SNP targets: A total of 130 SNPs were randomly selected from dbSNP Build 144 (https://www.ncbi.nlm.nih.gov/projects/SNP/, Supplemental Table S1) for most of the analyses in this study, covering the types of SNPs found in the human genome. A set of 1000 randomly selected common bi-allelic SNPs were used to evaluate the pipeline design rate.

Synthetic templates: Synthetic gBlocks^® Gene Fragments (Integrated DNA Technologies, https://www.idtdna.com/pages/products/genes/gblocks-gene-fragments) were used as known genotype controls during the initial assay validation stage (data not shown). gBlocks Gene Fragments representing the wild type and mutant allele were mixed together in an equal molar ratio, representing the heterozygous genotype. Concentrations were determined using the Nano Drop™ 2000 (Thermo Fisher Scientific).

Genomic DNA samples: Human genomic DNA samples were purchased from the Coriell Institute for Medical Research. A total of 136 unique samples in two sets were obtained, representative of the three original HapMap populations: Yoruba from Ibadan (YRI), Han Chinese from Beijing, China (CHB), and CEPH Utah residents (CEU) (Supplemental Table S2). DNA samples were quantified using a qPCR method targeting the RNase P (RPPH1) gene.

SNP genotyping was carried out using rhAmp SNP Assays, rhAmp Genotyping Master Mix, and rhAmp Reporter Mix with or without a passive reference dye (www.idtdna.com/rhAmp-Genotyping). Genotyping reactions were performed typically using 3 ng of dried down genomic DNA from either 46 (human SNP panel) or 90 (ADME SNP panel) samples (Coriell Institute) in 5 µL reactions. Unless otherwise noted, reactions were run on the QuantStudio 7 Flex instrument and analysis was performed using the QuantStudio Real-Time PCR Software (Thermo Fisher Scientific, CA). For DNA input testing, the gDNA was reduced from 3 ng to 125 pg per reaction. rhAmp SNP genotyping reactions were run with a thermal cycling profile of 95˚C for 10 minutes, followed by 40 cycles at 95˚C for 10 seconds, 60˚C for 30 seconds, and 68˚C for 20 seconds per the published protocol (www.idtdna.com/rhAmp-SNP-protocol). The same thermal cycling program was used on the CFX384 (BioRad, CA) platform, and reactions contained rhAmp Reporter Mix without passive reference dye. The IntelliQube (LGC Group) data was collected using a thermal profile of 93.5˚C for 10 minutes, followed by 40 cycles at 93.5˚C for 10 seconds, 60˚C for 30 seconds, and 68˚C for 20 seconds. The reactions contained rhAmp SNP genotyping master mix at a final 0.85x concentration, rhAmp SNP assay and reporter mix with passive reference dye both at final 1X concentration, and Coriell gDNA input at 1.6 ng per reaction. With the Biomark HD (Fluidigm, CA) platform, 23 gDNA samples from a subset of 90 (ADME SNP panel) at 150 ng per inlet was used according to the manufacturer’s recommendation. A modified protocol without additional HotStar Taq polymerase was performed. The thermal mix protocol was modified to the following: 37˚C for 2 minutes, 45˚C for 10 minutes, then 25˚C for 10 minutes. The reactions were run with the thermal cycling at 95˚C for 10 minutes, followed by 40 cycles at 95˚C for 10 seconds, 60˚C for 30 seconds, and 68˚C for 20 seconds. The call rate was defined as the percent of samples with an assigned SNP genotype call, and the call accuracy was the percent of called samples with the correct genotype assigned. The reported call rate and call accuracy were determined using auto-calls assigned by the QuantStudio Real-Time PCR Software to each of 46 samples for 130 assays.

SNP genotyping was carried out using TaqMan Genotyping Master Mix (Thermo Fisher Scientific) and either TaqMan SNP Genotyping Assays or TaqMan Drug Metabolism Genotyping Assays (Thermo Fisher Scientific) (assay ID numbers listed in Supplemental Table S3). SNPs were assessed using 3 ng dried down genomic DNA samples in 5 µL reactions. TaqMan SNP Genotyping Assay reactions were run with the following thermal cycling profile: 95˚C for 10 minutes, followed by 40 cycles at 95˚C for 15 seconds and 60˚C for 1 minute. TaqMan Drug Metabolism Genotyping Assay reactions were run with the following thermal cycling profile: 95˚C for 10 minutes, followed by 50 cycles at 95˚C for 15 seconds and 60˚C for 90 seconds. Thermal cycling profiles were based on the manufacturer recommended protocols. Cycling and allelic discrimination analysis was performed using the QuantStudio 7 Flex instrument and QuantStudio Real-Time PCR Software Autocaller (Thermo Fisher).

Universal reporter system: Of 10 million candidate sequences screened, a small set of universal primers and probes that met stringent thermodynamic properties (identical length, GC%, Tm, etc.) and had high sequence specificity against the human genome were selected. We further narrowed the probe set so that they did not form dimers and the pairwise edit distance was at least 4 for minimal interference among probes. A pair of the universal reporter probes were selected based on empirical testing (data not shown).

rhAmp SNP assay design pipeline: We have developed a sophisticated assay design pipeline that performs the following steps for each target SNP: 1) design all possible allele-specific primers (ASP), locus-specific primers (LSP) and their combinations; 2) evaluate the thermodynamic properties of the primers and amplicons, such as length, percent GC, Tm, sequence complexity, folding, etc.; 3) assess the impact of overlapping SNPs and repeats; 4) analyze the dimerization of assay primers, including self- and hetero-dimers; 5) evaluate primer compatibility, i.e. having similar Tm and length, particularly among ASPs; 6) determine the assay specificity against the human genome. Finally, based on the best combination of all the above parameters, an assay with the best overall quality is selected.

The schematic describing rhAmp SNP genotyping is shown in Figure 1. Each rhAmp SNP assay is comprised of one locus-specific primer (LSP) and two tailed allele-specific primers (ASPs). Each primer has a single RNA base and a 3’ end

blocking modification. The RNA base in the ASP is located immediately downstream of the target SNP location (Figure 1(A)).

The blocked primers are efficiently activated by the RNase H2 enzyme only in the presence of a perfectly matched RNA-DNA heteroduplex. Once the RNase H2 cleaves the primer, the 3’ end is available for extension by the mutant Taq DNA polymerase (Figure 1(B)). Each ASP has a unique 5’ tail sequence that becomes incorporated into the allele-specific amplicon. The tail region adds the necessary sequences for the binding of a universal primer and fluorescent probe (Figure 1(C)). The assays are designed such that the reference allele gives a signal in the FAM dye channel, and the alternative allele signal is detected in the VIC^® dye channel. The alternate allele fluorophore is Yakima Yellow^®, which has a very similar excitation and emission spectra as VIC dye. With such a similar excitation and emission profile, no spectral re-calibration is required.

While the RNase H2 enzyme exhibits significant differences in the rate of enzymatic cleavage of an RNA base in the presence or absence of a mismatch RNA-DNA heteroduplex [2] , the rhAmp SNP assay also requires a DNA polymerase to complete allele-specific PCR. A novel mutant Taq DNA polymerase was developed to function effectively in the same buffer as the RNase H2 enzyme and provide improved allelic discrimination in comparison to the wild-type Taq DNA polymerase. Figure 2 shows the performance difference in the allelic discrimination (AD) plot between two master mixes, one containing the wild type Taq DNA polymerase (Figure 2(A)) and the other with mutant Taq DNA polymerase (Figure 2(B)), with an example assay designed to detect SNP rs2269829. A total of 46 different human gDNA samples were tested at 3 ng of gDNA input per reaction. With wild-type Taq DNA polymerase, the three genotype clusters can be automatically called, however, the cluster separation is relatively poor, likely due to ASP mispriming, and some of the no template control replicates drift from the origin. Replacing the Taq with the mutant version improves the separation, tightness, and angles of the genotype clusters.

Of 1000 common human SNPs randomly selected from the public dbSNP database, rhAmp SNP assay designs were successfully generated for 950 targets (95%). Functional test results from a randomly selected subset of 130 assays tested with 46 genomic DNA samples (human SNP panel) are summarized in Table 1. Two examples of rhAmp SNP genotyping assays targeting human SNPs, rs6068816 and rs4148946 are shown in Figure 3. Overall performance of rhAmp SNP assays shows 98% call rate and 99% accuracy (concordance with published genotypes in the NCBI database) while maintaining a high assay design rate.

Effect of DNA input on rhAmp SNP genotyping

DNA input can vary due to factors such as limited biological samples, purification loss, and other causes. Low sample input may impact the quality of SNP genotyping calls due to lower fluorescent signals. Therefore, it is important to determine the DNA input range that can provide robust genotyping calls. We tested the lower limit of input DNA for the rhAmp SNP system (Figure 4). A rhAmp SNP assay targeting a human SNP (rs4657751) was tested in the presence of gDNA at 0 (no template control or NTC), 0.125 (blue), 0.5 (green) and 3 (orange) ng per PCR reaction on the QuantStudio 7 Flex instrument. Genotypes for 46 individual gDNA samples are accurately auto-called with 0.5 and 3 ng gDNA and can be manually called with 0.125 ng gDNA. For some assays tested, accurate genotypes could be manually called using sample input as low as 25 pg

(data not shown).

Compatibility of rhAmp SNP genotyping with various qPCR platforms

All rhAmp SNP assays in this study were initially evaluated on the QuantStudio 7 Flex (Thermo Fisher) and CFX384 (BioRad) instruments using 3 ng genomic DNA or 1000 copies gBlocks Gene Fragments. To demonstrate the compatibility of small volume platforms, some of the assays were tested on the IntelliQube (LGC Group) and Biomark HD system (Fluidigm). An example of allelic discrimination plots for a human ADME SNP rs7668258 in the UGT2B7 gene is shown on four different platforms: QuantStudio 7 Flex (Figure 5(A)), CFX384 (Figure 5(B)), IntelliQube (Figure 5(C)) and Biomark HD (Figure 5(D)). For the QuantStudio 7 Flex and CFX384, the assay was tested with 3 ng gDNA from 90 individuals and a reaction volume of 5 µL. The same gDNA samples were tested on the IntelliQube with 1.6 ng per reaction and reaction volume of 1.6 µL. For the Biomark HD, the assay was tested with 23 Coriell gDNA samples from a subset of the 90 ADME SNP gDNA panel samples, each at 150 ng per inlet. As noted in the methods, the thermal mix protocol is modified to achieve maximum assay sensitivity and specificity. rhAmp SNP assays perform well on these four major SNP genotyping platforms, with similar cluster angles and concordant genotyping calls achieved across all samples tested.

A robust SNP genotype call requires relatively high fluorescent signal, large cluster angle separation and tight clusters. In a study of 18 target SNPs, rhAmp SNP assays generated, on average, at least two-fold higher signal than TaqMan for both alleles (Supplemental Figure S1). The allelic discrimination plot and raw fluorescence of an ADME SNP rs776746 assayed by rhAmp SNP and TaqMan genotyping chemistries was compared (Figure 6). The rhAmp SNP assay achieves higher signal for all samples (Figure 6(A)), and more uniform signal for both alleles in heterozygote samples (Figure 6(B)), resulting in a heterozygote cluster angle closer to the ideal 45 degrees.