1. RainDrop assay using 4 μL of genomic DNA Concentration of amplifiable DNA ≥ 50 ng? Copies/assay ≥ 10? T790M positive T790M negative reRainDrop assay using 8 μL of genomic DNA Yes No Yes No Copies/assay ≥ 10? T790M positive Yes T790M negative No Flowchart for mutation call. If the measured event in the T790M gate was ≥ 10 events/assay, the assay was considered to be “positive”. If the event within a gated region was < 10 events/assay, the assay was considered to be “negative”. An assay was inconclusive if the amount of amplifiable DNA was less than 50 ng, at which point then a further assay was performed using 8 μL of DNA (double the volume of the initial assay). Supplementary Figure S1Watanabe et al.

2. Supplementary Figure S2 Quantification of performance of ddPCR analysis for EGFR T790M mutation. Plasmid containing the EGFR T790M mutation (MT plasmid: 2,000, 200, 20, or 0 copy) were serially diluted in 200,000 copies of wild-type EGFR sequence- containing plasmid (WT plasmid). Plasmids were encapsulated into droplets and subjected to the procedure described in Fig. 1. (A) Two-dimensional histogram of ddPCR assay. FAM, 6-carboxyfluorescein; TET, tetrachlorofluorescein. (B) After ddPCR assay, the copy numbers of spiked T790M mutant plasmid were counted using RainDrop Analyst software. Each filled circle represents an individual data point (n = 3). The dotted lines above and below the regression line (straight black line) display the 95% confidence interval. (C) Identical serial dilutions ranging from 0.01–1.00% T790M mutation copies per reaction were assayed in triplicates. The correlation coefficient (R 2 ) is given on the graph. Data shown here are representative of two independent experiments for each assay. WT plasmid: 2 × 10 5 2 × 10 5 2 × 10 5 2 × 10 5 MT plasmid: 2 × 10 3 2 × 10 2 2 × 10 1 - EGFR_wt-TET EGFR_T790M-FAM T790M WT T790M WT T790M WT T790M WT A BC Observed copy# (copies/sample) Expected copy# (copies/sample) Observed mutant% Expected mutant% Watanabe et al.

3. Supplementary Figure S3 Analytical sensitivity of 0.001% for the ddPCR assay. Twenty (2 × 10 1 ) copies of mutant plasmid were spiked into 2 × 10 6 copies of wild-type plasmid, resulting in 0.001% of the mutational fractional abundance. Spiked and non-spiked samples analyzed using the ddPCR assay. (A) Two-dimensional histogram of the non-spiked (left) and the spiked (right) assays. (B) Table showing wild-type and mutant events in both samples. (C) Mutant event is shown in box plot for spiked and non-spiked samples. Box indicates the range of the 95% confidence interval, and the line inside the box indicates the mean. Lines above and below the box denote maximum and minimum events, respectively. A BC WT plasmid: 2 × 10 6 2 × 10 6 MT plasmid: - 2 × 10 1 EGFR_wt-TET T790M WT EGFR_T790M-FAM T790M WT WT plasmid: 2 ×10 6 MT plasmid: - WT plasmid: 2 ×10 6 MT plasmid: 2 ×10 1 Wild-type copies T790M copies Wild-type copies T790M copies #11,996,5154.61,995,75623.6 #21,996,5898.51,996,19613.7 #31,995,7656.31,996,60915.8 #42,254,3539.82,167,05118.2 #52,214,3624.42,242,65316.2 #62,237,7157.22,306,53619.5 Mean6.817.8 95% CI1.52.4 WT plasmid: 2 × 10 6 2 × 10 6 MT plasmid: - 2 × 10 1 T790M copies/assay p < 0.01 (Wilcoxon test) Watanabe et al.

4. EGFR_wt-TET EGFR_T790M -FAM Wild-type control plasmid Human genomic DNA T790M WT T790M WT A549 genomic DNA T790M WT Determination of false-positive events from wild-type control DNA and normal human genomic DNA. Two- dimensional histogram of ddPCR assay with 2 × 10 5 wild-type control plasmid DNA (left), 400 ng human genomic DNA (middle), and 400 ng A549 genomic DNA (right). Supplementary Figure S4Watanabe et al.

5. InputA549 genomic DNA from FFPE cell block (400 ng) Sample # Amplifiable DNA (ng) Wild-type events False positive events #1102.063162.52.8 #2107.366417.53.1 #3106.165719.94.3 #4104.764840.02.5 #5104.864871.92.7 #6103.263892.82.7 #7103.764207.96.1 #8103.764214.52.6 Event 26.8 Mean 3.35 Std. 1.3 95% Upper Limit (One-tail Poisson Dist.) 7 Watanabe et al. EGFR_wt-TETEGFR_T790M -FAM T790M WT Determination of false-positive events in genomic DNA from formalin-fixed paraffin-embedded (FFPE) A549 cells. (A) Two-dimensional histogram of ddPCR assay with 400 ng genomic DNA form FFPE A549 cell block. (B) Determination of the limit of blank (LOB) from the 95% confidence interval of the Poisson model fit. Supplementary Figure S5 AB A549 gDNA from FFPE cell block

6. EGFR_wt-TET EGFR_T790M-FAM JME-013 140.8 ng JME-082 156.8 ng JME- 695 72.8 ng JME- 823 151.2 ng JME- 885 216.4 ng 98.8% 98.7% 99.3% 98.9% 98.5% T790M WT T790M WT T790M WT T790M WT T790M WT Supplementary Figure S6Watanabe et al. Detection of EGFR T790M mutant alleles in patients with EGFR T790M-containing primary tumors. Two- dimensional histogram of ddPCR assay with genomic DNA from FFPE samples. Sample ID and amount of input DNA are indicated on plots. Percentages of empty drops are also displayed on the plot, indicating that > 98% of drops are empty.

7. Operator 2 (% of MT allele) Operator 1 (% of MT allele) Supplementary Figure S7Watanabe et al. Comparison of T790M mutation detection in 6 samples by different operators. Operator-to-operator variation of T790M mutant events in DNA samples from 6 T790M-positive cases. The slope and correlation coefficient are given in the graph.

8. Operator 2 (% of MT allele) Operator 1 (% of MT allele) Supplementary Figure S8Watanabe et al. Reproducibility of ddPCR analysis from 16 T790M-positive samples. ddPCR assay was performed on different days by two different operators to confirm the reproducibility of the percentage of mutant allele. The slope and correlation coefficient are given in the graph.