach (FASS) to the analysis of a FTSS reactions was used (Table 1 and Supplementary Table S2). This is compared with a limit of detection of 50% via software detection and 20% by visual inspection when the same G>A transition mutations were analyzed via sequencing software. For G>T and G>C transversions, the analytic sensitivity for FTSS was 50%–20% by software detection and 20%–5% via visual inspection. It is worth noting that the limit of detection for variant c.35G>T (p.G12V) was unchanged at 5% whether analyzed by visual inspection as a sequencing reaction or as a FASS reaction. The improved detection of the sequencing reaction for this variant is likely due to a combination of zygosity and copy number effects increasing the concentration of the mutant allele since the KRAS mutant cell-line used for c.35G>T (p.G12V) mixtures was found to be homozygous for c.35G>T mutant with a copy number of 3 (Supplementary Table S1). It is also worth pointing out that the detection of variant c.38G>A (p.G13D) is more challenging since the cell-line that was used was heterozygous with a normal copy number of two (Supplementary Table S1). As a consequence, the actual concentration of mutant allele in the mixture was half the expected amount which was consistent with the calculated percent mutant allele detected by the PS assay (Table 1).

Table 1.  Summary results for gDNA mixtures of KRAS mutant and wild-type cancer cell-lines (sample type 1). (Click to enlarge)


When analyzed as a fragment analysis reaction, the analytic sensitivity of FTSS improves to a level that is comparable to the PS and HRM approaches for the same mutations (Table 1 and Supplementary Table S2). For G>A transitions, the limit of detection was found to be 10%–2% and 20%–10% for the PS and HRM assays, respectively. The limit of detection observed in this study for cell-line mixtures of the c.38G>A (p.G13D) variant is similar to the previously reported values of 15%–20%, 5%, and 10% for FTSS, PS, and HRM, respectively (13). For G>T transversions, the analytic sensitivity of PS and HRM assays was 5% with the PS assay demonstrating a 5% limit of detection for the G>C transversion mutant c.35>C (p.G12A).

The FASS, PS, and HRM approaches constitute a group of assays with equivalent analytic sensitivity (10%–5%); a second grouping of more sensitive assays (2%–1%) consists of the SBE, STA, and qPCR assays. The limit of detection for a G>A transition mutation (c.38G>A, p.G13D) was found to be 1% for SBE, STA, and qPCR (Table 1). The 1% analytical sensitivity for G>A transition mutations was confirmed for the SBE and STA assays using variant c.34G>A (p.G12S) (Supplementary Table S2). An analytical sensitivity of 1%–0.5% was observed for SBE, STA, and qPCR assay applications for G>T transversion mutations. As indicated above, achieving a limit of detection of 0.5% is an overestimate since the mutant cell line used for the c.35G>T (p.G12V) mixtures was found to be homozygous for the mutant allele with a copy number of 3 (Supplementary Table S1), thus the actual percentage of mutant allele in the mixture is >0.5%. The G>C transversion variant, c.35G>C (p.G12A), was found to be 2% and 1%, for the SBE and STA assays, respectively (Supplementary Table S2).

The second sample type used in this study, sample type 2, was FFPE mixtures of various percentages of KRAS mutant and wild-type cells from which gDNA was extracted prior to KRAS variant detection. A comparison of the results for the same mutation between sample type 1 (Table 1) and sample type 2 (Table 2) gives an indication of the robustness of the methodology used to detect that mutation since the fixation process used in the creation of FFPE blocks adversely influences the integrity of the gDNA in the sample. Again, the assays distributed into two groups, with FTSS, FASS, PS, and HRM demonstrating limit of detection between 20%–5% depending on the variant while the SBE, STA and qPCR assays were found to have analytic sensitivities ranging between 5%–1%. For the G>A transition mutation variant (c.38G>A, p.G13D), the limit of detection was improved from 20% by visual inspection of the sequencing electropherogram to 10% when the FASS approach was used (Table 2). For this type of mutation the effect of sample source resulted in a decline in the analytic sensitivity for the FASS approach from 5% for this variant in the sample type 1 mixtures. For G>A transition variant c.38G>A (p.G13D), the analytic sensitivity dramatically declined for the PS assay to 20% (from 5% for sample type 1) and 20% for HRM (from 10% for sample type 1). Similar to the other methods investigated the analytic sensitivity for the G>A transition variant c.38G>A (p.G13D) decreased the most, with the STA and the qPCR assays detecting the 5% mixture (compared with 1% for sample type 1) and the SBE assay detecting the 2% mixture (versus 1% for sample type 1).

  1    2    3