Table 3.

Mutations detected in constructed reference samples by four commercially available ctDNA detection platforms.

Input DNA (ng)	Mutant allele frequency (%)	Mutant copies per analysisa	ddPCR		Idylla		COBAS z480b		BEAMingc
			All mutations (%)	KRAS p.G12/G13 (%)	All mutations (%)	KRAS p.G12/G13 (%)	All mutations (%)	KRAS p.G12/G13 (%)	All mutations (%)	KRAS p.G12/G13 (%)
10	0.00d	0	nde	nd	nd	nd	nd	nd	nd	nd
10	0.04	1	nd	nd	nd	nd	nd	nd	4f	12
10	0.50	15	43	100	11	16	nd	nd	54	62
50	0.00d	0	nd	nd	nd	nd	12g	nd	nd	nd
50	0.02	3	25	58	4	8	17	nd	8	25
50	0.50	75	43	100	36	50	62	38	92	75
Overall sensitivity (%)			28	65	13	19	20	10	39	44
Sensitivity 50 ng (%)			34	79	20	29	40	19	50	50
Sensitivity 10 ng (%)			22	50	6	8	0	0	27	37

^aAverage number of mutant copies per analysis was calculated as Input DNA (ng) * 300 (Genome Equivalents/ng) * Mutant allele frequency (Mutant copies/Genome Equivalent).

^bFor COBAS z480, invalid results were obtained at 10 ng DNA input for all replicates, at 50 ng of DNA input invalid results were obtained in 25% of the replicates at 0 and 75 mutant copies. Invalid results were counted as not detected.

^cFor BEAMing, invalid results were obtained at 10 ng DNA input in 88%, 54% and 46% of the replicates at 0, 1 and 15 mutant copies respectively. At 50 ng input 7% of all replicates were reported as invalid. Invalid results were counted as not detected.

^dWildtype control samples without synthetic mutant fragment spike-in.

^end = not detected.

^fDetected mutations divided by the total number of mutations present over four replicates. Not all platforms target all mutations, and will have lower reported sensitivity as a result.

^gA false positive KRAS A59X was reported in all wildtype replicates. These false positives was based on a software error (personal communication with Roche Diagnostics).