Table 3 Training and testing accuracy of matching proteome to genome for SomaScan 5K data.
From: Large scale proteomic studies create novel privacy considerations
Cohort | Training | Testing | ||||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
COPDGene (N = 2646 genomes) | COPDGene (N = 9970 genomes) | ARIC (N = 12,219 genomes) | ||||||||||||||||
Ancestry | European American (N = 1877 proteomes) | African American (N = 769 proteomes) | European American (N = 1870 proteomes) | African American (N = 776 proteomes) | European American (N = 8987 proteomes) | African American (N = 2774 proteomes) | ||||||||||||
# Proteins | Top 1 (%) | In top 3 (%) | In top 1% (%) | Top 1 (%) | In top 3 (%) | In top 1% (%) | Top 1 (%) | In top 3 (%) | In top 1% (%) | Top 1 (%) | In top 3 (%) | In top 1% (%) | Top 1 (%) | In top 3 (%) | In top 1% (%) | Top 1 (%) | In top 3 (%) | In top 1% (%) |
20 | 85.56 | 93.61 | 99.15 | 60.73 | 76.20 | 96.62 | 83.90 | 92.09 | 98.66 | 60.05 | 77.32 | 97.55 | 52.77 | 70.54 | 96.44 | 35.63 | 52.34 | 80.52 |
40 | 99.04 | 99.63 | 99.89 | 94.93 | 97.66 | 99.48 | 97.97 | 98.93 | 99.63 | 94.59 | 97.68 | 99.74 | 94.08 | 97.28 | 99.71 | 86.87 | 94.41 | 99.24 |
60 | 99.52 | 99.79 | 99.89 | 97.92 | 98.83 | 99.48 | 98.72 | 99.30 | 99.63 | 97.29 | 98.84 | 99.74 | 97.36 | 98.88 | 99.78 | 94.27 | 97.56 | 99.76 |
100 | 99.79 | 99.89 | 99.89 | 98.83 | 99.09 | 99.48 | 99.36 | 99.52 | 99.63 | 98.45 | 99.23 | 99.74 | 98.83 | 99.49 | 99.80 | 96.75 | 98.81 | 99.95 |
150 | 99.84 | 99.89 | 99.89 | 99.09 | 99.22 | 99.48 | 99.47 | 99.63 | 99.63 | 98.84 | 99.48 | 99.87 | 99.05 | 99.53 | 99.80 | 97.61 | 98.90 | 99.86 |
All | 96.27 | 96.86 | 98.61 | 98.83 | 99.22 | 99.61 | 97.97 | 98.93 | 99.63 | 94.59 | 97.68 | 99.74 | 99.02 | 99.63 | 99.80 | 97.13 | 98.71 | 99.81 |