Table 1 |.
The cancer sequence data used to develop machine learning models included a variety of different tumor subtypes, sequencing approaches and manual review calls
| Variants | |||
|---|---|---|---|
| Training set | Hold out test set | Total | |
| Malignancy | |||
| Leukemia (n = 243) | 5,815 | 2,877 | 8,692 |
| Lymphoma (n = 23) | 8,692 | 628 | 1,891 |
| Breast (n = 135) | 8,986 | 4,320 | 13,306 |
| Small-cell lung (n = 18) | 9,177 | 4,601 | 13,778 |
| Glioblastoma (n = 17) | 844 | 412 | 1,256 |
| Melanoma (n = 1) | 185 | 100 | 285 |
| Colorectal (n = 1) | 842 | 419 | 1,261 |
| Gastrointestinal stromal (n = 1) | 70 | 31 | 101 |
| Malignant peripheral nerve sheath (n = 1) | 288 | 142 | 430 |
| Total | 27,470 | 13,530 | 41,000 |
| Sequencing methods | |||
| Capture sequencing | 9,479 | 4,755 | 14,234 |
| Exome sequencing | 9,367 | 4,677 | 14,044 |
| Genome sequencing | 8,264 | 4,098 | 12,722 |
| Variant calls | |||
| Somatic | 12,266 | 6,115 | 18,381 |
| Ambiguous | 7,189 | 3,454 | 10,643 |
| Fail | 5,909 | 2,945 | 8,854 |
| Germline | 2,106 | 1,016 | 3,122 |
The number of cases for each malignancy is given in parentheses.