Sensitivity: True positives (false negatives + true positives)*100
Specificity: True negatives (false positives + true negatives) *100
Category | Specification |
Sensitivity of new clone identification (30Mb) |
Copy loss >99% Copy gain ~75% |
Specificity of new clone identification at 30Mb |
Copy loss >99% Copy gain >99% |
Sensitivity of CNV calls in a defined clone at 5-15Mb | 97.9% |
Specificity of CNV calls in a defined clone at 5-15Mb | 97.0% |
Only high completeness (>8 reads in 60% of the amplicons) cells are used for calling CNV. Therefore the sequencing depth, panel uniformity, and number of cells called can affect the CNV calling performance.
For CNVs at 30Mb resolution, the sensitivity (75% for copy gain and >99% for copy loss) and specificity (>99% for gain & loss) of finding a subclone with CNV is based on:
- The size of the CNV event (10Mb vs 30Mb)
- The size of the clone (e.g., 5% vs 50%)
- The type of CNV in that clone (gain vs loss)
- The presence of heterozygous variants on the amplicons with CNVs (LoH vs only a gain/loss)
The sensitivity (97.9%) and specificity (97%) of calling CNV for each region once the clone has been identified is based on:
- The size of the clone (5% vs 50%)
- The type of CNV event (gain vs loss)
- The presence of heterozygous variants on the amplicons
Limit of detection (LoD): smallest number of cells needed to detect a copy loss or gain.
LoD depends on the number of cells containing the CNV, the type (gain vs loss), and the number of amplicons encompassed by the CNV. The sensitivity of detecting the clone increases drastically as the number of amplicons included in the CNV increases (resolution decreases).
A clone with A CNV of 30 Mb (6 amplicons) has a LoD of >50 cells (loss) and >350 cells (gain) at 75% and >99% sensitivity, respectively.